ARTICLES
Articles
Intensive blood-glucose control with sulphonylureas or insulin
compared with conventional treatment and risk of complications
in patients with type 2 diabetes (UKPDS 33)
UK Prospective Diabetes Study (UKPDS) Group*
Summary
Background Improved blood-glucose control decreases
the progression of diabetic microvascular disease, but
the effect on macrovascular complications is unknown.
There is concern that sulphonylureas may increase
cardiovascular mortality in patients with type 2 diabetes
and that high insulin concentrations may enhance
atheroma formation. We compared the effects of intensive
blood-glucose control with either sulphonylurea or insulin
and conventional treatment on the risk of microvascular
and macrovascular complications in patients with type 2
diabetes in a randomised controlled trial.
Methods 3867 newly diagnosed patients with type 2
diabetes, median age 54 years (IQR 48–60 years), who
after 3 months’ diet treatment had a mean of two fasting
plasma glucose (FPG) concentrations of 6·1–15·0
mmol/L were randomly assigned intensive policy with
a sulphonylurea (chlorpropamide, glibenclamide, or
glipizide) or with insulin, or conventional policy with diet.
The aim in the intensive group was FPG less than 6
mmol/L. In the conventional group, the aim was the best
achievable FPG with diet alone; drugs were added only if
there were hyperglycaemic symptoms or FPG greater than
15 mmol/L. Three aggregate endpoints were used to
assess differences between conventional and intensive
treatment: any diabetes-related endpoint (sudden death,
death from hyperglycaemia or hypoglycaemia, fatal or
non-fatal myocardial infarction, angina, heart failure,
stroke, renal failure, amputation [of at least one digit],
vitreous
haemorrhage,
retinopathy
requiring
photocoagulation, blindness in one eye, or cataract
extraction);
diabetes-related
death
(death
from
myocardial infarction, stroke, peripheral vascular disease,
renal disease, hyperglycaemia or hypoglycaemia, and
sudden death); all-cause mortality. Single clinical
endpoints and surrogate subclinical endpoints were also
assessed. All analyses were by intention to treat and
frequency of hypoglycaemia was also analysed by actual
therapy.
*Study organisation given at end of paper
Correspondence to: Prof Robert Turner, UKPDS Group,
Diabetes Research Laboratories, Radcliffe Infirmary,
Oxford OX2 6HE, UK
THE LANCET • Vol 352 • September 12, 1998
Findings Over 10 years, haemoglobin A1c (HbA1c) was 7·0%
(6·2–8·2) in the intensive group compared with 7·9%
(6·9–8·8) in the conventional group—an 11% reduction.
There was no difference in HbA1c among agents in the
intensive group. Compared with the conventional group,
the risk in the intensive group was 12% lower (95% CI
1–21, p=0·029) for any diabetes-related endpoint; 10%
lower (–11 to 27, p=0·34) for any diabetes-related death;
and 6% lower (–10 to 20, p=0·44) for all-cause mortality.
Most of the risk reduction in the any diabetes-related
aggregate endpoint was due to a 25% risk reduction
(7–40, p=0·0099) in microvascular endpoints, including
the need for retinal photocoagulation. There was no
difference for any of the three aggregate endpoints
between the three intensive agents (chlorpropamide,
glibenclamide, or insulin).
Patients in the intensive group had more
hypoglycaemic episodes than those in the conventional
group on both types of analysis (both p<0·0001). The
rates of major hypoglycaemic episodes per year were
0·7%
with
conventional
treatment,
1·0%
with
chlorpropamide, 1·4% with glibenclamide, and 1·8% with
insulin. Weight gain was significantly higher in the
intensive group (mean 2·9 kg) than in the conventional
group (p<0·001), and patients assigned insulin had a
greater gain in weight (4·0 kg) than those assigned
chlorpropamide (2·6 kg) or glibenclamide (1·7 kg).
Interpretation Intensive blood-glucose control by either
sulphonylureas or insulin substantially decreases the risk
of microvascular complications, but not macrovascular
disease, in patients with type 2 diabetes. None of the
individual drugs had an adverse effect on cardiovascular
outcomes. All intensive treatment increased the risk of
hypoglycaemia.
Lancet 1998; 352: 837–53
See Commentary page xxx
Introduction
Started in 1977, the UK Prospective Diabetes Study
(UKPDS) was designed to establish whether, in patients
with type 2 diabetes, intensive blood-glucose control
reduced the risk of macrovascular or microvascular
complications, and whether any particular therapy was
advantageous.
Most intervention studies have assessed microvascular
disease: improved glucose control has delayed the
837
ARTICLES
Figure 1: Trial profile
development
and
progression
of
retinopathy,
nephropathy, and neuropathy in patients with type 1
diabetes1,2 and those with type 2 diabetes.3 In the UK,
9% of patients with type 2 diabetes develop
microvascular disease within 9 years of diagnosis, but
20% have a macrovascular complication—and
macrovascular disease accounts for 59% of deaths in
these patients.4
Epidemiological studies of the general population have
shown an increased risk of cardiovascular disease with
concentrations of fasting glucose or haemoglobin A1c
(HbA1c) just above the normal range.5,6 The only
previous large-scale randomised trial in type 2 diabetes,
the University Group Diabetes Program (UGDP),7
followed 1000 patients assigned different therapies for
about 5·5 years (range 3–8 years) and found no evidence
that improved glucose control, by any therapy, reduced
the risk of cardiovascular endpoints. That study did,
however, report increased risk of cardiovascular
mortality in patients allocated the sulphonylurea,
tolbutamide, and this unexpected finding introduced
new hypotheses.8 These hypotheses included increased
myocardial damage from inhibition of ATP-K+ channel
opening in the presence of myocardial ischaemia9 due to
sulphonylurea binding to the cardiovascular SUR2
receptor—an event that could also increase the
likelihood of ventricular arrhythmia.10 An increase in
atherosclerosis with insulin treatment has also been
suggested, since plasma insulin concentrations are
supraphysiological.11,12
838
We report the final results of our study of intensive
blood-glucose control policy, with sulphonylurea or
insulin therapy, compared with conventional treatment
policy with diet, on the risk of microvascular and
macrovascular
clinical
complications.
We
also
investigated whether there was any particular benefit or
risk with sulphonylurea or insulin therapy.
Methods
Patients
Between 1977 and 1991, general practitioners in the catchment
areas of the 23 participating UKPDS hospitals were asked to
refer all patients with newly diagnosed diabetes aged 25–65
years. Patients generally attended a UKPDS clinic within 2
weeks of referral. Patients who had a fasting plasma glucose
(FPG) greater than 6 mmol/L on two mornings, 1–3 weeks
apart, were eligible for the study. An FPG of 6 mmol/L was
selected because this was just above the upper limit of normal
for our reference range. The exclusion criteria were: ketonuria
more than 3 mmol/L; serum creatinine greater than
175 µmol/L; myocardial infarction in the previous year; current
angina or heart failure; more than one major vascular event;
retinopathy requiring laser treatment; malignant hypertension;
uncorrected endocrine disorder; occupation that precluded
insulin therapy (eg, driver of heavy goods vehicle); severe
concurrent illness that would limit life or require extensive
systemic
treatment;
inadequate
understanding;
and
unwillingness to enter the study.
7616 patients were referred and 5102 were recruited (58%
male). The 2514 patients excluded were similar in age, sex, and
glycaemic status to those recruited. The study design and
protocol amendments, which conform with the guidelines of the
THE LANCET • Vol 352 • September 12, 1998
ARTICLES
Conventional (n=1138)
Intensive (n=2729)
All patients (n=3867)
Demographic
Age (years)*
M/F
Ethnicity (%) Caucasian/Indian Asian/Afro-Caribbean/Other
53·4 (8·6)
705/433
81/11/7/1
53·2 (8·6)
649/444
81/10/8/1
53·3 (8·6)
2359/1508
81/10/8/1
Clinical
Weight (kg)*
Body-mass index (kg/m2)*
Systolic blood pressure (mm Hg)*
Diastolic blood pressure (mm Hg)*
Smoking (%) never/ex/current
Alcohol (%) none/social/regular/dependent
Exercise (%) sedentary/moderately active/active/fit
78·1 (16·3)
27·8 (5·5)
135 (19)
82 (10)
34/35/31
26/56/18/2
20/37/39/4
77·3 (15·4)
27·5 (5·1)
135 (20)
83 (10)
35/35/30
24/56/17/1
21/34/40/5
Biochemical
FPG (mmol/L)†
HBA1c (%)*
Plasma insulin (pmol/L)‡
Triglyceride (mmol/L)‡
Total cholesterol (mmol/L)*
LDL-cholesterol (mmol/L)*
HDL-cholesterol (mmol/L)*
8·0 (7·1–9·6)
7·05 (1·42)
91 (52–159)
2·31 (0·84–6·35)
5·4 (1·02)
3·5 (0·99)
1·08 (0·24)
77·5 (15·5)
27·5 (5·2)
135 (20)
82 (10)
34/35/31
22/56/18/1
20/35/40/5
8·1 (7·1–9·8)
7·09 (1·54)
92 (52–159)
2·37 (0·85–6·63)
5·4 (1·12)
3·5 (1·0)
1·07 (0·25)
8·0 (7·1–9·7)
7·08 (1·51)
92 (52–160)
2·35 (0·84–6·55)
5·4 (1·1)
3·5 (1·0)
1·07 (0·24)
Medications
More than one asprin daily (%)
Diuretic (%)
Others (%) digoxin/antihypertensive/lipid lowering/HRT or OC
1·5
13
0·9/12/0·3/0·9
1·7
13
1·3/12/0·3/0·7
1·6
13
1·1/12/0·3/0·8
Surrogate clinical endpoints
Retinopathy (%)
Proteinuria (%)
Plasma creatinine (mmol/L)‡
Biothesiometer more than 25 volts (%)
36
2·1
81 (66–99)
11·4
36
1·7
82 (67–100)
11·8
36
1·9
81 (67–100)
11·5
Data are % of group, *mean (SD), †median (IQR), or ‡geometric mean (1 SD). HRT=hormone replacement therapy. OC=oral contraceptive therapy.
Table 1: Baseline characteristics of patients in conventional and intensive-treatment groups
Declarations of Helsinki (1975 and 1983), were approved by
the Central Oxford Research Ethics Committee and by the
equivalent committees at each centre. Each patient gave
informed witnessed consent.
about 50% of calories from carbohydrates; overweight patients
were advised to reduce energy content.13 After the run-in, a
mean FPG was calculated from measurements on 3 days over 2
weeks.
Dietary run-in
Definitions
Patients had a 3-month dietary run-in during which they
attended a monthly UKPDS clinic and were seen by a
physician and dietician. The patients were advised to follow
diets that were low saturated fat, moderately high fibre and had
Marked hyperglycaemia was defined as FPG greater than 15
mmol/L, symptoms of hyperglycaemia, or both, in the absence
of intercurrent illness. Hyperglycaemic symptoms included
thirst and polyuria.
Conventional
(n=896)
Chlorpropamide
(n=619)
Glibenclamide
(n=615)
Insulin
(n=911)
All patients
(n=3041)
Demographic
Age (years)*
M/F
Ethnicity (%) Caucasian/Indian Asian/Afro Caribbean/Other
54 (9)
555/341
83/9/7/1
54 (9)
359/260
79/10/11/0
54 (8)
381/234
84/8/7/1
54 (8)
656/346
82/8/9/1
54 (8)
1885/1156
82/8/9/1
Clinical
Weight (kg)*
Body-mass index (kg/m2)*
Systolic blood pressure (mm Hg)*
Diastolic blood pressure (mm Hg)*
Smoking (%) never/ex/current
Alcohol (%) none/social/regular/dependent
Exercise (%) sedentary/moderately active/active/fit
77 (16)
27·5 (5·3)
136 (19)
83 (10)
34/34/32
24/55/20/1
18/38/40/4
75 (15)
27·0 (4·9)
136 (19)
83 (10)
38/31/31
26/52/21/1
19/37/40/4
77 (14)
27·4 (5·0)
136 (19)
83 (10)
32/38/30
22/58/19/1
18/32/44/6
76 (14)
27·0 (4·8)
136 (20)
83 (11)
34/36/30
24/57/18/1
21/35/40/4
76 (15)
27·2 (5·0)
136 (19)
83 (10)
35/35/30
24/57/18/1
19/36/41/4
Biochemical
FPG (mmol/L)†
HBA1c (%)*
Plasma insulin (pmol/L)‡
Triglyceride (mmol/L)‡
Total cholesterol (mmol/L)*
LDL-cholesterol (mmol/L)*
HDL-cholesterol (mmol/L)*
Medications
More than one asprin daily (%)
Diuretic (%)
Others (%) digoxin/antihypertensive/lipid lowering/
HRT or OC
Surrogate clinical endpoints
Retinopathy (%)
Proteinuria (%)
Plasma creatinine (mmol/L)‡
Biothesiometer more than 25 volts (%)
7·9 (7·1–9·4)
6·2 (1·2)
89 (51–156)
2·43 (0·86–6·92)
5·4 (1·03)
3·5 (0·99)
1·07 (0·23)
8·0 (7·1–9·7)
6·3 (1·4)
90 (51–160)
2·58 (0·88–7·55)
5·5 (1·15)
3·5 (1·05)
1·08 (0·25)
8·0 (7·2–9·6)
6·3 (1·3)
91 (52–160)
2·37 (0·84–6·72)
5·5 (1·11)
3·5 (1·00)
1·09 (0·25)
8·1 (7·1–9·9)
6·1 (1·1)
90 (52–156)
2·48 (0·85–7·25)
5·4 (1·13)
3·5 (1·03)
1·07 (0·25)
8·0 (7·1–9·6)
6·2 (1·2)
90 (52–156)
2·46 (0·86–7·10)
5·4 (1·10)
3·5 (1·02)
1·08 (0·24)
1·2
13
0·5/12·2/0·1/0·3
1·5
12
1·0/11·2/0·3/0·3
1·1
15
1·3/11·3/0/0·5
1·8
14
1·3/10·7/0·2/0·7
1·4
14
1·0/11·6/0·3/0·5
38
2·2
80 (66–97)
12·1
40
1·7
81 (67–82)
10·1
30
2·1
82 (67–99)
15·2
38
1·5
81 (67–99)
12·1
38
1·9
81 (67–99)
12·3
Data are % of group, *mean (SD), †median (IQR), or ‡geometric mean (1 SD). HRT=hormone replacement therapy. OC=oral contraceptive therapy.
Table 2: Baseline characteristics of patients in conventional group and individual intensive groups
THE LANCET • Vol 352 • September 12, 1998
839
ARTICLES
Assigned therapy in 15 centres (32 406 person-years)
Assigned therapy in all 23 centres
(38 263 person-years)
Conventional
(n=896)
Chlorpropamide
(n=619)
Glibenclamide
(n=615)
Insulin
(n=911)
Total person-years
9491
6562
6573
Actual therapy (person years)
Diet alone
Chlorpropamide alone or in combination
Glibenclamide alone or in combination
Glipizide alone or in combination
Metformin alone or in combination
Insulin
5495 (58%)
621 (7%)
1699 (18%)
47 (0·5%)
1105 (12%)
1458 (15%)
409 (6%)
5266 (80%)
483 (7%)
28 (0·4%)
900 (14%)
615 (9%)
432 (7%)
126 (2%)
5467 (83%)
17 (0·3%)
1319 (20%)
681 (10%)
Conventional
(n=1138)
Intensive
(n=2729)
9780
11 188
27 075
1896 (19%)
66 (1%)
823 (8%)
58 (1%)
329 (3%)
7215 (74%)
6490 (58%)
743 (7%)
1715 (15%)
281 (3%)
1132 (10%)
1809 (16%)
3206 (12%)
6372 (24%)
6789 (25%)
1359 (5%)
2581 (10%)
10 413 (38%)
Table 3: Person-years of follow-up on assigned and actual therapies for first 15 and all centres
Randomisation
The flow of patients in the study is shown in figure 1.
Patients were stratified by ideal bodyweight (overweight was
>120% ideal bodyweight).14 Non-overweight patients were
randomly assigned intensive treatment with insulin (30%),
intensive treatment with sulphonylurea (40%: equal
proportions in the first 15 centres to chlorpropamide or
glibenclamide, and in the last eight centres to chorpropamide or
glipizide), or conventional treatment with diet (30%). The
non-balanced randomisation was chosen so that there were
sufficient patients in the two sulphonylurea groups to allow
comparison between the first-generation and second-generation
drugs. Overweight patients were randomly assigned treatment
with the additional possibility of metformin: intensive treatment
with insulin (24%), intensive treatment with sulphonylurea with
equal proportions of patients on chlorpropamide and
glibenclamide (32%), intensive treatment with metformin
(20%), and conventional treatment with diet (24%). The 342
overweight patients who were randomly allocated metformin
therapy are reported separately, as intended per protocol.15
Randomisation was by means of centrally produced,
computer-generated therapy allocations in sealed, opaque
envelopes which were opened in sequence. The numerical
sequence of envelopes used, the dates they were opened, and
the therapies stipulated were monitored. The trial was open
once patients were randomised. No placebo treatments were
given.
Conventional treatment policy
concentrations, a letter was sent from the coordinating center
with advice on necessary changes in therapy. Patients assigned
insulin started on once daily ultralente insulin (Ultratard HM,
Novo-Nordisk, Crawley, UK or Humulin Zn, Eli-Lilly,
Basingstoke, UK) or isophane insulin. If the daily dose was
more than 14 units (U) or pre-meal or bed-time home bloodglucose measurements were more than 7 mmol/L, a
short-acting insulin, usually soluble (regular) insulin was
added—ie, basal/bolus regimen. Patients on more than 14 U
insulin per day, or on short-acting insulins, were particularly
encouraged to do regular home-glucose monitoring.
Protocol and amendments
The original protocol for the first 15 centres stipulated that
patients
continue
their
assigned
treatment
(diet,
chlorpropamide, glibenclamide, metformin, or insulin) for as
long as possible to achieve maximum exposure to each therapy
alone and thus find out whether there were differences in
response to each agent. Additional therapies were added to
those allocated to diet, sulphonylurea, or metformin only when
marked hyperglycaemia developed. For patients on
sulphonylureas, metformin was added; but if marked
hyperglycaemia recurred, patients were changed to insulin
therapy. Metformin was used to a maximum of 2550 mg per
day.
When the progressive hyperglycaemia in all groups became
apparent, the protocol was amended to allow the early addition
of metformin when, on maximum doses of sulphonylurea, FPG
was greater than 6 mmol/L in symptomless patients in the
intensive group. Patients were changed to insulin therapy if
marked hyperglycaemia recurred.
When the last eight centres were recruited in 1988, patients
allocated sulphonylurea had insulin added early, rather than
metformin, when on maximum doses of sulphonylurea FPG
was greater than 6 mmol/L.
The aim in this group was to maintain FPG below 15 mmol/L
without symptoms of hyperglycaemia. Patients attended
UKPDS clinics every 3 months and received dietary advice
from a dietician with the aim of maintaining near-normal
bodyweight.
If marked hyperglycaemia or symptoms occurred, patients
were secondarily randomised to treatment with sulphonylurea
or insulin therapy, with the additional option of metformin in
overweight patients; this was a separate stratified randomisation
from the original randomisation, but with the same proportions
allocated sulphonylurea and insulin.13 If marked hyperglycaemia
recurred in participants secondarily allocated sulphonylurea,
metformin was added, and in those secondarily allocated
metformin, glibenclamide was added. Patients with marked
hyperglycaemia or symptoms on both agents were changed to
insulin. Throughout, the aim of FPG below 15 mmol/L without
symptoms was maintained. Clinical centres were advised by
automatically generated letters when patients allocated
conventional treatment received inappropriate pharmacological
therapy.
1148 UKPDS patients were in the Hypertension in Diabetes
Study (HDS).16 This study, which started in 1987, randomly
allocated hypertensive patients to a tight blood-pressure-control
treatment that aimed for a blood pressure of 150/85 mm Hg or
lower with either captopril or atenolol or, to a less tight bloodpressure-control treatment that aimed for a blood pressure of
180/105 mm Hg or lower but avoided the use of captopril and
atenolol. The UKPDS Acarbose Study17 started in 1994 and
randomly allocated 1946 patients to additional double-blind,
placebo-controlled therapy with acarbose for 3 years—
irrespective of their blood-glucose and blood-pressure control
allocations.
Intensive treatment policy
Clinic visits
The aim of intensive treatment was FPG less than 6 mmol/L
and,
in
insulin-treated
patients,
pre-meal
glucose
concentrations of 4–7 mmol/L. These patients also continued
to receive dietary advice from a dietician. The daily doses of the
sulphonylureas used were: chlorpropamide 100–500 mg;
glibenclamide 2·5–20 mg; and glipizide 2·5–40 mg.
Whenever glucose concentrations were above target
Patients attended morning clinics every 3 months or more
frequently as needed to attain glycaemic control. From 1990,
the routine clinic visits were every 4 months. Patients fasted
from 2200 h the night before for plasma glucose and other
biochemical measurements, and did not take their allocated
treatment on the morning of the clinic visit.
At each visit plasma glucose, blood pressure, and weight
840
Embedded studies
THE LANCET • Vol 352 • September 12, 1998
ARTICLES
Figure 2: Cross-sectional and 10-year cohort data for FPG, HbA1c, weight, and fasting plasma insulin in patients on intensive or
conventional treatment
were measured, and therapy was adjusted if necessary. From
a checklist we asked about all medications, hypoglycaemic
episodes, home blood-glucose measurements, illness, time
off work, admissions to hospital, general symptoms including
any drug side-effects, and clinical events. Hypoglycaemic
episodes were defined as minor if the patient was able to
treat the symptoms unaided, or major if third-party help or
medical intervention was necessary. Details of all major
hypoglycaemic episodes were audited to ensure the coding was
appropriate.
At entry, randomisation, 6 months, 1 year, and annually
thereafter a fasting blood sample was taken for measurement of
HbA1c, plasma creatinine (annually from 1989), triglyceride,
total cholesterol, LDL-cholesterol, HDL-cholesterol, insulin,
and insulin antibodies. Every year, urinary albumin and
creatinine were measured in a random urine sample.
At entry and then every 3 years all patients had a full clinical
examination. At these reviews, a 12-lead electrocardiogram was
recorded and Minnesota coded13 and a posterior-anterior chest
radiograph taken for measurement of cardiac diameter. Doppler
blood pressure was measured in both legs and in the right arm.
Visual acuity was measured with a Snellen chart until 1989 and
subsequently with an Early Treatment of Diabetic Retinopathy
Study (ETDRS) chart.13 The best attainable vision was assessed
with the patient’s usual spectacles or with a pinhole. Direct
ophthalmoscopy with pupil dilation was carried out every
THE LANCET • Vol 352 • September 12, 1998
3 years. Since 1982, retinal colour 30º photographs of four
fields per eye (nasal, disc, macula, and temporal-to-macula
fields) were taken with additional stereo photographs of the
macula; poor quality photographs were repeated. Two assessors
at a single centre reviewed the photographs for diabetic
retinopathy; any fields with retinopathy were graded by two
other assessors by a modified ETDRS final scale.13
Neuropathy was assessed clinically by knee and ankle reflexes
and by biothesiometer (Biomedical Instruments Co, Newbury,
OH, USA) readings at the lateral malleolus and at the end of
the big toe.13 Autonomic neuropathy was assessed by: R-R
intervals measured on electrocardiograms at expiration and
inspiration on deep breathing for five cycles; change in R-R
interval on standing; basal heart rate during deep breathing;
lying and standing blood pressure; and, in men, self-reported
erectile dysfunction. These assessments, including visual
acuity, grading of photographs, and Minnesota coding, were
carried out by staff from whom the allocations and actual
therapies were concealed.
Biochemistry
Methods have been reported previously.18 Plasma glucose
analysers were monitored monthly in each clinical centre by
the UKPDS Glucose Quality Assurance Scheme; the
mean interlaboratory imprecision was 4% and values were
841
ARTICLES
Figure 3: Cross-sectional and 10-year cohort data for FPG, HbA1c, weight, and fasting plasma insulin in patients on chlorpropamide,
glibenclamide, or insulin, or conventional treatment
within 0·1 mmol/L of those obtained by UK External Quality
Assessment Scheme. Plasma creatinine, urea, and urate were
measured in the clinical chemistry laboratories at the clinical
centres. Blood, plasma and urine samples were transported
overnight at 4°C to the central biochemistry laboratory for all
other measurements. HbA1c was measured by high-performance
liquid
chromatography
(Biorad
Diamat
Automated
Glycosylated Haemoglobin Analyser, Hemel Hempstead, UK),
and the normal range is 4·5–6·2%.18 By comparison with the
US National Glycohemoglobin Standardization Program,
HbA1c(UKPDS)=1·104
HbA1c(DCCT)–0·7336,
(r=0·99,
n=40). From 1988 urine albumin was measured by an
immunoturbidimetric method (reference range 1·4–36·5
mg/L).18 Microalbuminuria has been defined for this study as a
urinary albumin concentration greater than 50 mg/L due to
initial storage of urine samples at –20°C between 1979 and
1988, and clinical-grade proteinuria as urinary albumin
concentrations greater than 300 mg/L.19 Insulin was measured
by double-antibody radioimmunoassay (Pharmacia RIA 100
842
Pharmacia Upjohn, Milton Keynes, UK) with 100%
cross-reaction to intact proinsulin and 25% to 32/33 split
proinsulin.
Clinical endpoints
21 clinical endpoints were predefined in the study protocol in
198113 and are listed later. Particular disorders were defined:
myocardial infarction by WHO clinical criteria with
electrocardiogram/enzyme changes or new pathological Q wave;
angina by WHO clinical criteria and confirmed by a new
electrocardiogram abnormality or positive exercise test; heart
failure (not associated with myocardial infarction), by clinical
symptoms confirmed by Kerley B lines, râles, raised jugular
venous pressure, or third heart sound; major stroke by
symptoms or signs for 1 month or longer; limb amputation as
amputation of at least one digit; blindness in one eye by WHO
criteria with Snellen-chart visual acuity of 6/60 or worse, or
ETDRS logMAR 1·0 or worse, for 3 months; and renal failure
by dialysis or plasma creatinine greater than 250 µmol/L not
THE LANCET • Vol 352 • September 12, 1998
ARTICLES
Figure 4: Proportion of patients with aggregate and single endpoints by intensive and conventional treatment and relative risks
related to any acute intercurrent illness. The clinical decision
for photocoagulation or cataract extraction was made by
ophthalmologists independent of the trial.
Aggregate endpoints were defined by the Data-Monitoring
and Ethics Committee in 1981 as time to the first occurrence
of: any diabetes-related endpoint (sudden death, death from
hyperglycaemia or hypoglycaemia, fatal or non-fatal myocardial
infarction, angina, heart failure, stroke, renal failure,
amputation [of at least one digit], vitreous haemorrhage, retinal
photocoagulation, blindness in one eye, or cataract extraction);
diabetes-related death (death from myocardial infarction,
stroke,
peripheral
vascular
disease,
renal
disease,
hyperglycaemia or hypoglycaemia, and sudden death); all-cause
mortality. These aggregates were used to assess the difference
between conventional and intensive treatment.
To investigate differences among chlorpropamide, insulin,
and glibenclamide, four additional clinical-endpoint aggregates
were used: myocardial infarction (fatal and non-fatal) and
sudden death; stroke (fatal and non-fatal); amputation or death
due to peripheral vascular disease; and microvascular
complications (retinopathy requiring photocoagulation, vitreous
haemorrhage, and or fatal or non-fatal renal failure).
Surrogate endpoints
Subclinical, surrogate variables were assessed every 3 years.
The criteria were: for neuropathy—loss of both ankle or both
knee reflexes or mean biothesiometer reading from both toes 25
V or greater; for autonomic neuropathy—R-R interval less than
the age-adjusted normal range (a ratio <1·03 of the longest R-R
interval at approximately beat 30 to the shortest at
approximately beat 15); for orthostatic hypotension—systolic
fall of 30 mm Hg or more, or diastolic fall of 10 mm Hg or
more; and for impotence—no ejaculation or erection.
Retinopathy was defined as one microaneurysm or more in one
eye or worse retinopathy, and progression of retinopathy as a
two-step change in grade. Poor visual acuity was: logMAR
more than 0·3 (unable to drive a car), more than 0·7 (US
definition of blindness), and logMAR 1·0 or greater (WHO
definition of blindness). Deterioration of vision was defined as a
THE LANCET • Vol 352 • September 12, 1998
three-line deterioration in reading an ETDRS chart. Ischaemic
heart disease by Minnesota coding was either WHO grade 1
(possible coronary heart disease) or grade 2 (probable coronary
heart
disease).
Left-ventricular
hypertrophy
was
a
cardiothoracic ratio 0·5 or greater.
The study closed on Sept 30, 1997. All available information
for each endpoint, such as admission notes, operation records,
death certificates, and necropsy reports, were gathered. The
file, with no reference to assigned or actual therapy, was
reviewed independently by two physicians who assigned
appropriate International Classification of Disease–9 codes.20
Any disagreements between the two assessors were discussed and
the evidence reviewed; if agreement was not possible the file
was submitted to two different assessors for final arbitration.
Statistical analysis
When the UKPDS started in the late 1970s, it was thought that
improved blood-glucose control might reduce the incidence of
diabetes-related endpoints by 40%. This seemed reasonable
since the risk of cardiovascular events in patients with diabetes
is at least twice that of people with normal glucose tolerance
and microvascular complications do not occur in the
normoglycaemic population. The first three aggregate
endpoints were defined and, for death and major cardiovascular
events (the stopping criteria), the original power calculation to
find a 40% difference between the intensive and conventional
groups was a sample size of 3600 with 81% power at the 1%
level of significance.
However, by 1987 no risk reduction was seen in any of these
aggregates and it became obvious a 40% advantage was unlikely
to be obtained. The publication of other intervention studies of
chronic diseases in the mid 1980s suggested that a more
realistic goal would be a difference of 15%. Accordingly, the
study was extended to include randomisation of 3867 patients
with a median time from randomisation of 11 years to the end
of the study in 1997. In 1992, at the 1% level of significance,
the power for any diabetes-related endpoint and for diabetesrelated death was calculated as 81% and 23%, respectively.
There was the same proportion of patients in the
843
ARTICLES
Figure 5: Proportion of patients with aggregate and single endpoints by individual intensive treatment and conventional treatment
and relative risks
Key as for figure 4.
non-overweight and overweight stratifications assigned
intensive and conventional treatment, and, within the intensive
group, sulphonylurea or insulin treatment, and thus the
non-overweight and overweight patients are analysed together.
The 3867 patients from all 23 centres were included in the
analyses of conventional and intensive treatment.
The analysis among chlorpropamide, glibenclamide, or
insulin in the intensive group used only 3041 patients from the
first 15 centres where patients had remained for longer periods
on monotherapy until marked hyperglycaemia occurred.
Intention-to-treat analysis was used to compare outcomes
between the intensive and conventional treatment groups and
between the patients on conventional treatment and those on
each of the intensive treatment agents.
All analyses of significance were two-sided (2p). Life-table
analyses were done with log-rank tests. Hazard ratios, used to
estimate relative risks, were obtained from Cox proportionalhazards models. In the text, the relative risks are quoted in terms
of risk reduction. For the clinical endpoint aggregates, 95% CI
are quoted. For single endpoints and surrogate variables 99% CI
are given to make allowance for potential type I errors. Mean
(SD), geometric mean (1SD interval), or median (IQR) have
844
been quoted for the biometric and biochemical variables, with
Wilcoxon, t test, or χ2 for comparison tests. Risk reductions for
categorical variables were derived from relative risks obtained
from frequency tables. Survival-function estimates were
calculated by the product-limit (Kaplan-Meier) method. Yearly
averaged data for weight and FPG were calculated as the median
of three consecutive visits for each patient—ie, the annual visit,
and the 3 month visit before and after this. HbA1c data were from
the annual assessment but overall values for HbA1c during a
period were the median for each patient for each allocation.
Glucose control and HbA1c were assessed both cross-sectionally
and in the cohort with 10 years’ follow-up. Urine albumin was
assessed in mg/L with no adjustment for urine creatinine
concentration.21 Data for albuminuria at the triennial visit were
the median of that year and the years before and after.
Hypoglycaemic episodes in each year were analysed both by
intention to treat and by actual therapy.
Safety
The Data-monitoring and Ethics Committee reviewed the
endpoint analyses every 6 months to decide whether to stop or
modify the study according to predetermined guidelines. These
THE LANCET • Vol 352 • September 12, 1998
ARTICLES
Figure 5: Continued
guidelines included a difference of 3 SD or more by log-rank
test in the three aggregate endpoints between intensive and
conventional blood-glucose control groups.13 The stopping
criteria were not attained.
Results
Background and biochemical data
4763 (93%) of 5102 patients had mean FPG of 7·0
mmol/L or more (American Diabetes Association
criteria),22 and 4396 (86%) of 5102 had values greater
than 7·8 mmol/L (WHO criteria).23
Baseline characteristics of the 3867 patients assigned
conventional or intensive treatment are given in table 1.
The baseline characteristics of the 3041 patients in the
comparison of conventional treatment with each of the
three intensive agents are in table 2.
The median follow-up for endpoint analyses was 10·0
years (IQR 7·7–12·4). The median follow-up for the
comparison of conventional treatment with each of the
three intensive agents was 11·1 years (9·0–13·0). The
THE LANCET • Vol 352 • September 12, 1998
percentage of total person-years for which the assigned
or other therapies were taken in the conventional or
intensive groups are shown in table 3.
At the end of the trial, the vital status of 76 (2·0%)
patients who had emigrated was not known; 57 and 19
in intensive and conventional groups, respectively, which
reflects the 70/30 randomisation. A further 91 (2·4%)
patients (65 in the intensive group) could not be
contacted in the last year of the study for assessment of
clinical endpoints. The corresponding numbers for
comparison of the individual intensive agents were 69
(2·7%) emigrated and 63 (2·1%) not contactable.
In the conventional group, the FPG and HbA1c
increased steadily over 10 years from randomisation in
both the cohort study of 461 patients and in the
cross-sectional data at each year (figure 2). In the
intensive group, there was an initial decrease in FPG
and HbA1c in the first year, both in the 10 year cohort of
1180 patients and in the cross-sectional data, with a
subsequent increase similar to that in the conventional
845
ARTICLES
Figure 6: Kaplan-Meier plots of aggregate endpoints: any diabetes-related endpoint and diabetes-related death for conventional or
intensive treatment, and by individual intensive therapy
Key as for figures 3 and 4.
group (figure 2). A difference between the assigned
groups in HbA1c was maintained throughout the study.
The median HbA1c values over 10 years were
significantly lower in the intensive than in the
conventional group (7·0% [6·2–8·2] vs 7·9% [6·9–8·8],
p<0·0001). Median HbA1c for 5-year periods of followup in the intensive and conventional groups were 6·6%
(5·9–7·5) and 7·4% (6·4–8·5) for the first period,
7·5% (6·6–8·8) and 8·4% (7·2–9·4) for the second, and
8·1% (7·0–9·4) and 8·7% (7·5–9·7) for the third period
(all p<0·0001) .
The median HbA1c values over 10 years with
chlorpropamide (6·7%), glibenclamide (7·2%), and
insulin (7·1%) were each significantly lower than that
with conventional treatment (7·9%, p<0·0001). HbA1c
was significantly lower in the chlorpropamide group
than in the glibenclamide group (p=0·008) but neither
differed from the insulin group (figure 3).
There was a significant increase in weight in the
intensive group compared with the conventional group,
by (mean) 3·1 kg (99% CI –0·9 to 7·0, p<0·0001) for
846
the cohort at 10 years (figure 2). Patients assigned either
of the sulphonylureas gained more weight than the
conventional group, whereas patients assigned insulin
gained more weight than those assigned a sulphonylurea
(figure 3). In the cohort at 10 years, those assigned
chlorpropamide gained 2·6 kg more (1·6–3·6, p<0·001);
those assigned glibenclamide gained 1·7 kg more
(0·7–2·7, p<0·001); and those assigned insulin gained
4·0 kg more (3·1–4·9, p<0·0001) than those assigned
conventional therapy (figure 3). The cross-sectional data
were similar to the cohort data.
Median fasting plasma insulin increased in the
intensive group, and was 17·9 pmol/L (95% CI
0·5–35·3) greater than in the conventional group over
the first 10 years (p<0·0001, figure 2). Fasting plasma
insulin in participants assigned to sulphonylureas
increased more than in those in the conventional group
over the first 3 years, and in those assigned to insulin
this increase was even greater from 6 years as higher
insulin doses were given (figure 3).
The median insulin doses at 3 years, 6 years, 9 years,
THE LANCET • Vol 352 • September 12, 1998
ARTICLES
Figure 7: Kaplan-Meier plots of aggregate endpoints: microvascular disease, myocardial infarction, and stroke for intensive and
conventional treatment and by individual intensive therapy
Microvascular disease=renal failure, death from renal failure, retinal photocoagulation, or vitreous haemorrhage. Myocardial infarction=non-fatal, fatal,
or sudden death. Stroke=non-fatal and fatal. Key as for figures 3 and 4.
and 12 years in patients assigned intensive treatment
with insulin were 22 U (IQR 14–34), 28 U (18–45),
34 U (20–50), and 36 U (23–53), respectively. Median
doses of insulin for patients with body-mass indices less
than 25 kg/m2 and greater than 35 kg/m2 were 16 U
(10–24) and 36 U (23–50) at 3 years; the corresponding
doses were 24 U (14–36) and 60 U (40–82) at 12 years.
The maximum insulin dose was 400 U per day.
Systolic and diastolic blood pressure were significantly
higher throughout the study in patients assigned
chlorpropamide than in those assigned any of the other
therapies. For example, at 6 years’ follow-up the mean
blood pressure in the chlorpropamide group was 143/82
mm Hg compared with 138/80 mm Hg in each of the
other allocations (p<0·001). The proportion of patients
on therapy for hypertension was higher among those
assigned chlorpropamide (43%) than among those
assigned conventional treatment, glibenclamide, or
insulin (34%, 36%, and 38%, respectively; p=0·022).
THE LANCET • Vol 352 • September 12, 1998
Aggregate and single endpoints
The number of patients who developed aggregate or
single clinical endpoints in the intensive and
conventional groups are shown in figure 4; similarly,
figure 5 shows the comparison between the three
intensive
groups
and
conventional
treatment.
Kaplan-Meier plots for any diabetes-related endpoint—
ie, the complication-free interval—and diabetes-related
deaths are shown in figure 6 and those for microvascular
endpoints, myocardial infarction, and stroke in figure 7.
The number needed to treat to prevent one patient
developing any of the single endpoints over 10 years was
19·6 patients (95% CI 10–500). The complication-free
interval, expressed as the follow-up to when 50% of the
patients had at least one diabetes-related endpoint, was
14·0 years in the intensive group compared with 12·7
years in the conventional group (p=0·029).
Patients assigned intensive treatment had a significant
25% risk reduction in microvascular endpoints
(p=0·0099) compared with conventional treatment—
847
ARTICLES
Figure 8: Proportion of patients with selected surrogate endpoints at 3-year intervals
most of which was due to fewer cases of retinal
photocoagulation (figure 4): the reduction in risk was of
borderline significance for myocardial infarction
(p=0·052) and cataract extraction (p=0·046).
There was no significant difference between the three
intensive treatments on microvascular endpoints or the
risk reduction for retinal photocoagulation (figure 5).
Few patients developed renal failure, died from renal
disease, or had vitreous haemorrhage.
Surrogate endpoints
Figure 8 shows the proportion of patients with surrogate
endpoints (two-step progression of retinopathy,
biothesiometer threshold, microalbuminuria, proteinuria, and two-fold increase in plasma creatinine) found
at 3-year visits. After 6 years’ follow-up, a smaller
proportion of patients in the intensive group than in the
conventional group had a two-step deterioration in
retinopathy: this finding was significant even when
retinal photocoagulation was excluded (data not shown).
When the three intensive treatments were compared,
patients assigned chlorpropamide did not have the same
risk reduction as those assigned glibenclamide or insulin
(p=0·0056) for the progression of retinopathy at 12
years’ follow-up, and adjustment for the difference in
mean systolic or diastolic blood pressure by logistic
regression analysis did not change this finding.
There was no difference between conventional and
intensive treatments in the deterioration of visual acuity
with a mean ETDRS chart reduction of one letter per
3 years in each group. At 12 years the proportion of
patients blind in both eyes (logMAR>0·7) did not differ
between the intensive and conventional groups (6/734
[0·8%]) vs 5/263 [1·9%], p=0·15). 11% of patients in
both groups did not have adequate vision for a driving
licence (logMAR > 0·3 in both eyes).
Proportions of patients with absent ankle reflexes did
848
not differ between intensive and conventional groups
(35 vs 37%, p=0·60); similar proportions had absent
knee reflexes (11 vs 12%, p=0·42).
The heart-rate responses to deep breathing and
standing did not differ between the intensive and
conventional groups, but at 12 years the basal heart rate
was significantly lower in the intensive than in the
conventional group (median 69·8 [IQR 62·5–78·9] vs
74·4 [65·2–83·3] bpm, p<0·001). β-blockers were taken
by 16% and 19% (p=0·58) of patients in the intensive
and conventional groups.
The proportion of patients with impotence did not
differ at 12 years in the intensive and conventional
groups (46·8 vs 54·7%, respectively; p=0·09).
There was no difference between the intensive and
conventional treatment groups, or between the three
intensive allocations, in the proportion of patients who
had a silent myocardial infarction, cardiomegaly,
evidence of peripheral vascular disease by doppler blood
pressure, or absent peripheral pulses.
Hyperglycaemia and hypoglycaemia
The proportion of patients with one or more major, or
any, hypoglycaemic episode in a year was significantly
higher in the intensive group than in the conventional
group (figure 9). When the three intensive treatments
were compared by actual therapy, major hypoglycaemic
episodes or any episode were most common in patients
on insulin therapy (figure 10). During the first few years
of therapy, any hypoglycaemic episodes were also
frequent
in
patients
on
glibenclamide
or
chlorpropamide, but fell as FPG increased. By
intention-to-treat analysis, there was less difference
between the allocations as more patients in the
conventional group had sulphonylurea or insulin therapy
added. One insulin-group patient died at home,
unattended: this death was attributed to hypoglycaemia.
THE LANCET • Vol 352 • September 12, 1998
ARTICLES
Figure 9: Major and any hypoglycaemic episodes per year by intention-to-treat analysis and actual therapy for intensive and
conventional treatment
Data from the first 15 centres. The numbers of patients studied at 5, 10, and 15 years in the intensive and conventional groups by actual therapy were
1317, 395; 762, 150; and 120, 14 respectively.
In the conventional group, one patient died from
hyperglycaemic, coma after a febrile illness.
Over the first 10 years, the mean proportion of
patients per year with one or more major hypoglycaemic
episodes while taking their assigned intensive or
conventional treatment was 0·4% for chlorpropamide,
0·6% for glibenclamide, 2·3% for insulin, and 0·1%
for diet; the corresponding rates for any hypoglycaemic
episode were 11·0%, 17·7%, 36·5%, and 1·2%.
By intention-to-treat analyses, major hypoglycaemic
episodes occurred with chlorpropamide (1·0%),
glibenclamide (1·4%), insulin (1·8%), and diet (0·7%)
and any hypoglycaemic episodes in 16%, 21%, 28%,
and 10%, respectively. Hypoglycaemic episodes in
patients on diet therapy were reactive and occurred
either after meals or after termination of glucose
infusions given while in hospital (eg, postoperatively).
Discussion
We found that an intensive blood-glucose-control policy
with an 11% reduction in median HbA1c over the first 10
years decreased the frequency of some clinical
complications of type 2 diabetes. The intensive
treatment group had a substantial, 25% reduction in the
risk of microvascular endpoints, most of which was due
to fewer patients requiring photocoagulation. There
THE LANCET • Vol 352 • September 12, 1998
was evidence, albeit inconclusive, of a 16% risk
reduction (p=0·052) for myocardial infarction, which
included non-fatal and fatal myocardial infarction and
sudden death, but diabetes-related mortality and allcause mortality did not differ between the intensive and
conventional groups. The study did not have sufficient
power to exclude a beneficial effect on fatal outcomes.
The progression of subclinical, surrogate variables of
microvascular disease was also decreased, in agreement
with other studies of improved glucose control.1–3 The
median complication-free interval was 1·3 years longer
in the intensive group.
The UGDP raised concerns that the sulphonylurea,
tolbutamide, may increase the risk of cardiovascular
death, and several mechanisms by which sulphonylureas
might have an adverse effect were suggested. However,
we found no difference in the rates of myocardial
infarction or diabetes-related death between participants
assigned sulphonylurea or insulin therapies. Studies in
animals suggested that first-generation sulphonylureas,
such as chlorpropamide, might increase the risk of
ventricular fibrillation,10 but this suggestion was not
supported by our findings since the rate of sudden death
was similar in the groups assigned chlorpropamide,
glibenclamide, or insulin. Thus, the UKPDS data do
not support the suggestion of adverse cardiovascular
effects from sulphonylureas.
849
ARTICLES
Figure 10: Major and any hypoglycaemic episodes by intention-to-treat analysis and actual therapy by individual intensive therapy
and conventional treatment
Data from first 15 centres. The numbers of patients studied at 5, 10 and 15 years in the intensive groups with chlorpropamide, glibenclamide and
insulin and the conventional group by actual therapy were 380, 378, 559, 395; 171, 175, 416, 150; and 21, 16, 83, 14 respectively.
Exogenous insulin has also been suggested as
potentially harmful treatment because in-vitro studies
with raised insulin concentrations induced atheroma,24
and epidemiological studies showed an association
between high plasma insulin concentrations and
myocardial infarction.25,26 We did not find an increase in
myocardial infarctions in patients assigned insulin
therapy, even though their fasting plasma insulin
concentrations were higher than those in any other
group. The macrovascular subclinical surrogate
endpoints did not differ between intensive and
conventional groups, perhaps because 10 years’
follow-up is too short to find changes in atheroma or
because the endpoints were not sufficiently sensitive.
Since there was no evidence, however, for a harmful
cardiovascular effect of sulphonylurea or insulin therapy,
it appears that the beneficial effect of an intensive
glucose control with these agents outweighs the
theoretical risks.
The 0·9% difference in HbA1c between the intensive
(7·0%) and conventional (7·9%) groups over 10 years,
an 11% reduction, is smaller than the 1·9% difference
(9·0% and 7·1%; 20% reduction) in HbA1c in the
DCCT.2 The DCCT studied younger patients with
type 1 diabetes and used slightly different methods that
focused on surrogate variables. The risk reductions seem
850
proportional given the HbA1c differences: for progression
of microvascular disease, 21% for retinopathy in
UKPDS and 63% in the DCCT; and, for albuminuria,
34% and 54% respectively.8 Our data suggest that
clinical benefit can be obtained at lower HbA1c values
than those in the DCCT.
Few patients had late ophthalmic complications such
as vitreous haemorrhage or blindness and this may be
because the follow-up was not long enough or, more
likely, because of the decrease in retinal damage and
blindness after photocoagulation.27,28
The reduction in the progression of albuminuria by
intensive treatment was probably accompanied by
a reduced risk for development of renal failure,
since there was a 67% risk reduction in the proportion
of patients who had a two-fold increase in plasma
creatinine and 74% risk reduction in those who
had a doubling of their plasma urea. This result is
potentially important since, although less than 1% of
UKPDS patients developed renal failure, in many
populations type 2 diabetes is the principal cause of
renal failure.
No difference in the risk reduction of microvascular
clinical endpoints was seen between the three intensive
treatments, and thus, improved glycaemic control,
rather than any one therapy, is the principal factor.
THE LANCET • Vol 352 • September 12, 1998
ARTICLES
Nevertheless, patients assigned chlorpropamide did not
have the same risk reduction in progression of the
retinopathy as those assigned glibenclamide or insulin,
and this difference was not accounted for, statistically,
by higher blood pressure in the chlorpropamide group.
There was no difference in the progression of
albuminuria between any treatment groups.
Increased blood pressure has been reported with
chlorpropamide,29 which can also cause water
retention.30 Other sulphonylureas that do not raise the
blood pressure may be preferred.
Intensive blood-glucose control had disadvantages
such as greater weight gain than occurred in the
conventional group. There was also an increased risk of
hypoglycaemic episodes, particularly in patients treated
with insulin; each year about 3% had a major episode
and 40% a minor or major hypoglycaemic episode.
Although the increased risk of hypoglycaemia with
insulin was less than that in the DCCT,31 this risk
limited the extent to which normoglycaemia could be
obtained in our patients with type 2 diabetes32—as it
does in patients with type 1 diabetes.2
The relation between glycaemia and outcome in our
study is complex. Although a difference in HbA1c
between conventional and intensive groups was
maintained throughout, HbA1c progressively increased.
The risks of hypoglycaemia and of weight gain,
particularly in patients treated by insulin, are perceived
by patients as difficulties that limit their ability to
achieve improved glucose control (data not shown).
Although early addition of other agents may have
delayed the increasing hyperglycaemia, each of the
available oral hypoglycaemic agents (sulphonylureas,33
metformin,32,34 thiazolidinediones,35 and acarbose36 have
limited glucose-lowering efficacy and many patients
eventually required insulin to avoid marked
hyperglycaemia. Our patients on intensive treatment
with insulin achieved lower HbA1c values than those seen
in several studies of intensive glucose control in patients
with type 2 diabetes.37–39 Recent recommendations40 set
an HbA1c below 7% as a goal but, to our knowledge, this
has been achieved only in intervention studies with high
insulin doses, generally above 100 U per day, in small
groups of obese patients who received detailed attention
over a short period.41,42 Studies of glycaemic control in
type 2 diabetes with insulin therapy in the community
report mean HbA1c values of 8·5%43 and 9·0%.44 Current
therapy of type 2 diabetes, including insulin regimens,
may need to be reviewed. The US National Health and
Nutrition Education Examination Survey III, by the
same assay method as the DCCT, found that 51% of
insulin-treated patients and 42% of those on oral
hypoglycaemic agents had HbA1c values greater than 8%,
(Maureen Harris, National Institute of Diabetes and
Digestive and Kidney Diseases, USA; personal
communication).
About 50% of patients with newly diagnosed type 2
diabetes already have diabetic tissue damage,13 but lack
of benefit in these patients from early treatment has
meant variation in the guidelines for screening
populations.45,46 The UKPDS shows that improved
blood-pressure47 and glucose control reduce the risk of
the diabetic complications that cause both morbidity
and premature mortality, and increase the case for
formal screening programmes for early detection of
diabetes in the general population.
THE LANCET • Vol 352 • September 12, 1998
Our study, despite the median of 10 years’ follow-up
is still short compared with the median life expectancy of
20 years in UKPDS patients diagnosed at median age 53
years. To investigate longer-term responses, we will
carry out post-study monitoring of UKPDS for a further
5 years, to establish whether the improved glucose
control achieved will substantially decrease the risk of
fatal and non-fatal myocardial infarctions with longer
follow-up.
UKPDS shows that an intensive glucose-control
treatment policy that maintains an 11% lower HbA1c—
ie, median 7·0% over the first 10 years after diagnosis of
diabetes—substantially reduces the frequency of
microvascular endpoints but not diabetes-related
mortality or myocardial infarction. The disadvantages of
intensive treatment are weight gain and risk of
hypoglycaemia. There was no evidence that intensive
treatment with chlorpropamide, glibenclamide, or
insulin had a specific adverse effect on macrovascular
disease.
UKPDS Study Organisation
Writing Committee—Robert C Turner, Rury R Holman, Carole A Cull,
Irene M Stratton, David R Matthews, Valeria Frighi, Susan E Manley,
Andrew Neil, Heather McElroy, David Wright, Eva Kohner, Charles
Fox, and David Hadden.
Coordinating centre—Chief investigators: R C Turner, R R Holman.
Additional investigators: D R Matthews, H A W Neil. Statisticians:
I M Stratton, C A Cull, H J McElroy, Z Mehta (previously A Smith,
Z Nugent). Biochemist: S E Manley. Research associate: V Frighi.
Consultant statistician: R Peto. Epidemiologist: A I Adler (previously
J I Mann). Administrator: P A Bassett, (previously S F Oakes).
Endpoint assessors: D R Matthews (Oxford), A D Wright
(Birmingham), T L Dornan (Salford). Retinal-photography grading:
E M Kohner, S Aldington, H Lipinski, R Collum, K Harrison,
C MacIntyre, S Skinner, A Mortemore, D Nelson, S Cockley, S Levien,
L Bodsworth, R Willox, T Biggs, S Dove, E Beattie, M Gradwell,
S Staples, R Lam, F Taylor, L Leung (Hammersmith). Dietician:
E A Eeley (Oxford). Biochemistry laboratory staff: M J Payne,
R D Carter, S M Brownlee, K E Fisher, K Islam, R Jelfs, P A Williams,
F A Williams, P J Sutton, A Ayres, L J Logie, C Lovatt, M A Evans,
L A Stowell. Consultant biochemist: I Ross (Aberdeen). Applications
programmer: I A Kennedy. Database clerk: D Croft. ECG coding:
A H Keen, C Rose (Guy’s Hospital). Health economists: M Raikou,
A M Gray, A J McGuire, P Fenn (Oxford). Quality-of-life questionnaire:
Z Mehta (Oxford), A E Fletcher, C Bulpitt, C Battersby
(Hammersmith), J S Yudkin (Whittington). Mathematical modeller:
R Stevens (Oxford).
Clinical centres—M R Stearn, S L Palmer, M S Hammersley,
S L Franklin, R S Spivey, J C Levy, C R Tidy, N J Bell, J Steemson,
B A Barrow, R Coster, K Waring, L Nolan, E Truscott, N Walravens,
L Cook, H Lampard, C Merle, P Parker, J McVittie, I Draisey (Oxford);
L E Murchison, A H E Brunt, M J Williams, D W Pearson,
X M P Petrie, M E J Lean D Walmsley, F Lyall, E Christie, J Church,
E Thomson, A Farrow, J M Stowers, M Stowers, K McHardy,
N Patterson (Aberdeen); A D Wright, N A Levi, A C I Shearer,
R J W Thompson, G Taylor, S Rayton, M Bradbury, A Glover,
A Smyth-Osbourne, C Parkes, J Graham, P England, S Gyde, C Eagle,
B Chakrabarti, J Smith, J Sherwell (Birmingham); N W Oakley,
M A Whitehead, G P Hollier, T Pilkington, J Simpson, M Anderson,
S Martin, J Kean, B Rice, A Rolland, J Nisbet (London, St George’s);
E M Kohner, A Dornhorst, M C Doddridge, M Dumskyj, S Walji,
P Sharp, M Sleightholm, G Vanterpool, C Rose, G Frost, M Roseblade,
S Elliott, S Forrester, M Foster, K Myers, R Chapman (London,
Hammersmith); J R Hayes, R W Henry, M S Featherston,
G P R Archbold, M Copeland, R Harper, I Richardson, S Martin,
M Foster, H A Davison (City Hospital, Belfast); L Alexander,
J H B Scarpello, D E Shiers, R J Tucker, J R H Worthington, S Angris,
A Bates, J Walton, M Teasdale, J Browne, S Stanley, B A Davis,
R C Strange (Stoke-on-Trent); D R Hadden, L Kennedy, A B Atkinson,
P M Bell, D R McCance, J Rutherford, A M Culbert, C Hegan,
H Tennet, N Webb, I Robinson, J Holmes, M Foster, J Rutherford,
S Nesbitt (Royal Victoria Hospital, Belfast); A S Spathis, S Hyer,
M E Nanson, L M James, J M Tyrell, C Davis, P Strugnell, M Booth,
H Petrie, D Clark, B Rice, S Hulland, J L Barron (Carshalton);
J S Yudkin, B C Gould, J Singer, A Badenoch, S Walji, M McGregor,
L Isenberg, M Eckert, K Alibhai, E Marriot, C Cox, R Price,
M Fernandez, A Ryle, S Clarke, G Wallace, E Mehmed, J A Lankester,
E Howard, A Waite, S MacFarlane (London, Whittington);
851
ARTICLES
R H Greenwood, J Wilson, M J Denholm, R C Temple, K Whitfield,
F Johnson, C Munroe, S Gorick, E Duckworth, M Fatman, S Rainbow
(Norwich); L J Borthwick, D J Wheatcroft, R J Seaman, R A Christie,
W Wheatcroft, P Musk, J White, S McDougal, M Bond, P Raniga
(Stevenage); J L Day, M J Doshi, J G Wilson, J R Howard-Williams,
H Humphreys, A Graham, K Hicks, S Hexman, P Bayliss, D Pledger
(Ipswich); R W Newton, R T Jung, C Roxburgh, B Kilgallon, L Dick,
M Foster, N Waugh, S Kilby, A Ellingford, J Burns (Dundee); C V Fox,
M C Holloway, H M Coghill, N Hein, A Fox, W Cowan, M Richard,
K Quested, S J Evans (Northampton); R B Paisey, N P R Brown,
A J Tucker, R Paisey, F Garrett, J Hogg, P Park, K Williams, P Harvey,
R Wilcocks, S Mason, J Frost, C Warren, P Rocket, L Bower (Torbay);
J M Roland, D J Brown, J Youens, K Stanton-King, H Mungall, V Ball,
W Maddison, D Donnelly, S King, P Griffin, S Smith, S Church,
G Dunn, A Wilson, K Palmer (Peterborough); P M Brown,
D Humphriss, A J M Davidson, R Rose, L Armistead, S Townsend,
P Poon (Scarborough); I D A Peacock, N J C Culverwell,
M H Charlton, B P S Connolly, J Peacock, J Barrett, J Wain,
W Beeston, G King, P G Hill (Derby); A J M Boulton, A M Robertson,
V Katoulis, A Olukoga, H McDonald, S Kumar, F Abouaesha,
B Abuaisha, E A Knowles, S Higgins, J Booker, J Sunter, K Breislin,
R Parker, P Raval, J Curwell, H Davenport, G Shawcross, A Prest,
J Grey, H Cole, C Sereviratne (Manchester); R J Young, T L Dornan,
J R Clyne, M Gibson, I O’Connell, L M Wong, S J Wilson, K L Wright,
C Wallace, D McDowell (Salford); A C Burden, E M Sellen, R Gregory,
M Roshan, N Vaghela, M Burden, C Sherriff, S Mansingh, J Clarke,
J Grenfell (Leicester); J E Tooke, K MacLeod, C Seamark, M Rammell,
C Pym, J Stockman, C Yeo, J Piper, L Leighton, E Green, M Hoyle, K
Jones, A Hudson, A J James, A Shore, A Higham, B Martin (Exeter).
UKPDS Data Committee—C A Cull, V Frighi, R R Holman,
S E Manley, D R Matthews, H A W Neil, I M Stratton, R C Turner.
Data-monitoring and Ethics Committee —W J H Butterfield,
W R S Doll, R Eastman, F R Ferris, R R Holman, N Kurinij, R Peto,
K McPherson, R F Mahler, T W Meade, G Shafer, R C Turner,
P J Watkins (Previous members: H Keen, D Siegel).
Policy advisory group—C V Fox, D R Hadden, R R Holman,
D R Matthews, R C Turner, A D Wright, J S Yudkin.
Steering Committee for glucose study —D J Betteridge, R D Cohen,
D Currie, J Darbyshire, J V Forrester, T Guppy, R R Holman,
D G Johnston, A McGuire, M Murphy, A M el-Nahas, B Pentecost,
D Spiegelhalter, R C Turner, (previous members: K G M M Alberti,
R Denton, P D Home, S Howell, J R Jarrett, V Marks, M Marmot,
J D Ward).
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
Acknowledgments
We thank the patients and many NHS and non-NHS staff at the centres
for their cooperation; Philip Bassett for editorial assistance; and
Caroline Wood, Kathy Waring, and Lorraine Mallia for typing the
papers. The study was supported by grants from the UK Medical
Research Council, British Diabetic Association, UK Department of
Health, US National Eye Institute and US National Institute of
Diabetes, Digestive and Kidney Disease (National Institutes of Health),
British Heart Foundation, Wellcome Trust, Charles Wolfson Charitable
Trust, Clothworkers’ Foundation, Health Promotion Research Trust,
Alan and Babette Sainsbury Trust, Oxford University Medical Research
Fund Committee, Novo-Nordisk, Bayer, Bristol-Myers Squibb,
Hoechst, Lilly, Lipha, and Farmitalia Carlo Erba. We also thank
Boehringer Mannheim, Becton Dickinson, Owen Mumford, Securicor,
Kodak, Cortecs Diagnostics. We thank Glaxo Wellcome, Smith Kline
Beecham, Pfizer, Zeneca, Pharmacia and Upjohn, and Roche for
funding epidemiological, statistical, and health-economics analyses.
22
23
24
25
26
27
28
References
1
2
3
4
5
Reichard P, Berglund B, Britz A, Cars I, Nilsson BY, Rosenqvist U.
Intensified conventional insulin treatment retards the microvascular
complications of insulin-dependent diabetes mellitus (IDDM): the
Stockholm Diabetes Intervention Study (SDIS) after 5 years. J Intern
Med 1991; 230: 101–08.
DCCT Research Group. The effect of intensive treatment of diabetes
on the development and progression of long-term complications
insulin-dependent diabetes mellitus. N Engl J Med 1993; 329:
977–86.
Ohkubo Y, Kishikawa H, Araki E, et al. Intensive insulin therapy
prevents the progression of diabetic microvascular complications in
Japanese patients with non-insulin-dependent diabetes mellitus: a
randomized prospective 6-year study. Diabetes Res Clin Pract 1995;
28: 103–17.
UKPDS Group. UK Prospective Diabetes Study 17: A nine-year
update of a randomized, controlled trial on the effect of improved
metabolic control on complications in non-insulin-dependent
diabetes mellitus. Ann Intern Med 1996; 124: 136–45.
Fuller J, McCartney P, Jarrett RJ. Hyperglycaemia and coronary
heart disease: the Whitehall Study. J Chronic Dis 1979; 32: 721–28.
852
29
30
31
32
33
34
Balkau B, Shipley M, Jarrett RJ, et al. High blood glucose
concentration is a risk factor for mortality in middle-aged
nondiabetic men. Diabetes Care 1998; 21: 360–67.
University Group Diabetes Program. Effects of hypoglycemic agents
on vascular complications in patients with adult-onset diabetes VII:
mortality and selected nonfatal events with insulin treatment.
JAMA 1978; 240: 37–42.
University Group Diabetes Program. A study of the effects of
hypoglycemic agents on vascular complications in patients with
adult-onset diabetes. Diabetes 1976; 25: 1129–53.
Smits P, Thien T. Cardiovascular effects of sulphonylurea
derivatives. Diabetologia 1995; 38: 116–21.
Pogatsa G. Potassium channels in the cardiovascular system.
Diabetes Res Clin Pract 1995; 28 (suppl 1): S91-S98.
Stout RW. Insulin and atherosclerosis. In: Stout RW, ed. Diabetes
and atherosclerosis. Dordrecht: Kluwer Academic Publishers, 1992:
165–201.
Genuth S. Exogenous insulin administration and cardiovascular risk
in non-insulin-dependent and insulin-dependent diabetes mellitus.
Ann Intern Med 1996; 124: 104–09.
UKPDS Group. UK Prospective Diabetes Study VIII: study design,
progress and performance. Diabetologia 1991; 34: 877–90.
Metropolitan Life Insurance Company. Net weight standard for men
and women. Stat Bull Metrop Insur Co 1959; 40: 1–4.
UKPDS Group. Effect of intensive blood-glucose control with
metformin on complications in overweight patients with type 2
diabetes (UKPDS 34). Lancet 1998; 352: 854–65.
Hypertension in Diabetes Study IV. Therapeutic requirements to
maintain tight blood pressure control. Diabetologia 1996; 39: 1554–61.
Holman RR, Cull CA, Turner RC. Glycaemic improvement over one
year in a double-blind trial of acarbose in 1,946 NIDDM patients.
Diabetologia 1996; 39 (suppl 1): A44.
UKPDS Group. UK Prospective Diabetes Study XI: biochemical
risk factors in type 2 diabetic patients at diagnosis compared with
age-matched normal subjects. Diabet Med 1994; 11: 534–44.
Manley SE, Burton ME, Fisher KE, Cull CA, Turner RC. Decreases
in albumin/creatinine and N-acetylglucosaminidase/creatinine ratios
in urine samples stored at –20ºC. Clin Chem 1992; 38: 2294–99.
World Health Organisation. International Classification of
Procedures in Medicine. Geneva: World Health Organisation, 1978.
UKPDS Group. UK Prospective Diabetes Study IX: relationships of
urinary albumin and N-acetylglucosaminidase to glycaemia and
hypertension at diagnosis of type 2 (non-insulin-dependent) diabetes
mellitus and after 3 months diet therapy. Diabetologia 1992; 36:
835–42.
American Diabetes Association. Report of the expert committee on
the diagnosis and classification of diabetes mellitus. Diabetes Care
1998; 21 (suppl 1); 55–19.
World Health Organization. Diabetes mellitus. WHO technical
report series no 727. Geneva: WHO, 1985.
Stout RW. Insulin and atheroma: 20-yr perspective. Diabetes Care
1990; 13: 631–54.
Pyörälä K. Relationship of glucose tolerance and plasma insulin to
the incidence of coronary heart disease: results from two population
studies in Finland. Diabetes Care 1979; 2: 131–41.
Desprès JP, Lamarche B, Mauriège P, et al. Hyperinsulinemia as an
independent risk factor for ischemic heart disease. N Engl J Med
1996; 334: 952–57.
Davies EG, Petty RG, Kohner EM. Long term effectiveness of
photocoagulation for diabetic maculopathy. Eye 1989; 3: 764–67.
British Multicentre Group. Photocoagulation for proliferative
diabetic retinopathy: a randomised controlled clinical trial using the
xenon-arc. Diabetologia 1984; 26: 109–15.
Schmitt JK, Moore JR. Hypertension secondary to chlorpropamide
with amelioration by changing to insulin. Am J Hypertens 1993; 6:
317–19.
Melander A. Sulphonylureas in the treatment of
non-insulin-dependent diabetes. Baillieres Clin Endocrinol Metab
1988; 2: 443–53.
DCCT Research Group. Adverse events and their association with
treatment regimens in the Diabetes Control and Complications Trial.
Diabetes Care 1995; 18: 1415–27.
UKPDS Group. UK Prospective Diabetes Study 16: overview of six
years’ therapy of type 2 diabetes - a progressive disease. Diabetes
1995; 44: 1249–58.
UKPDS Group. UK Prospective Diabetes Study 26: sulphonylurea
failure in non-insulin dependent diabetic patients over 6 years.
Diabet Med 1998; 15: 297–303.
UKPDS Group. UK Prospective Diabetes Study 28: a randomised
trial of efficacy of early addition of metformin in
sulphonylurea-treated non-insulin dependent diabetes. Diabetes Care
1998; 21: 87–92.
THE LANCET • Vol 352 • September 12, 1998
ARTICLES
35 Kumar S, Boulton AJ, Beck-Nielsen H, et al. Troglitazone, an insulin
action enhancer, improves metabolic control in NIDDM patients.
Diabetologia 1996; 39: 701–09.
36 Chiasson JL, Josse RG, Hunt JA, et al. The efficacy of acarbose in
the treatment of patients with non-insulin-dependent diabetes
mellitus. A multicenter controlled clinical trial. Ann Intern Med 1996;
121: 928–35.
37 Birkeland KI, Rishaug U, Hanssen KE, Vaaler S. NIDDM: a rapid
progressive disease. Diabetologia 1996; 39: 1629–33.
38 Yki-Järvinen H, Kauppila M, Kujansuu E, et al. Comparison of
insulin regimens in patients with non-insulin-dependent diabetes
mellitus. N Engl J Med 1992; 327: 1426–33.
39 Chow CC, Tsang LWW, Sorensen JP. Comparison of insulin with or
without continuation of oral hypoglycaemic agents in the treatment
of secondary failure in NIDDM patients. Diabetes Care 1995; 18:
307–14.
40 American Diabetes Association. Standards of medical care for
patients with diabetes mellitus. Diabetes Care 1998; 21 (suppl 1):
S23-S31.
THE LANCET • Vol 352 • September 12, 1998
41 Abraira C, Colwell JA, Nuttall FQ. Veterans Affairs Cooperative
Study on glycemic control and complications in Type II diabetes
(VACSDM). Diabetes Care 1995; 18: 1113–23.
42 Henry RR, Gumbiner B, Ditzler T, Wallace P, Lyon R, Glauber HS.
Intensive conventional insulin therapy for type II diabetes: metabolic
effects during a 6 month outpatient trial. Diabetes Care 1993; 16:
21–31.
43 Hayward RA, Manning WG, Kaplan SH, Wagner EH, Greenfield S.
Starting insulin therapy in patients with Type 2 diabetes.
JAMA 1997; 278: 1663–700.
44 Dunn NR, Bough P. Standards of care of diabetic patients in a
typical English community. Br J Gen Pract 1996; 46: 401–05.
45 American Diabetes Association. Clinical practice recommendations
1998. Diabetes Care 1998; 21 (suppl 1): S20-S22.
46 de Courten M, Zimmet P. Screening for non-insulin-dependent
diabetes mellitus: where to draw the line? Diabet Med 1997; 14: 5–98.
47 UKPDS Group. Tight blood pressure control and risk of
macrovascular and microvascular complications in type 2 diabetes:
UKPDS 38. BMJ (in press).
853
ARTICLES
Effect of intensive blood-glucose control with metformin on
complications in overweight patients with type 2 diabetes
(UKPDS 34)
UK Prospective Diabetes Study (UKPDS) Group*
Summary
Background In patients with type 2 diabetes, intensive
blood-glucose control with insulin or sulphonylurea
therapy decreases progression of microvascular disease
and may also reduce the risk of heart attacks. This study
investigated whether intensive glucose control with
metformin has any specific advantage or disadvantage.
Methods Of 4075 patients recruited to UKPDS in 15
centres, 1704 overweight (>120% ideal bodyweight)
patients with newly diagnosed type 2 diabetes, mean age
53 years, had raised fasting plasma glucose (FPG;
6·1–15·0 mmol/L) without hyperglycaemic symptoms
after 3 months’ initial diet. 753 were included in a
randomised controlled trial, median duration 10⋅7 years,
of conventional policy, primarily with diet alone (n=411)
versus intensive blood-glucose control policy with
metformin, aiming for FPG below 6 mmol/L (n=342). A
secondary analysis compared the 342 patients allocated
metformin with 951 overweight patients allocated
intensive blood-glucose control with chlorpropamide
(n=265), glibenclamide (n=277), or insulin (n=409). The
primary outcome measures were aggregates of any
diabetes-related clinical endpoint, diabetes-related death,
and all-cause mortality. In a supplementary randomised
controlled trial, 537 non-overweight and overweight
patients, mean age 59 years, who were already on
maximum sulphonylurea therapy but had raised FPG
(6·1–15.0
mmol/L)
were
allocated
continuing
sulphonylurea therapy alone (n=269) or addition of
metformin (n=268).
Findings Median glycated haemoglobin (HbA1c) was 7·4%
in the metformin group compared with 8·0% in the
conventional group. Patients allocated metformin,
compared with the conventional group, had risk
reductions of 32% (95% CI 13–47, p=0·002) for any
diabetes-related endpoint, 42% for diabetes-related death
(9–63, p=0·017), and 36% for all-cause mortality (9–55,
p=0·011). Among patients allocated intensive bloodglucose control, metformin showed a greater effect than
chlorpropamide, glibenclamide, or insulin for any
diabetes-related endpoint (p=0·0034), all-cause mortality
(p=0·021), and stroke (p=0·032). Early addition of
metformin in sulphonylurea-treated patients was
*Study organisation given at end of paper
Correspondence to: Prof Robert Turner, UKPDS Group,
Diabetes Research Laboratories, Radcliffe Infirmary,
Oxford OX2 6HE, UK
854
associated with an increased risk of diabetes-related
death (96% increased risk [95% CI 2–275], p=0·039)
compared with continued sulphonylurea alone. A
combined analysis of the main and supplementary studies
showed fewer metformin-allocated patients having
diabetes-related endpoints (risk reduction 19% [2–33],
p=0·033). Epidemiological assessment of the possible
association of death from diabetes-related causes with
the concurrent therapy of diabetes in 4416 patients
did not show an increased risk in diabetes-related
death in patients treated with a combination of
sulphonylurea and metformin (risk reduction 5% [233 to
32], p=0·78).
Interpretation Since intensive glucose control with
metformin appears to decrease the risk of diabetesrelated endpoints in overweight diabetic patients, and is
associated with less weight gain and fewer
hypoglycaemic
attacks
than
are
insulin
and
sulphonylureas, it may be the first-line pharmacological
therapy of choice in these patients.
Lancet 1998; 352: 854–65
See Commentary page xxx
Introduction
The UK Prospective Diabetes Study reported that
intensive blood-glucose control with sulphonylureas or
insulin substantially reduced the risk of complications
but not macrovascular disease.1
Metformin is a biguanide that decreases blood glucose
concentration by mechanisms different from those of
sulphonylurea or insulin. It lowers, rather than
increases, fasting plasma insulin concentrations2 and
acts by enhancing insulin sensitivity, inducing greater
peripheral uptake of glucose, and decreasing hepatic
glucose output.3 The improved glucose control is
achieved without weight gain.4 Biguanides also decrease
concentrations of plasminogen-activator inhibitor type 1
(PAI-1)5 and may thus increase fibrinolytic activity. This
effect may be secondary either to enhanced insulin
sensitivity or to lower insulin concentrations, because
therapy with troglitazone (a thiazolidinedione) also
decreases production of PAI-1 and increases insulin
sensitivity.6
The only long-term outcome data on biguanides
available were from the University Group Diabetes
Program (UGDP) study of phenformin. An unexpected
outcome was higher mortality from cardiovascular
causes with phenformin than with placebo, and for total
mortality for phenformin than with a combination of
THE LANCET • Vol 352 • September 12, 1998
ARTICLES
Figure 1: Trial profile for diet/metformin study in overweight diet-treated patients
insulin and placebo allocations.7 The study design did
not allow comparison of phenformin with the
sulphonylurea used in the UGDP (tolbutamide). One
death from lactic acidosis occurred in the phenformin
group. Phenformin was withdrawn from clinical use in
many countries, partly because of the UGDP data and
partly because of the association with lactic acidosis.8
Metformin is now the only biguanide in general use,
since it has a 10–20-fold lower risk of lactic acidosis
than phenformin, and is regarded as a safe drug
provided it is not used in at-risk patients, such as those
in renal failure.9
Metformin was included as a randomisation option in
overweight patients in the UK Prospective Diabetes
Study (UKPDS) from 1977 as part of the original
protocol in the first 15 centres. The primary aim was to
compare conventional treatment (primarily with diet
alone) with intensive treatment with metformin,10–12 with
a secondary aim of comparing the group allocated
metformin
with
overweight
patients
allocated
sulphonylurea or insulin therapies.
In 1990, increasing glycaemia despite maximum
sulphonylurea therapy was noted. Following a UKPDS
protocol amendment, normal-weight and overweight
patients allocated sulphonylurea treatment, who had
fasting plasma glucose (FPG) concentrations of
6⋅1–15⋅0 mmol/L but no symptoms on maximum doses,
were then assigned either continuing treatment with
sulphonylurea alone or addition of metformin to
sulphonylurea.
We report here on whether addition of metformin
reduces the risk of clinical complications of diabetes.
brief, between 1977 and 1991, general practitioners in 23
centres in the UK referred patients with newly diagnosed type
2 diabetes, aged 25–65 years, for possible inclusion in UKPDS.
5102 diabetic patients with FPG above 6⋅0 mmol/L on two
mornings were recruited. The patients were advised to follow a
diet high in carbohydrates and fibre and low in saturated fats,
with energy restriction in overweight patients. After 3 months
on diet, 4209 eligible patients with FPG above 6⋅0 mmol/L
were randomised by a stratified design: 2022 (48%) were nonoverweight patients (<120% ideal bodyweight13) and 2187
(52%) were overweight. Patients were allocated conventional
treatment with diet or intensive treatment with sulphonylurea
or insulin with metformin as an additional intensive therapy
option in overweight patients in the first 15 centres. We report
here results for the overweight participants who had FPG
between 6·1 and 15·0 mmol/L (n=1704) without symptoms of
hyperglycaemia, after diet treatment.
This paper reports on two randomised controlled trials in
patients in the first 15 centres, in which metformin was a
therapeutic option.
Trial in overweight, diet-treated patients of intensive
blood-glucose control with metformin versus
conventional treatment
The 1704 overweight patients were randomly assigned
conventional treatment, primarily with diet (24%), or intensive
treatment with chlorpropamide (16%), glibenclamide (16%),
insulin (24%), or metformin (20%). This report primarily
compares the 411 overweight patients assigned conventional
treatment and 342 overweight patients assigned intensive
treatment with metformin, as designated in the protocol10
(figure 1). The paper also reports the secondary analysis
comparing the outcomes between overweight patients allocated
metformin (n=342) with the 951 patients allocated intensive
therapy with chlorpropamide (n=265), glibenclamide (n=277),
or insulin (n=409).
Methods
Conventional treatment policy
Patients
The 411 overweight patients assigned the conventional
approach continued to receive dietary advice at 3-monthly
UKPDS has been described in the accompanying paper.1,10 In
THE LANCET • Vol 352 • September 12, 1998
855
ARTICLES
Figure 2: Trial profile for sulphonylurea-treated patients with randomisation to metformin
clinical visits with the aim of attaining normal bodyweight and
FPG to the extent that is feasible in clinical practice. If marked
hyperglycaemia developed (defined by the protocol as FPG
above 15 mmol/L or symptoms of hyperglycaemia1) patients
were secondarily randomised to additional non-intensive
pharmacological therapy with the other four treatments
(metformin, chlorpropamide, glibenclamide, and insulin) in the
same proportions as in the primary randomisations, with the
aim of avoiding symptoms and maintaining FPG below 15
mmol/L.1 If patients assigned sulphonylurea therapy developed
marked hyperglycaemia, metformin was added to their
regimen; if marked hyperglycaemia recurred, the allocation was
changed to insulin therapy.
maintaining FPG below 6⋅0 mmol/L. If marked hyperglycaemia
again developed, treatment was changed to insulin, initially
ultralente (Ultratard HM, Novo, or Humulin Zn, Lilly) or
isophane (NPH) insulin, with the addition of short-acting
(regular) insulin, usually soluble insulin before meals when
premeal or bedtime blood-glucose concentrations were
above 7⋅0 mmol/L. If the glucose control was not satisfactory,
other regimens could be introduced (eg, soluble/isophane
regimens).
Intensive treatment policy with metformin
1234 patients, both non-overweight and overweight, were
assigned to intensive treatment with sulphonylurea in the first
15 centres. Of these, 537 who were treated with maximum
doses of sulphonylurea and had FPG of 6·1–15·0 mmol/L
without symptoms of hyperglycaemia, were randomly
assigned in equal proportions early addition of metformin to
the sulphonylurea (n=269) or continued sulphonylurea alone
(n=268; figure 2). If those allocated sulphonylurea
alone later developed protocol-defined marked hyperglycaemia,
metformin was added. If patients with early or later addition of
metformin developed protocol-defined marked hyperglycaemia,
oral therapy was stopped and changed to insulin therapy.
The aim of the intensive approach for glucose control with
metformin, sulphonylurea, or insulin therapies, in addition to
dietary advice, was to obtain near-normal FPG (ie, <6⋅0
mmol/L). If FPG increased, patients were kept on the allocated
monotherapy alone until marked hyperglycaemia developed, so
that the clinical effects of each therapy could be assessed.
342 overweight patients were assigned intensive control with
metformin. Treatment started with one 850 mg tablet per day,
then 850 mg twice daily, and then 1700 mg in the morning and
850 mg with the evening meal (maximum dose=2550 mg). If
on any dose, symptoms of diarrhoea or nausea occurred,
patients were asked to reduce the dose to that which previously
did not cause symptoms.
When marked hyperglycaemia developed in those allocated
metformin, glibenclamide was added with the aim of
856
Trial in non-overweight and overweight sulphonylureatreated patients of addition of metformin versus
continued sulphonylurea alone
Combined analysis of two randomised controlled trials
The unexpected finding of an increased risk of mortality in
THE LANCET • Vol 352 • September 12, 1998
ARTICLES
Demographic
Age (years)*
M/F
Ethnicity (%) Caucasian/Indian Asian/
Afro-Caribbean/other
Clinical
Weight (kg)*
Body-mass index (kg/m2)
Systolic blood pressure (mm Hg)*
Diastolic blood pressure (mm Hg)*
Smoking (%) never/ex/current
Alcohol (%) none/social/regular/
dependent
Exercise (%) sedentary/moderately
active/active/fit
Biochemical
FPG (mmol/L)†
HBA1c (%)*
Plasma insulin (pmol/L)‡
Triglyceride (mmol/L)‡
Total cholesterol (mmol/L)*
LDL cholesterol (mmol/L)*
HDL cholesterol (mmol/L)*
Medications
More than one aspirin daily (%)
Diuretic (%)
Others (%) digoxin/antihypertensives/
lipid lowering/HRT or OC
Surrogate clinical endpoints
Retinopathy (%)
Proteinuria (%)
Plasma creatinine (mmol/L)‡
Biothesiometer more than 25 V (%)
Conventional
(n=411)
Metformin
(n=342)
Insulin
(n=409)
Chlorpropamide
(n=265)
Glibenclamide
(n=277)
All patients
(n=1704)
53 (9)
193 (47%)/218
86/6/7/1
53 (8)
157 (46%)/185
85/4/10/1
53 (8)
192 (47%)/217
88/4/8/0
53 (9)
119 (45%)/146
86/6/8/0
53 (9)
127 (46%)/150
87/4/8/1
53 (8)
784 (46%)/920
86/5/8/1
87 (15)
31·8 (4·9)
140 (18)
86 (10)
39/36/25
30/56/14/0·5
87 (17)
31·6 (4·8)
140 (18)
85 (9)
43/32/25
27/58/14/1·5
85 (14)
31·0 (4·2)
139 (19)
85 (10)
37/34/39
27/57/15/1·2
85 (15)
31·2 (4·5)
141 (18)
86 (9)
38/30/32
28/54/17/1·1
86 (14)
31·5 (4·4)
139 (19)
85 (9)
34/35/31
25/56/19/1·1
86 (15)
31·4 (4·6)
140 (18)
86 (10)
38/34/28
27/56/15/1·1
24/40/34/3
29/34/35/3
24/37/36/4
21/38/38/3
21/34/40/5
24/36/36/4
8·0 (7·1–9·3)
7·1 (1·5)
114 (71–183)
2·96 (1·03–8·47)
5·5 (1·0)
3·66 (1·04)
1·04 (0·22)
8·1 (7·2–9·8)
8·2 (7·2–10·0)
8·0 (7·2–9·6)
8·2 (7·3–9·6)
8·1 (7·1–9·7)
7·3 (1·5)
7·2 (1·5)
7·2 (1·7)
7·2 (1·5)
7·2 (1·5)
116 (66–203)
116 (71–186)
111 (65–189)
114 (68–189)
114 (69–190)
2·79 (1·01–7·74)
2·89 (1·02–8·19)
2·85 (1·03–7·86)
2·65 (0·99–7·10)
2·84 (1·02–7·92)
5·6 (1·3)
5·6 (1·1)
5·6 (1·2)
5·6 (1·2)
5·6 (1·2)
3·67 (1·16)
3·69 (1·04)
3·59 (1·10)
3·59 (1·07)
3·65 (1·08)
1·06 (0·23)
1·05 (0·23)
1·05 (0·23)
1·07 (0·26)
1·05 (0·23)
1·5
20
0·5/16/0·4/0·4
1·5
17
0·9/15/0/0·3
2·9
20
1·7/12/0/0·3
1·9
20
1·9/15/0·7/0·4
1·1
19
0·4/16/0/0·7
1·8
19
0·9/15/0·1/0·4
33
3·1
78 (64–96)
13·6
38
2·0
77 (63–95)
13·7
39
1·1
77 (63–94)
15·4
37
2·2
79 (65–96)
19·9
29
2·6
79 (65–97)
14·3
36
2·2
79 (66–96)
15·2
Data are % of group, *mean (SD),†median (IQR), or ‡geometric mean (1 SD).
HRT=hormone replacement therapy; OC=oral contraceptive therapy.
Table 1: Baseline characteristics of patients in conventional group and in individual intensive-treatment groups
sulphonylurea-treated patients allocated addition of metformin
led us to undertake a further statistical analysis. Following a
test for heterogeneity between the two trials described above,15
a combined analysis of addition of metformin in patients on
diet therapy and in those on sulphonylurea therapy was done.
The datasets were merged by taking time from randomisation
to metformin or not, to an event, or to a censor date. A formal
meta-analysis16 was also done.
Epidemiological assessment
We excluded 623 of the patients (537 in randomised controlled
trial in patients on maximum sulphonylurea treatment of early
or late addition of metformin, and 86 patients who had
insufficient baseline data or were not in the main three ethnic
groups). The aim of the epidemiological assessment in 4416
participants was to find out whether the combination of
sulphonylurea and metformin was associated with an increase
in mortality from diabetes-related causes. 457 patients were
treated by sulphonylurea and metformin: 107 patients assigned
conventional therapy in the main randomisation who received
the combination after recurrent episodes of protocol-defined
marked hyperglycaemia; 257 patients assigned sulphonylurea
or metformin in the main randomisation, or those with marked
hyperglycaemia after the initial 3 months’ period, who had the
other therapy added when marked hyperglycaemia developed;
and 93 who refused allocated insulin. All these patients were
treated by combined therapy because of the progressive
hyperglycaemia of type 2 diabetes,11 but if marked
hyperglycaemia recurred, the treatment of these patients was
changed to insulin. The combination of sulphonylurea and
metformin was compared with all other therapies in terms of
diabetes-related deaths by means of a Cox proportionalhazards model, with the actual therapy as a time-dependent
covariate, and allowance for age, sex, ethnic group, and FPG
after 3 months’ diet.
THE LANCET • Vol 352 • September 12, 1998
Clinic visits
Patients were seen every month for the first 3 months and then
every 3 months or more frequently if required to attain control
criteria. Patients attended fasting for plasma glucose and other
biochemical measurements, blood pressure and bodyweight
were measured, and therapy was adjusted if necessary. Details
were recorded of actual therapies, hypoglycaemic episodes, and
home blood-glucose monitoring. At each visit, patients were
asked whether they had experienced hypoglycaemic symptoms.
Physicians recorded hypoglycaemic episodes as minor when the
patient was able to treat the symptoms unaided, or major if
third-party help or medical intervention was necessary. The
number of patients, in an allocation and taking the allocated
therapy, who had one or more minor or major hypoglycaemic
episodes in a year was recorded, and the mean over 10 years
calculated. Hypoglycaemic episodes in each year were analysed
both by intention to treat and by actual therapy.
Clinical endpoint analyses
The closing date for the study was Sept 30, 1997.
Endpoints were aggregated for analysis to keep to a minimum
the numbers of statistical tests.12 The three predefined primary
outcome analyses were the time to the first occurrence of: any
diabetes-related clinical endpoint (sudden death, death from
hyperglycaemia or hypoglycaemia, fatal or non-fatal myocardial
infarction, angina, heart failure, stroke, renal failure,
amputation [of at least one digit], vitreous haemorrhage,
retinopathy requiring photocoagulation, blindness in one
eye, or cataract extraction); diabetes-related death (death from
myocardial infarction, stroke, peripheral vascular disease,
renal disease, hypoglycaemia, or hyperglycaemia, and sudden
death); and all-cause mortality. Four additional clinical
endpoint aggregates were used to assess the effect of
therapies on different types of vascular disease in secondary
outcome analyses: myocardial infarction (fatal and non-fatal
857
ARTICLES
Figure 3: Median FPG, median HbA1c, mean change in bodyweight, and median change in fasting plasma insulin in cohorts of
patients followed up to 10 years by assigned treatment (shown by continuous lines)
Cross-sectional data at each year are shown by individual symbols for all patients assigned regimen.
and sudden death); stroke (fatal and non-fatal); amputation (of
at least one digit) or death due to peripheral vascular disease
(including death from gangrene); and microvascular
complications
(retinopathy
requiring
photocoagulation,
vitreous haemorrhage, and fatal or non-fatal renal failure).
Subclinical, surrogate variables1 were assessed every 3 years.
Biochemistry
Methods have been previously reported.1,17 The normal range
was
4·5–6·2%.
for
glycated
haemoglobin
(HbA1c)
Microalbuminuria has been defined for this study as urinary
albumin concentration above 50 mg/L and clinical grade
proteinuria as more than 300 mg/L.
Assignment
All randomisations were done at the level of the individual
patient, by means of therapy allocations in sealed opaque
envelopes, which were opened in sequence. The numerical
858
sequence of envelopes used, the dates they were opened, and
the therapies stipulated were monitored. No placebo was given.
Statistical analysis
Analyses were by intention to treat. Life-table analyses were
done with log-rank tests and hazard ratios, used to estimate
relative risks, were obtained from Cox proportional-hazards
models. For the primary and secondary outcome analyses of
clinical endpoint aggregates, 95% CIs are quoted. For single
endpoints 99% CIs are quoted, to make allowance for potential
type 1 errors.1 Further details are given in the accompanying
paper.1
Results
Intensive blood-glucose control with metformin versus
conventional treatment in overweight patients
Table 1 shows the baseline data for overweight patients
THE LANCET • Vol 352 • September 12, 1998
ARTICLES
Figure 4: Proportion of patients who reported one or more episodes of major hypoglycaemia or any hypoglycaemia per year,
assessed by actual therapy and by allocation (intention to treat)
Numbers of patients studied at 5 years, 10 years, and 15 years in actual therapy analysis=168, 60, and 6 for conventional group; 220, 101, and 6 for
metformin group; 235, 166, and 26 for insulin group; 148, 60, and 5 for chlorpropamide group; and 161, 71, and 6 for glibenclamide group.
at the time of randomisation to conventional treatment
or
intensive
treatment
with
chlorpropamide,
glibenclamide, insulin, or metformin. The mean bodymass index for overweight patients with type 2 diabetes
was 31·4 kg/m2 (SD 4·6); 99⋅5% of patients had bodymass index greater than 25 kg/m2, and 54⋅0% had bodymass index greater than 30 kg/m2.
The median follow-up (to the last known date at
which vital status was known or to the end of the trial)
was 10·7 years. Vital status was not known at the end of
the trial for 13 (1·8%) patients who had emigrated.
A further 43 (2·5%) patients could not be contacted in
the last year of the study for assessment of clinical
endpoints.
Figure 3 shows the median FPG and HbA1c in the
cohort of 482 patients with data available studied over
10 years and cross-sectional data for all those assigned
each therapy. In the metformin group there was a
decrease in FPG and HbA1c in the first year, with a
subsequent gradual rise in both variables. From 10
years, FPG in the metformin group approached that of
the conventional treatment group. The median HbA1c
during the 10 years of follow-up was 7·4% in the
metformin group and 8⋅0% in the conventional
treatment group. The patients assigned intensive control
with sulphonylurea or insulin had similar HbA1c to the
metformin group. The median HbA1c values in the
metformin group and conventional control group were
6·7% and 7·5%, respectively, in the first 5 years of
follow-up, 7·9% and 8·5% in the second 5 years, and
8·3% and 8·8% in the last 5 years. The cross-sectional
THE LANCET • Vol 352 • September 12, 1998
data, of all patients at each year, were similar to the
cohort data.
For the cohorts followed up for 10 years, the change
in bodyweight was similar in the metformin and
conventional control groups, and less than the increase
in bodyweight observed in patients assigned intensive
control with sulphonylureas or insulin. There was a
decrease in fasting plasma insulin in the patients
assigned metformin, which persisted throughout followup (figure 3).
Of the 4292 person-years of follow-up among
patients assigned conventional control, 2395 (56%)
were treated by diet. The remaining 44% of personyears required, as per protocol, additional non-intensive
pharmacological therapies. Of the 3682 person-years of
follow-up among the overweight patients assigned
metformin, 3035 (82%) were treated with metformin
alone or in combination. The median dose of metformin
was 2550 mg/day (IQR 1700–2550). For the
conventional control group, there were 3557 (83%) of
person-years with crossover to metformin therapy.
Figure 4 shows the proportion of patients per year
who had a major hypoglycaemic episode according to
actual therapy and intention to treat. The rate of any
hypoglycaemic episodes was higher in patients taking
metformin as allocated than in those on diet alone but
lower than the rates in those taking sulphonylureas as
allocated. The rate of hypoglycaemic episodes increased
over time among patients treated with insulin, as higher
insulin doses were required, and decreased among those
on sulphonylurea therapy, as glucose concentrations
859
ARTICLES
Figure 5: Kaplan-Meier plots in diet/metformin study for any
diabetes-related clinical endpoint and diabetes-related death
Intensive, in this figure, indicates chlorpropamide, glibenclamide, and
insulin groups. Similar plots of data for sulphonylurea/metformin study
are superimposed showing relative time of commencement.
increased. Over 10 years of follow-up among patients
taking therapy as allocated, the proportions of patients
per year who had one or more major hypoglycaemic
attacks
in
the
conventional,
chlorpropamide,
glibenclamide, insulin, and metformin groups were
0·7%, 0·6%, 2·5%, 0·3%, and 0% respectively; for any
hypoglycaemic episode the corresponding proportions
were 0·9%, 12·1%, 17·5%, 34·0%, and 4·2%.
Among all patients assigned treatments (intentionto-treat analyses), major hypoglycaemic episodes
occurred in 0·7%, 1·2%, 1·0%, 2·0%, and 0⋅6%,
respectively, of the conventional, chlorpropamide,
glibenclamide, insulin, and metformin groups, and any
hypoglycaemic episodes in 7·9%, 15·2%, 20·5%, 25·5%,
and 8·3%, respectively. Hypoglycaemic episodes in
patients on diet therapy were reactive hypoglycaemic
attacks, either after meals or, in some patients, after
termination of glucose infusions while in hospital (eg,
postoperatively).
860
Aggregate and single endpoints (diet vs metformin
study)
Patients assigned intensive blood-glucose control with
metformin had a 32% lower risk (p=0·0023) of
developing any diabetes-related endpoint than those
allocated conventional blood-glucose control (figures 5
and 6). These endpoints included macrovascular and
microvascular complications and represented the effect
of intensive policy with metformin on complication-free
survival. The group assigned metformin had a
significantly greater risk reduction than those assigned
intensive therapy with sulphonylurea or insulin
(p=0·0034).
The metformin group had a lower risk of diabetesrelated death than the conventional treatment group
(figures 5 and 6), with no significant difference between
the metformin group and those assigned therapy with
sulphonylurea or insulin. There were no deaths from
lactic acidosis.
Cardiovascular disease accounted for 62% of the total
mortality in the overweight patients in the conventional
treatment group. The metformin group had a 36%
lower risk (p=0·011) of all-cause mortality than the
conventional group (figure 6). There was a greater risk
reduction than in the groups assigned intensive therapy
with sulphonylurea or insulin (p=0·021). The
metformin group had a 39% lower risk (p=0·010) of
myocardial infarction than the conventional treatment
group, but did not differ from the other intensive
treatment group (figure 6). There were no significant
differences between the metformin group and the
conventional group in the other aggregate endpoints.
For all macrovascular diseases together (myocardial
infarction, sudden death, angina, stroke, and peripheral
disease), the metformin group had a 30% (5–48,
p=0·020) lower risk than the conventional treatment
group but did not differ significantly from the other
intensive groups.
Data for the single endpoints are shown in figures 7
and 8. There was no difference in the rate of death due
to non-diabetes-related endpoints (accidents, cancer,
other specified causes, or unknown causes).
Surrogate endpoints—The metformin group had a
lower rate of progression to retinopathy than the
conventional group, of borderline significance
(p=0·044), at 9 years; there was no difference at 12
years. The result was similar to that in the other
intensive therapy group. The proportion of patients with
urine albumin above 50 mg/L did not differ significantly
between the intensive treatment, metformin, and
conventional groups (24%, 23%, and 23% respectively).
There was no difference between the treatment groups
in any of the surrogate indices of macrovascular disease.
Addition of metformin in patients receiving
sulphonylurea
Table 2 shows the demographic data for the patients
whose response to maximum sulphonylurea treatment
was not adequate (FPG 6·1–15·0 mmol/L) and who
were assigned continuing intensive policy with
sulphonylurea alone or with early addition of
metformin. The mean body-mass index of normal and
overweight patients in this study was 29·6 kg/m2 (SD
5·5); 17% had body-mass index below 25 kg/m2 and
39% had values above 30 kg/m2.
THE LANCET • Vol 352 • September 12, 1998
ARTICLES
Figure 6: Incidence of clinical endpoints among patients assigned intensive control with metformin (n=342), intensive control with
chlorpropamide, glibenclamide, or insulin (intensive; n=951), or conventional control (n=411)
Relative risk (RR) is for metformin or intensive group compared with conventional group.
The median duration from the initial randomisation
to subsequent randomisation of addition or no addition
of metformin was 7·1 years. The median follow-up after
randomisation was 6·6 years. Vital status was not known
in ten (2%) patients who had emigrated and a further
five (1%) who could not be contacted.
Figure 9 shows the median FPG and HbA1c in the
cohorts studied for 4 years after second randomisation
to addition or no addition of metformin therapy
compared with data for all the overweight patients in the
comparison of intensive control with metformin and
conventional control. There was a decrease in FPG in
patients on sulphonylurea therapy who were assigned
addition of metformin, whereas FPG concentrations in
those on sulphonylurea therapy alone approached those
of overweight patients in the conventional treatment
group. HbA1c values in patients with addition of
metformin decreased initially but approached those of
the patients remaining on sulphonylurea alone after 3
years. The median HbA1c over 4 years in the cohort with
addition of metformin was 7·7% compared with 8·2% in
those on sulphonylurea alone. There were no significant
differences in bodyweight or plasma insulin between the
groups allocated addition of metformin or continued
sulphonylurea therapy alone.
The patients assigned addition of metformin took this
drug for 62% of their person-years of follow-up. For
those randomly assigned continuing sulphonylurea
alone, there were 75% of person-years without
metformin therapy.
Aggregate and single endpoints (addition of metformin
study)
Figure 10 shows the aggregates of endpoint data and
figure 11 the single endpoint data.
The addition of metformin to sulphonylurea was
associated with a 96% increased (p=0·039) risk of
diabetes-related death. Addition of metformin to
Figure 7: Kaplan-Meier plots in diet/metformin study for microvascular disease (renal failure or death from renal failure, retinopathy
requiring photocoagulation, or vitreous haemorrhage), myocardial infarction (non-fatal and fatal, including sudden death), stroke
(non-fatal and fatal) and cataract extraction
Similar plots of data for sulphonylurea/metformin study are superimposed showing relative time of commencement.
THE LANCET • Vol 352 • September 12, 1998
861
ARTICLES
Figure 8: Incidence of single endpoints in diet vs metformin study
Relative risk (RR) is for comparison with conventional control.
sulphonylurea therapy also increased the risk of death
from any cause (60% increase, p=0·041). There were no
significant differences between the groups for the other
aggregate endpoints. In a subgroup analysis, there was
no significant difference between patients allocated
metformin in addition to chlorpropamide or
glibenclamide (data not shown).
The data for the single endpoints are shown in
figure 11.
Combined analysis of both trials
Heterogeneity tests confirmed the different outcomes
between the two trials for any diabetes-related endpoint
(p=0·034), diabetes-related death (p=0·00256), and allcause mortality (p=0·0173), with a non-significant trend
for myocardial infarction (p=0·068). Figure 10 shows
the results for the two trials combined, with a 12%
reduced risk for any diabetes-related endpoint
(p=0·033). A formal meta-analysis gave similar results
for diabetes-related endpoints (observed minus expected
22·7, variance 104·9, p=0·026) and for myocardial
infarction (observed minus expected 12·2, variance
43·9, p=0·065).
Epidemiological analysis
The 4417 patients had 45 527 person-years of followup; 5181 (11%) of these person-years were treated with
sulphonylurea plus metformin therapy. 39 (8%) of the
490 diabetes-related deaths occurred while patients
were receiving sulphonylurea plus metformin therapy. A
Cox proportional-hazards model, with adjustment for
age, sex, ethnic group, and FPG after 3 months’ diet,
Sulphonylurea alone (n=269)
Sulphonylurea plus metformin (n=268)
All patients (n=537)
Demographic
Age (years)*
M/F
Ethnicity (%) Caucasian/Indian Asian/Afro-Caribbean/other
58 (9)
164 (61%)/108
77/13/10/0
59 (8)
158 (59%)/118
77/11/12/0
59 (9)
322 (60%)/226
77/11/11/1
Clinical
Weight (kg)*
Body-mass index (kg/m2)
Systolic blood pressure (mm Hg)*
Diastolic blood pressure (mm Hg)*
Smoking (%) never/ex/current
Alcohol (%) none/social/regular/dependent
Exercise (%) sedentary/moderately active/active/fit
82 (16)
29·4 (5·7)
138 (21)
81 (11)
31/40/29
37/44/18/0·4
14/38/45/3
83 (16)
29·7 (5·3)
140 (20)
83 (11)
35/40/24
32/51/16/1·1
22/35/39/4
83 (16)
29·6 (5·5)
139 (21)
82 (11)
33/40/27
34/52/13/0·8
18/37/42/3
Biochemical
FPG (mmol/L)†
HBA1c (%)*
Plasma insulin (pmol/L)‡
Triglyceride (mmol/L)‡
Total cholesterol (mmol/L)*
LDL cholesterol (mmol/L)*
HDL cholesterol (mmol/L)*
9·2 (7·8–10·9)
7·6 (1·8)
102 (58–180)
1·61 (0·91–2·86)
5·9 (1·0)
3·67 (0·96)
1·08 (0·28)
9·0 (7·6–11·3)
7·5 (1·7)
102 (58–181)
1·64 (0·89–3·04)
5·6 (1·1)
3·53 (0·93)
1·10 (0·30)
9·1 (7·7–11·1)
7·5 (1·7)
102 (58–181)
1·63 (0·90–2·95)
5·6 (1·1)
3·60 (0·95)
1·09 (0·29)
Medications
More than one aspirin daily (%)
Diuretic (%)
Others (%) digoxin/antihypertensives/lipid lowering/HRT or OC
6·4
13
1·9/24/0·4/0·8
4·5
16
1·5/25/0/0·4
5·5
15
1·7/25/0·4/0·8
Data are % of group, *mean (SD), †median (IQR), or ‡geometric mean ( 1 SD).
HRT=hormone replacement therapy; OC=oral contraceptive therapy.
Table 2: Baseline characteristics of patients assigned sulphonylurea treatment and subsequently randomised to continuing
sulphonylurea treatment alone or with early addition of metformin
862
THE LANCET • Vol 352 • September 12, 1998
ARTICLES
Figure 9: Median FPG and median HbA1c in cohorts of patients followed to 10 years from primary randomisation in diet vs metformin
study, and cohorts of patients followed to 4 years from second randomisation to sulphonylurea alone or sulphonylurea plus
metformin in sulphonylurea vs metformin study
with current therapies as a time-dependent variable,
showed a non-significant risk reduction in diabetesrelated death for sulphonylurea plus metformin
compared with all other treatments of 5% (95%CI -33
to 32, p=0·78).
Discussion
The main trial reported in this paper evaluated the
effect of metformin in diet-treated overweight patients
with type 2 diabetes. The study design parallels that in
the accompanying paper,1 comparing conventional
blood-glucose control primarily with diet alone and
intensive treatment with sulphonylurea or insulin. The
data shown here suggest that metformin therapy in diettreated overweight patients reduced the risk for any
diabetes-related endpoint, diabetes-related death, and
all-cause mortality. These possible benefits were not
seen in the second trial reported here, which suggests an
increased risk for diabetes-related deaths and all-cause
mortality when metformin is given in addition to
sulphonylurea
therapy
in
non-overweight
and
overweight patients. Because the difference in the effect
of metformin between diet-treated and sulphonylureatreated patients could be extremes of the play of chance,
a combined analysis of all the data was undertaken. This
showed that addition of metformin had a comparable
effect to that seen with intensive therapy with
sulphonylurea or insulin reported in the accompanying
paper1 with a net reduction of 19% in any diabetesrelated endpoint (p=0·033).
The trend to a reduced risk for microvascular
endpoints with metformin therapy was comparable to
Figure 10: Incidence of clinical endpoints in sulphonylurea vs metformin study and diet vs metformin study
Relative risk (RR) is for comparison with conventional or sulphonylurea alone. Results of a combined analysis of these two studies shown also.
THE LANCET • Vol 352 • September 12, 1998
863
ARTICLES
Figure 11: Incidence of single endpoints in sulphonylurea vs metformin study
Relative risk (RR) is for sulphonylurea plus metformin vs sulphonylurea alone.
that reported in the accompanying paper for intensive
glucose control1 but did not achieve statistical
significance.
Clinical use of metformin in overweight patients
In diet-treated overweight patients metformin similarly
improved HbA1c levels as with sulphonylurea and insulin
therapy but did not induce weight gain and was
associated with fewer episodes of hypoglycaemia. Given
the equivalent HbA1c levels obtained, the possible
additional benefit of metformin observed in overweight
diet-treated patients, of a reduced risk for any diabetesrelated endpoint, all mortality, and stroke is not
explicable on the basis of glycaemic control. The
improvements in the predominantly cardiovascular
outcomes seen with metformin may be due to the
decrease in PAI-1 that accompanies the metformininduced increase in insulin sensitivity.3 PAI-1 can
inhibit fibrinolysis; thus decrease in PAI-1 could lessen
the likelihood of extension of a thrombolysis. In
addition, metformin lowers systemic methylglyoxal
concentrations in patients with type 2 diabetes,18 which
suggests that it may have an aminoguanidine-like action.
However, these postulated mechanisms may not be
relevant since, in the combined analysis, the effect of
metformin on cardiovascular outcomes was not
substantiated.
Clinical use of metformin in patients already treated
with sulphonylurea
When metformin was prescribed in the trial in both
non-overweight and overweight patients already treated
with sulphonylurea there was a significant increase in
risk of diabetes-related death and all-cause mortality
rather than a beneficial effect on the primary outcome.
The different outcomes seen in these two trials may be
explained by differences in the patients studied. The
sulphonylurea-treated patients were on average 5 years
older; more hyperglycaemic (baseline median FPG 9·1
vs 8·1 mmol/L); less overweight; and followed up on
864
average for 5 years less. Secondly, it is important to note
that the differences in outcome relate to a relatively
small number of endpoints. The epidemiological
analysis did not corroborate an association of diabetesrelated deaths with combined sulphonylurea and
metformin therapy although the CIs were wide.
The UKPDS studied metformin primarily in obese
patients, since when the study started (1970s),
metformin was generally prescribed only in such
patients. Obesity is common among patients with type 2
diabetes.19 At entry to UKPDS, body-mass index was
above 25 kg/m2 in 75% of patients and above 30 kg/m2
in 35%.
Since metformin seems to give risk reduction of
diabetes-related endpoints in overweight patients with
type 2 diabetes, does not induce weight gain, and is
associated with fewer hypoglycaemic attacks than
sulphonylurea or insulin therapy,10 it could be chosen as
the first-line pharmacological therapy in such patients.
Although these findings may not apply to nonoverweight patients, metformin seems to lower
glycaemia in patients with type 2 diabetes, irrespective
of the degree of obesity.1
Conclusion
The addition of metformin in patients already treated
with sulphonylureas requires further study. On balance,
metformin treatment appears to be advantageous as a
first-line pharmacological therapy in diet-treated
overweight patients with type 2 diabetes.
UKPDS Study Organisation
Participating centres—Radcliffe Infirmary, Oxford; Royal Infirmary,
Aberdeen; Birmingham General Hospital; St George’s Hospital,
London; Hammersmith Hospital, London; Belfast City Hospital; North
Staffordshire Royal Infirmary, Stoke-on-Trent; Royal Victoria Hospital,
Belfast; St Helier Hospital, Carshalton; Whittington Hospital, London;
Norfolk and Norwich Hospital, Norwich; Lister Hospital, Stevenage;
Ipswich Hospital; Ninewells Hospital, Dundee; Northampton Hospital;
Torbay Hospital; Peterborough General Hospital; Scarborough
Hospital; Derbyshire Royal Infirmary; Manchester Royal Infirmary;
THE LANCET • Vol 352 • September 12, 1998
ARTICLES
Hope Hospital, Salford; Leicester General Hospital; Royal Devon and
Exeter Hospital.
Writing committee—Robert C Turner, Rury R Holman, Irene M
Stratton, Carole A Cull, David R Matthews, Susan E Manley, Valeria
Frighi, David Wright, Andrew Neil, Eva Kohner, Heather McElroy,
Charles Fox, David Hadden
7
8
9
Acknowledgments
We thank the patients and many NHS and non-NHS staff at the centres
for their cooperation.
Major grants for this study were obtained from the UK Medical
Research Council, British Diabetic Association, the UK Department of
Health, the National Eye Institute and the National Institute of
Digestive, Diabetes and Kidney Disease in the National Institutes of
Health, USA, the British Heart Foundation, Novo-Nordisk, Bayer,
Bristol Myers Squibb, Hoechst, Lilly, Lipha, and Farmitalia Carlo Erba.
Other funding companies and agencies, the supervising committees, and
all participating staff are listed in reference 11.
10
11
12
13
14
References
1
2
3
4
5
6
UKPDS Group. Intensive blood-glucose control with sulphonylureas
or insulin compared with conventional treatment and risk of
complications in patients with type 2 diabetes (UKPDS 33). Lancet
1998; 352: 837–53.
UKPDS Group. UK Prospective Diabetes Study 24: relative efficacy
of sulfonylurea, insulin and metformin therapy in newly diagnosed
non-insulin dependent diabetes with primary diet failure followed for
six years. Ann Intern Med 1998; 128: 165–75.
Cusi K, Consoli A, DeFronzo RA. Metabolic effects of metformin
on glucose and lactate metabolism in noninsulin-dependent diabetes
mellitus. J Clin Endocrinol Metab 1996; 81: 4059–67.
Bailey CJ. Biguanides and NIDDM. Diabetes Care 1992; 15:
755–72.
Nagi DK, Yudkin JS. Effects of metformin on insulin resistance, risk
factors for cardiovascular disease, and plasminogen activator
inhibitor in NIDDM subjects: a study of two ethnic groups. Diabetes
Care 1993; 16: 621–29.
Nolan JJ, Ludvik B, Beersden P, Joyce M, Olefsky J. Improvement in
glucose tolerance and insulin resistance in obese subjects treated
THE LANCET • Vol 352 • September 12, 1998
15
16
17
18
19
20
with troglitazone. N Engl J Med 1994; 331: 1188–93.
University Group Diabetes Program. A study of the effects of
hypoglycemic agents on vascular complications on patients with
adult-onset diabetes: V– evaluation of phenformin therapy. Diabetes
1975; 24 (suppl 1): 65–184.
Nattrass M, Alberti KG. Biguanides. Diabetologia 1978; 14:
71–74.
Bailey CJ, Turner RC. Metformin. N Engl J Med 1996; 334:
574–79.
UKPDS Group. UK Prospective Diabetes Study VIII: study design,
progress and performance. Diabetologia 1991; 34: 877–90.
UKPDS Group. UK Prospective Diabetes Study 16: overview of six
years’ therapy of type 2 diabetes—a progressive disease. Diabetes
1995; 44: 1249–58.
UKPDS Group. UK Prospective Diabetes Study 17: a nine-year
update of a randomized, controlled trial on the effect of improved
metabolic control on complications in non-insulin-dependent
diabetes mellitus. Ann Intern Med 1996; 124: 136–45.
Metropolitan Life Insurance Company. Net weight standard for men
and women. Statist Bull 1959; 40: 1–4.
UKPDS Group. UK Prospective Diabetes Study 28: a randomised
trial of efficacy of early addition of metformin in sulphonylureatreated non-insulin dependent diabetes. Diabetes Care 1998; 21:
87–92.
Rothman KJ. Modern epidemiology. Boston: Little, Brown, 1986.
Early Breast Cancer Trialists Collaborative Group. Treatment of
early breast cancer. Oxford: Oxford University Press, 1990.
UKPDS Group. UK Prospective Diabetes Study XI: biochemical
risk factors in type 2 diabetic patients at diagnosis compared with
age-matched normal subjects. Diabet Med 1994; 11: 534–44.
Beisswenger P, Howell S, Touchette A, Lal S, Szwergold B, Rohlf J.
Metformin reduces systemic methylgloxal levels in NIDDM. Diabetes
1997; 46 (suppl 1): 74A (abstr).
Modan M, Karasik A, Halkin H, et al. Effect of past and concurrent
body mass index on prevalence of glucose intolerance and type 2
(non-insulin-dependent) diabetes and on insulin response: the Israel
study of glucose intolerance, obesity and hypertension. Diabetologia
1986; 29: 82–89.
Yki-Jarvinen H, Nikkil K, Ryysy L, Tulokas T, Vanamo R, Hekkil
M. Comparison of bedtime insulin regimens in NIDDM: metformin
prevents insulin-induced weight gain. Diabetologia 1996; 39 (suppl
1): A33 (abstr).
865
Reviews/Commentaries/Position Statements
R E V I E W
A R T I C L E
Combination Therapies With Insulin in
Type 2 Diabetes
HANNELE YKI-JÄRVINEN, MD, FRCP1
T
he U.K. Prospective Diabetes Study
(UKPDS) demonstrated that intensive glucose control with insulin or
sulfonylureas markedly reduces the risk
of microvascular complications (1). For
myocardial infarction, the reduction in
risk (16% for a 0.9% decrease in HbA1c)
was of borderline significance but corresponded closely to epidemiological predictions (14% decrease for a 1% drop in
HbA1c) (2). These data demonstrated that
neither insulin nor sulfonylureas, despite
causing hyperinsulinemia and weight gain,
have adverse effects on cardiovascular outcome. Glycemic control deteriorated continuously, however, even in intensively
treated patients in the UKPDS (1).
In the UKPDS, the worsening of glycemic control has been attributed to the
natural course of type 2 diabetes and lack
of efficacy of current antihyperglycemic
therapies (1). Insulin therapy consisted of
a single injection of ultralente or isophane
insulin. If the daily dose exceeded 14 U,
regular insulin was added and homeglucose monitoring was encouraged (1).
Combination therapy regimens with insulin and oral agents were not used. We
now know that 14 U of long-acting insulin is insufficient to control fasting glycemia in most type 2 diabetic patients (3).
Since 1977, when the UKPDS was started,
several studies have tried to define the optimal insulin treatment regimen for type 2
diabetic patients. These studies are the focus of this review and include studies
comparing insulin alone to combination
therapy with insulin and sulfonylureas
(subject to meta-analyses in 1991 and
1992) (4,5) and more recent trials using
metformin, glitazones, or acarbose in insulin combination therapy regimens.
They do not contain data on cardiovascu-
lar end points but only on surrogate
markers of risk of micro- and macrovascular complications, mostly data on glycemia, body weight, insulin doses, lipids,
and in a few studies, also accurate data on
the frequency of hypoglycemias.
According to a Medline search (1966 –
2000), insulin alone has been compared
with insulin combination therapy in a total of 34 prospective studies that lasted at
least 2 months and reported data on HbA1
or HbA1c in type 2 diabetic patients. Studies comparing glycemic control, weight
gain, hypoglycemias, and insulin requirements between the two modes of treatment in insulin-naı̈ve patients are listed in
Table 1 and in previously insulin-treated
patients are listed in Table 2. The studies
have been ranked according to glycemic
control at the end of the trial.
GLYCEMIC CONTROL AND
INSULIN REQUIREMENTS
Insulin-naı̈ve and previously insulintreated patients.
In insulin-naı̈ve patients in a total of 15
comparisons (10 studies), glycemic control was similar in most (11 of 15) comparisons and better with the insulin
combination than the insulin-alone regimen in four comparisons (Table 1). In all
studies, the daily insulin dose was lower
with insulin combination therapy than
with insulin alone. The weighted mean
for the insulin-sparing effect of two drugs
(sulfonylureas and metformin) in addition to insulin was 62%, i.e., 1.5–2.0 times
that with regimens combining either metformin alone (⫺32%) or sulfonylureas
alone (⫺42%) (Table 1) with insulin. These
data imply that oral agents still have significant glucose-lowering effects even in
patients who are poorly controlled on oral
drugs. One may also predict from these
data that if the insulin dose is lowered less
than ⬃30% when patients are transferred
from insulin alone to insulin combined
with sulfonylurea or metformin, glycemic
control will be better during insulin combination therapy. This is documented by
analysis of data from studies in previously
insulin-treated patients (Table 2). In these
comparisons, glycemic control was better
in most (19 of 25) comparisons, but the
insulin dose was decreased by only 19%
in the combination regimens using metformin and by 21% in comparisons using
insulin and sulfonylureas (Table 2). Thus,
although in most comparisons (30 of 45)
glycemic control has been better with insulin combination therapy regimens than
with insulin alone, the difference may be
at least partly explained by how the insulin dose has been decreased during insulin combination therapy. All glitazones
have improved glycemic control when
added to previous insulin treatment (Table 2). In a study directly comparing the
insulin-sparing effects of troglitazone
(600 mg/day) and metformin (1,700 mg/
day), troglitazone had a greater (⫺53%)
insulin-sparing effect than metformin
(⫺31%). This was explained by insulinsensitizing effects of troglitazone but not
metformin (6).
WEIGHT GAIN
From the 1Department of Medicine, Division of Diabetes, University of Helsinki, Helsinki, Finland.
Address correspondence and reprint requests to Hannele Yki-Järvinen, MD, FRCP, Department of Medicine, Division of Diabetes, P.O. Box 340, 00029 HUS, Helsinki, Finland. E-mail: ykijarvi@helsinki.fi.
Received for publication 19 October 2000 and accepted in revised form 2 January 2001.
Abbreviations: GADA, GAD antibodies; UKPDS, U.K. Prospective Diabetes Study.
A table elsewhere in this issue shows conventional and Système International (SI) units and conversion
factors for many substances.
What determines weight gain during
insulin therapy?
Although most patients with type 2 diabetes are overweight, weight loss precedes the diagnosis of type 2 diabetes (7).
This weight loss is due to hyperglycemiainduced wasting of energy, as glucosuria
and as energy used to overproduce glucose (8). When glucose control is improved with insulin and/or sulfonylureas,
energy loss in the urine decreases or
ceases, weight increases, and basal metabolic rate (kJ/min) (8 –10) and dietary intake (11) remain unchanged. The increase
in body weight increases basal metabolic
rate, but this is counterbalanced by improved glycemic control, which decreases
basal metabolic rate because less energy is
758
DIABETES CARE, VOLUME 24, NUMBER 4, APRIL 2001
● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●
DIABETES CARE, VOLUME 24, NUMBER 4, APRIL 2001
Metformin alone
MET ⫹ bedtime NPH
Metformin and sulfonylureas
GLYB ⫹ MET ⫹ bedtime NPH
GLYB ⫹ MET ⫹ morning NPH
GLYB ⫹ MET ⫹ bedtime NPH
GLYB ⫹ MET ⫹ bedtime NPH
Weighted mean
Sulfonylurea regimens
GLIMEP ⫹ bedtime 30/70
GLYB ⫹ bedtime NPH
GLYB ⫹ bedtime NPH
GLYB ⫹ morning NPH
GLYB ⫹ Ins
GLYB ⫹ lispro t.i.d.
GLYB ⫹ bedtime NPH
GLYB ⫹ Ins
GLYB ⫹ Ins
GLICL ⫹ Ins
Weighted mean
Combination
regimen
6
12
6
6
6
2
2
4
4
12
12
3
3
6
12
Duration
(months)
7.6% Comb
7.8% Comb
8.1% Comb
8.2% Ins
8.4% Comb
8.4% Comb
8.5% Ins
8.8% Comb
9.8% Comb
11.8% Ins§
7.6% Comb
7.7% Comb
8.0% Comb
8.4% Ins
7.2% Comb
End
HbA1c*
No difference
No difference
No difference
No difference
No difference
No difference
No difference
Better with GLYB
Better with GLYB
Better with GLICL
No difference
No difference
No difference
No difference
Better with MET‡
Glycemia
Yes
Yes
No
No
Yes
No
No
Yes
Yes
No
Yes
No
No
No
Yes
Placebo
control
Less with MET‡
No difference
No difference
Less with MET
Less with MET
No difference
No difference
No difference
No difference
No difference
No difference
Less with GLYB
No difference
Less with Ins
—
Parallel
Parallel
Parallel
Parallel
Parallel
Parallel
Parallel
Parallel
Parallel
Parallel
Parallel
Crossover
Crossover
Parallel
Weight
gain
Parallel
Crossover/
parallel
More with GLYB
No difference
No difference
No difference
ND
No difference
No difference
—
—
—
No difference
No difference
No difference
No difference
No difference
Hypoglycemia
⫺37
⫺55
⫺38
⫺33
⫺43
⫺36
⫺56
⫺50
⫺21
⫺35
⫺42
⫺62
⫺58
⫺44
⫺74
⫺62
⫺32
Difference in
insulin dose (%)†
The trials are grouped according to the oral agent used and then ranked within these groups based on glycemic control at the end of treatment with the better regimen. Only trials lasting 2 months or longer
are included; *HbA1c at the end of treatment in the group with better control (even if not significantly better in one group versus the other); †% difference in insulin doses at the end of treatment with a
combination regimen versus insulin alone; ‡significant difference; §HbA1 Comb, combination regimen; GLICL, gliclazide; GLIMEP, glimepiride; GLYB, glyburide; Ins, regimen containing insulin alone; MET,
metformin; 30/70 ⫽ an insulin mixture containing 30% regular insulin and 70% NPH.
25
12
29
29
55
32
32
28
27
56
12
15
15
16
12
Reference
no.
Table 1—Studies comparing combination treatment regimens with insulin to insulin alone in insulin-naive type 2 diabetic patients
Yki-Järvinen
needed for glucose overproduction. Because dietary intake remains unchanged
(11), weight gain is proportional to reduction of glucosuria and can indeed be
predicted based on fasting glucose concentrations (11). Because glucosuria
appears when the fasting glucose concentration exceeds 10 –12 mmol/l, weight
gain is inevitable if insulin therapy is postponed until significant glucosuria occurs.
In our experience, a 5-mmol/l (90-mg/dl)
decrease in fasting glucose or a decrease
in HbA1c by 2.5% from a baseline of 15
mmol/l (270 mg/dl) is associated with a
5-kg weight gain during 1 year (or 2
kg/1% decrease in HbA1c) (11). Thus, the
main predictors of weight gain are initial
glycemia and its response to treatment
(11). The patient with poor glycemic control before initiation of insulin therapy
but with a good treatment response is at
greatest risk for weight gain.
Choice of oral agent and weight gain
in insulin-naı̈ve patients.
Only one trial has compared the
combination of insulin with that of insulin and metformin alone in previously
insulin-naı̈ve patients (12). In this
study, which lasted 12 months, the bedtime insulin-metformin regimen was
superior to three other bedtime insulin
regimens with respect to glycemic control, weight gain, and hypoglycemias
(Table 1) (12). The ability of metformin
to counteract weight gain and improve
glycemia, when combined with insulin,
has been confirmed in abstract reports
(13,14). In these studies, weight gain
was less despite comparable glycemia
(13) or weight gain was similar despite
better glycemic control (14) in patients
using metformin and insulin compared
with those using insulin and sulfonylureas or insulin alone. Data are heterogenous regarding the ability of metformin to
influence weight gain when combined
with both insulin and sulfonylureas, compared with regimens containing insulin
alone (Table 1). In two comparisons,
weight gain was less with the combination
regimen than with insulin alone (15,16),
whereas two other comparisons revealed
no difference (12,15) (Table 1, Fig. 1).
The ability of metformin to counteract
weight gain during insulin combination
therapy has been attributed to a decrease
in dietary intake (11).
759
760
Metformin regimens
MET ⫹ insulin
MET ⫹ insulin
MET ⫹ insulin
MET ⫹ insulin
Weighted mean
Sulfonylurea regimens
GLYB ⫹ insulin
GLYB ⫹ insulin
SU ⫹ insulin
GLYB ⫹ insulin
TOLAZ ⫹ insulin
GLYB ⫺ insulin
GLYB ⫹ insulin
GLYB ⫹ insulin
GLIP ⫹ insulin
GLYB ⫹ insulin
GLYB ⫹ insulin
GLYB insulin
GLYB ⫹ insulin
TOLAZ ⫹ insulin
GLYB ⫹ insulin
GLYB ⫹ insulin
Weighted mean
Glitazone
ROSI ⫹ insulin
TRO ⫹ insulin
PIO ⫹ insulin
␣-Glucosidase inhibitor
ACARB ⫹ insulin
ACARB ⫹ insulin
Combination
regimen
6
12
6
6
4
3
3
12
4
3
12
3
4
3
4
11
2
2
2
12
2
6
4
3
6
Duration
(months)
8.3%
7.3%
7.8%
7.9%
8.6%
6.0% Comb
7.0% Comb
7.5% Comb
8.3% Comb§
8.8% Comb§
8.8% Comb§
9.1% Comb
9.6% Comb
9.8% Comb§
10.2 Comb§
10.3% Ins§
11.0% Comb§
12.4% Comb§
12.6% Comb§
12.9% Ins
13.0% Comb§
6.5% Comb
7.8% Comb
7.8% Comb
9.8% Comb
End HbA1c*
or HbAI
Yes
Yes
Yes
Yes
Yes
No
Yes
No
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
No
Yes
Yes
Placebo
Parallel
Parallel
Parallel
Parallel
Parallel
Crossover
Parallel
Parallel
Crossover
Crossover
Parallel
Crossover
Parallel
Crossover
Crossover
Crossover
Crossover
Crossover
Crossover
Crossover
Crossover
Parallel
Parallel
Crossover
Parallel
Parallel/
crossover
Better with ACARB
Better with ACARB
Better with ROSI
Better with TRO
Better with PIO
No difference
Better with GLYB
No difference
Better with GLYB
Better with GLYB
Better with GLYB
Better with GLYB
Better with GLYB
No difference
Better with GLYB
Better with GLYB
No difference
Better with GLYB
No difference
No difference
Better with GLYB
Better† with MET
Better with MET
Better with MET
Better with MET
Glycemia
—
No difference
More with ROSI
More with TRO
More with PIO
No difference
No difference
No difference
No difference
Fixed
—
—
—
—
No difference
No difference
No difference
No difference
—
No difference
No difference
Less with MET‡
Less with MET
No difference
ND
Weight gain
No difference
No difference
More with ROSI
More with TRO
More with PIO
—
—
—
More with GLY
—
—
—
—
—
—
—
—
—
—
—
—
Less with MET‡
—
—
—
Hypoglycemias
Fixed
—
⫾0
⫺46
—
⫺25
⫺20
⫺35
⫺20
⫺23
⫺47
⫺7
Fixed
⫺35
⫺3
⫺7
⫺2
Fixed
Fixed
⫺22
⫾0
⫺21㛳
⫺23
⫺26
⫺3
⫺20
⫺19
Difference in
insulin dose (%)†
The trials are grouped according to the oral agent used and then ranked within these groups based on glycemic control at the end of treatment with the better regimen. Only trials lasting 2 months or longer
are included. *HbA1c at the end of treatment in the group with better control (even if not significantly better in one group versus the other); †% difference in insulin doses at the end of treatment with a
combination regimen versus insulin alone; ‡statistically significant difference between insulin combination therapy versus insulin alone; §HbA1 reference range higher than that for HbA1c; 㛳trials in which the
insulin dose was fixed are not included in the calculation of the weighted mean. ACARB, acarbose; Comb, combination regimen; GLIP, glipizide; Irs, regimen containing insulin; PIO, pioglitazone; ROSI,
rosiglitazone; SU, various sulfonylureas; TOLAZ, tolazamide; TRO, troglitazone.
65
66
54
52
53
45
58
46
26
47
59
60
61
62
49
48
50
44
63
64
51
24
42
43
57
Reference
no.
Table 2—Studies comparing combination treatment regimens with insulin to insulin alone in previously insulin-treated type 2 diabetic patients
Combination therapies with insulin in type 2 diabetes
DIABETES CARE, VOLUME 24, NUMBER 4, APRIL 2001
Yki-Järvinen
Figure 1—Weight gain in previously insulin-treated patients during treatment with insulin combination regimens containing metformin (MET) (24,42,43), various sulfonylureas (SU) (26,44 –
51), and glitazones (GLIT) (52–54) and in insulin-naı̈ve patients treated with insulin and MET
(12), SU⫹MET (12,15,16), and SU (12,25,27–29).
Choice of oral agent and weight gain
in previously insulin-treated patients.
Switching patients from treatment with
insulin alone to insulin combination
therapy with metformin has been associated with less weight gain in two of
three studies, whereas no difference was
found in any of the 16 comparisons in
which insulin plus sulfonylurea was
compared with insulin alone, despite
better control in 10 of 16 comparisons
(Table 2). It is unclear whether this is
because weight was not accurately recorded or because the larger dose of exogenous insulin or the greater number
of insulin injections used in insulin
alone as compared with the insulin
combination regimen had independent
weight-promoting effects. In studies
comparing insulin-glitazone treatment
with insulin alone, glycemic control was
better in each study with the insulinglitazone combination than with insulin
alone. Better glycemic control was also
associated with greater weight gain in
each of the three studies. Although the
amount of weight gain relative to the
improvement in glycemic control
seemed slightly greater than with sulfonylureas (Fig. 1), data on insulinglitazone combination therapy are still
sparse. The significance of weight gain
with glitazones is also difficult to judge,
because glitazones may be beneficial in
redistributing fat from visceral to subDIABETES CARE, VOLUME 24, NUMBER 4, APRIL 2001
cutaneous sites (17–19). A small fraction of weight gain with glitazones
could be due to peripheral edema
(20,21).
HYPOGLYCEMIAS
Frequency of hypoglycemias in type
1 versus type 2 diabetes.
In both patients with type 1 (22) and type 2
(12) diabetes, the frequency of hypoglycemias is inversely proportional to glycemic
control. In the Diabetes Control and Complications Trial, in patients with HbA1c between 7 and 8%, the frequency of severe
hypoglycemias requiring assistance in the
provision of treatment was 0.62 per patient
per year (22). In contrast, in the FINFAT
study (12), the Kumamoto study (23), or
other studies of type 2 diabetes in which
comparable glycemic control was achieved
(24,25), there were no severe hypoglycemias. The frequency of biochemical hypoglycemias (blood glucose ⬍3.5 mmol/l)
was 1.9 per patient per year in patients
treated with insulin plus metformin and approximately twice as high in the other
groups in the FINFAT study (12). In the
latter study, HbA1c averaged between 7
and 8% in all groups for 1 year. These
data, although derived from separate
studies, suggest that hypoglycemias are
much less of a problem in type 2 diabetic
patients than in type 1 diabetic patients.
Does the oral agent influence the
occurrence of hypoglycemias
independent of glycemic control?
The occurrence of hypoglycemia has been
sparsely reported [eight comparisons in
insulin-naı̈ve patients (Table 1) and five
comparisons in previously insulin-treated
patients (Table 2)]. In insulin-naı̈ve patients, use of insulin combination therapy
with metformin has been associated with
less hypoglycemias than with insulin
alone, despite better glycemic control
with the insulin-combination regimen
(12). No difference was observed between
insulin-alone and insulin-sulfonylurea
regimens in five of seven studies; in two
studies (25,26), there were more cases of
hypoglycemia with insulin and sulfonylurea than with insulin alone (Tables 1 and
2). No difference in the incidence of hypoglycemia was observed between insulin
alone compared with insulin plus sulfonylurea and metformin regimens (Tables
1 and 2). In the latter studies, there was
also no difference in glycemic control. In
all three studies comparing insulinglitazone combination therapy to insulin
alone, the frequency of hypoglycemia was
higher and glycemic control was better
with the insulin-combination regimen.
These data suggest that with the possible
exception of metformin, use of insulin
combination therapy is accompanied by a
similar frequency of hypoglycemia than is
use of insulin alone.
CHANGES IN SERUM
TRIGLYCERIDES AND
OTHER LIPIDS AND
LIPOPROTEINS
Insulin-naı̈ve patients.
Data on changes in serum triglycerides
and glycemia in insulin-naı̈ve patients are
summarized in Table 3. As judged from
the weighted means of insulin-alone regimens, a decrease in HbA1c from ⬃10 to
8% (i.e., by 2%) is associated with a 0.7to 0.8-mmol/l decrease in serum triglycerides from an initial concentration of
2.4 –2.7 mmol/l. With insulin combination therapy regimens, a decrease of
HbA1c by 2% decreases serum triglycerides by 0.4 – 0.6 mmol/l (Table 3). In all
except one study, insulin alone lowered
serum triglycerides slightly more than
insulin combination therapy, although
there was no significant difference in the
lowering of serum triglycerides with the
two modes of therapy in any of the studies
761
Combination therapies with insulin in type 2 diabetes
Table 3—Changes in glycosylated hemoglobin and serum triglycerides during treatment with insulin alone as compared with combination
therapies with insulin and oral agents in insulin-naive patients
Ref.
no.
12
Combination
regimen
Metformin alone
MET ⫹ bedtime NPH
Baseline Change Baseline Change Baseline Change Baseline Change Better† Better†
Duration HbA1c in HbA1c HbA1c in HbA1c S-Tg in S-Tg S-Tg in S-Tg regimen regimen
S-Tg glycemia
Comb Comb
Ins
Ins
(months)
Ins
Ins
Comb
Comb
12
9.9
⫺2.0‡
9.8
⫺2.5‡
2.6
⫺0.9‡
2.4
⫺0.7‡
NS
Comb
12
15
15
16
Metformin and sulfonylureas
GLYB ⫹ MET ⫹ bedtime NPH
GLYB ⫹ MET ⫹ morning NPH
GLYB ⫹ MET ⫹ bedtime NPH
GLYB ⫾ MET ⫹ bedtime NPH
Weighted mean
12
3
3
6
9.9
9.6
9.6
10.7
9.9
⫺2.0‡
⫺1.6‡
⫺1.6‡
⫺2.3‡
⫺1.9
9.9
9.5
9.9
10.2
9.9
⫺2.1‡
⫺1.7‡
⫺1.9‡
⫺1.5‡
⫺1.9
2.6
2.4
2.4
1.8
2.4
⫺0.9‡
⫺0.6‡
⫺0.6‡
⫺0.4‡
⫺0.7
2.3
2.6
2.5
2.1
2.4
⫺0.4‡
⫺0.3
⫺0.6‡
⫺0.2
⫺0.4
NS
NS
NS
NS
NS
NS
NS
NS
25
12
29
29
55
28
27
Sulfonylurea regimens
GLIMEP ⫹ bedtime 30/70
GLYB ⫹ bedtime NPH
GLYB ⫹ bedtime NPH
GLYB ⫹ morning NPH
GLYB ⫹ insulin
GLYB ⫹ insulin
GLYB ⫹ insulin
Weighted mean
6
12
6
6
6
4
4
9.9
9.9
11.2
11.2
11.5*
9.7*
10.4*
10.5
⫺2.0‡
⫺2.0‡
⫺3.0‡
⫺3.0‡
⫺2.2‡
⫺0.8
0.2
⫺2.2
9.7
9.8
10.5
11.1
10.4*
10.1*
10.6*
10.2
⫺2.1‡
⫺2.0‡
⫺2.3‡
⫺2.6‡
⫺2.2‡
⫺1.3
⫺0.8‡
⫺2.1
3.1
2.6
2.4
2.4
2.0
3.1
3.6
2.7
⫺1.0‡
⫺0.8‡
⫺0.7‡
⫺0.7‡
⫺0.4
⫺0.6
⫺1.1‡
⫺0.8
3.2
2.7
2.0
2.2
2.2
2.8
3.6
2.7
⫺0.3‡
⫺0.8‡
⫺0.3‡
⫺0.3
⫺0.3
⫺0.8‡
⫺1.2‡
⫺0.6
NS
NS
NS
NS
NS
Comb
NS
NS
NS
NS
NS
NS
Comb
Comb
The trials are grouped according to the oral agent used and then ranked within these groups based on glycemic control at the end of treatment with the better regimen.
Only trials lasting at least 2 months are included. *HbA1 value; †denotes a statistically significant difference between insulin combination therapy versus insulin
alone; ‡significant difference at the end versus start of the treatment period. Comb, combination regimen; Ins, regimen containing insulin alone; S-Tg, serum
triglycerides (mmol/l). For other abbreviations, see Table 1.
(Table 3). LDL and HDL cholesterol concentrations remained unchanged in all
studies, with no differences between regimens (12,15,16,25,27–29).
Previously insulin-treated patients.
As summarized in Table 4, the greater improvement in glycemic control with insulin combination therapy than with insulin
alone in 11 of 14 studies has not been
consistently (4 of 11 studies) associated
with a greater decrease in serum triglycerides. These data demonstrate that factors
other than average glucose concentrations
determine the degree of lowering of serum triglycerides. Overall, the available
comparisons of changes in serum lipid
and lipoprotein concentrations in both
insulin-naı̈ve and previously insulintreated patients do not allow definitive
conclusions and do not support choice of
one treatment regimen over another.
BLOOD PRESSURE
Regarding blood pressure, in a follow-up
study of the patients participating in the
FINMIS study (15,30), blood pressure increased significantly in the entire group of
100 patients during 1 year. Weight gain
correlated both with the increase in blood
pressure and with an increase in the LDL
762
cholesterol concentrations (30). Three
shorter studies reported data on blood
pressure (15,25,27) but found no
changes in blood pressure or differences
between regimens.
SPECIAL QUESTIONS
Choice of insulin regimen during
insulin combination therapy:
NPH insulin or insulin glargine?
Regarding basal insulinization, the commonly used intermediate-acting insulin
(NPH) is not ideal for once-daily use. In
the FINMIS study, in patients with type 2
diabetes with a mean BMI of 29 kg/m2,
injection of NPH insulin at 9:00 P.M. resulted in maximal glucose lowering between 4:00 and 8:00 A.M., but the effect
was gone by 3:00 P.M., i.e., 18 h after the
injection, and dinnertime glucose concentrations were unnecessarily high. The
recently approved long-acting insulin analog insulin glargine seems to overcome
these problems. In a study comparing
NPH plus oral agents to insulin glargine
plus oral agents in 423 insulin-naı̈ve type
2 diabetic patients for 1 year, all hypoglycemias were 35% lower and nocturnal hypoglycemias were 56% lower with insulin
glargine than with NPH (Fig. 2). Dinner-
time glucose levels were also significantly
lower with insulin glargine than with
NPH (Fig. 2).
Regular insulin or short-acting
insulin analogs compared with NPH
during combination therapy.
Regular insulin three times per day plus a
sulfonylurea has been compared with a
single injection of NPH taken at bedtime
and a sulfonylurea. No difference in glycemic control was found, but weight gain
was significantly greater with three injections of regular insulin than with a single
injection of bedtime NPH insulin (31).
Greater weight gain with no difference in
glycemic control has also been reported
with three injections of lispro plus sulfonylurea compared with NPH plus sulfonylurea (32) (Table 5).
Timing of the intermediate-acting
insulin injection.
The pros and cons of timing of the intermediate-acting insulin injection has been
examined in three studies (15,33,34). In
two studies, a bedtime injection was recommended because it resulted in less
weight gain (15) or less hypoglycemias
(34) than a morning injection. In the third
DIABETES CARE, VOLUME 24, NUMBER 4, APRIL 2001
Yki-Järvinen
Table 4—Changes in glycosylated hemoglobin and serum triglycerides during treatment with insulin alone as compared with insulin combination therapy in previously insulin-treated patients
Ref.
no.
Combination
regimen
24
42
43
57
Metformin
MET ⫹ insulin
MET ⫹ insulin
MET ⫹ insulin
MET ⫹ insulin
Weighted mean
45
46
26
47
48
61
51
Sulfonylurea
GLYB ⫹ insulin
SU ⫹ insulin
GLYB ⫹ insulin
TOLAZ ⫹ insulin
GLYB ⫹ insulin
GLYB ⫹ insulin
GLYB ⫹ insulin
Weighted mean
52
54
Glitazones
TRO ⫹ insulin
ROSI ⫹ insulin
Weighted mean
65
␣-Glucosidase inhibitor
ACARB ⫹ insulin
Duration
(months)
Baseline Change Baseline Change Baseline Change Baseline Change Better† Better†
in S-Tg regimen regimen
S-Tg
in S-Tg
HbA1c in HbA1c HbA1c in HbA1c S-Tg
glycemia
S-Tg
Comb Comb
Ins
Ins
Ins
Ins
Comb
Comb
6
4
6
3
9
9.6
11.5
8.9
9.8
⫺1.6§
0.0
⫺0.2
⫺0.5§
⫺0.6
9.1
9.6
11.7
8.9
10.0
⫺2.5§
⫺1.9§
⫺1.9
⫺1.1§
⫺1.9
2.5
2.4
2.8
2.2
2.5
⫺0.4
⫺0.1
⫺0.0
⫺1.0§
⫺0.4
2.3
2.0
2.9
2.2
2.4
⫺0.1
⫺0.4§
⫺0.3§
⫺0.9§
⫺0.4
NS
Comb
Comb
NS
Comb
Comb
Comb
Comb
3
12
4
3
11
4
2
6.7
10.2
9.2*
10.7*
10.3
10.4
14.0*
10.1
⫺0.4
⫺2.4§
⫺0.1
⫺1.5§
⫺1.3§
0.0
⫺0.6
⫺1.0
6.3
9.8
9.2*
10.7*
11.1
10.9
14.0*
10.1
⫺0.3
⫺2.3§
⫺0.9§
⫺1.9
⫺2.0§
⫺1.3§
⫺1.0
⫺1.4
1.5
2.3
1.2
2.1
1.7
1.4
2.1
1.8
⫺0.08
⫺0.6
0.1
0.0
0.1
⫺0.1
⫺0.4
⫺0.1
1.5
2.5
1.2
2.1
2.4
1.8
2.1
2.0
0.2
⫺0.8
0.1§
⫺0.5§
⫺0.7
⫺0.1
⫺0.2
⫺0.2
NS
NS
Ins
Comb
NS
NS
NS
NS
NS
Comb
Comb
NS
Comb
Comb
6
6
8.9
9.4
9.2
0.1
⫺0.1
⫺0.0
9.0
9.3
9.2
⫺1.2
⫺0.4
⫺1.3
2.6
3.0
2.8
⫺0.5
⫺0.3
0.1
2.5
2.7
2.6
⫺0.1
⫺0.4
⫺0.2
NS
NS
Comb
Comb
6
8.7
0.1
8.8
⫺0.6§
2.1†
—
2.2‡
—
Comb
Comb
The trials are grouped as in Table 3. Only trials lasting at least 2 months are included. *HbA1; †statistically significant difference between insulin combination therapy
versus insulin alone; ‡serum triglycerides 120 min after a standardized meal challenge; §significant difference at the end versus start of the treatment period. There
were no significant differences between changes in serum triglycerides with insulin alone versus insulin combination therapy. Abbreviations as in Tables 2 and 3.
study, no differences in glycemia or
weight gain were found (33).
PREDICTION OF INSULIN
REQUIREMENTS
Variation in hepatic insulin sensitivity
seems to be much more important than
insulin absorption in determining insulin
requirements during combination therapy with NPH insulin (3). Hepatic insulin
sensitivity cannot be routinely measured
but correlates with various indexes of
obesity (3). In type 2 diabetic patients
with a mean BMI of 29 kg/m2, to achieve
an average HbA1c of ⬃7.5% from a baseline value of 10%, the mean bedtime NPH
insulin dose for body weights of 70, 80,
90, and 100 kg has been 0.2, 0.3, 0.4, and
0.5 IU/kg (3,11). However, interindividual variation is large and has varied
20-fold between 8 and 168 IU per day
(12), which implies that these average
predictions are not accurate enough to be
used on an individual level.
Autoantibodies to glutamic acid deDIABETES CARE, VOLUME 24, NUMBER 4, APRIL 2001
carboxylase (GADA) predict an increased
likelihood of insulin requirement in both
young and old adults with type 2 diabetes
(35). In 3,672 newly diagnosed patients
in the UKPDS, 34% of those aged 25–34
years and 7% of those aged 55– 65 years
had GADA (35). Among patients older
than 55 years at diagnosis, 34% of those
with GADA and 5% with autoantibodies
to neither GADA nor islet cell cytoplasm
required insulin therapy. In these older
patients, only the presence of GADA but
not phenotypic features such as BMI predicted insulin requirement. There are no
studies comparing insulin combination
regimens with insulin alone in these patients, who are often classified as having
type 2 diabetes but actually have type 1
diabetes (36). In patients in whom signs
of absolute insulin deficiency (rapid
weight loss, ketonuria) ultimately develop, the presence of GADA may guide
the choice of a basal-bolus–type full insulin-replacement regimen (37).
PREDICTORS OF A
GLYCEMIC RESPONSE TO
INSULIN COMBINATION
THERAPY
In studies in which the insulin dose is titrated aggressively to reach glycemic targets, the decrease in HbA1c will be directly
proportional to its initial level. Of other
factors, obesity predicts a poor response
to any type of insulin therapy, especially if
insufficient doses of insulin are used
(30,38). In addition, and as discussed
above, GADA may predict poor response
to combination therapy.
PRACTICAL ALGORITHM TO
INITIATE INSULIN THERAPY
Initiation of insulin therapy on an ambulatory basis in type 2 diabetic patients has
been shown to be as safe and effective as
an inpatient program (39). Regardless of
the insulin treatment regimen chosen, the
insulin dose should be adjusted to reach
glycemic targets. Considering the large
interindividual variation in insulin re763
Combination therapies with insulin in type 2 diabetes
lin combination therapy with NPH or insulin glargine is used. However, because
especially NPH insulin is unable to adequately control postdinner glycemia, fasting glucose must be in the normal range
(4 – 6 mmol/l) for average glycemic control, measured using HbA1c to be ⬍7.5%.
A simple method of initiating insulin therapy, developed based on experience from
the FINFAT study, is shown in Table 5 (12).
The patient is assumed to be insulin-naı̈ve
and on maximal doses of sulfonylureas
and metformin. The recommendation to
discontinue the sulfonylurea (glyburide)
but not metformin after insulin combination therapy is started is based on the inability of some patients to adequately
titrate the dose of bedtime NPH because
of hypoglycemia (12). An increase in mild
hypoglycemias was also reported by Riddle and Schneider (25) with glimepiride
combined with a single injection of 30/70
insulin at 6:00 P.M. compared with two injections of 30/70 insulin. Hypoglycemias
may not be a problem with peakless insulins such as insulin glargine (40). Discontinuation of sulfonylurea when insulin
therapy is started may retard achievement
of good glycemic control unless the insulin dose is rapidly increased (12,25).
Glitazones could be an additional or alternative component in the oral hypoglycemic agent regimen, but there are no
studies in insulin-naı̈ve patients.
Figure 2—Upper panel: Diurnal glucose profiles after 52 weeks of treatment of 423 patients
using oral hypoglycemic agents with either insulin glargine (F) or NPH (E). Lower panel: The
percentage of patients experiencing any symptomatic hypoglycemia (ALL) or nocturnal hypoglycemia (NOCTURNAL) in the same study (40).
quirements, it is difficult to define the correct insulin dose by performing dose
adjustments only at outpatient visits, unless these are very frequent. In our experience of treating insulin-naı̈ve patients
(12,15), the best method of defining the
insulin dose is teaching the patient to selfadjust the dose based on results of home
glucose monitoring. This is easiest to perform if the dose adjustment is based on
764
measurement of fasting plasma glucose
only. Fasting plasma glucose is not influenced by size, composition, or rate of
absorption of meals as much as by postprandial glucose levels, and its measurement does not interfere with daily activities. The maximal action of NPH given at
bedtime is exerted on fasting glucose,
which, therefore, is a particularly suitable
target for titration of the dose when insu-
CONCLUDING REMARKS
Against the emerging epidemic of type 2 diabetes, studies comparing different insulin
treatment regimens are sparse and include
only a small number of patients treated for a
maximum of 1 year (Tables 1 and 2). Data
on effects of insulin-combination therapy
versus insulin alone on diabetic microvascular and macrovascular complications are
nonexistent. The main reason for the paucity of data may be the reluctance of private
funding agencies to support studies using
pharmacological agents and the reluctance
of industry to support studies with established preparations. The development
of new agents such as glitazones and
insulin analogs have increased the number of patients included in various trials,
but many company-initiated trials are designed to fulfill licensing requirements
and must be performed in multiple centers to save time. Although some company-initiated trials are of superb quality,
others suffer from inadequate glycemic
control and may lack the comparisons the
DIABETES CARE, VOLUME 24, NUMBER 4, APRIL 2001
Yki-Järvinen
Table 5—Studies comparing insulin combination regimens with different insulin injection regimens in type 2 diabetic patients (oral agents
similar in all regimens)
Reference
no.
Regimen 1
n
Regimen 2
n
Glycemia
Weight gain
Hypoglycemia
208
28
No difference
No difference
No difference
Less† with bedtime NPH
Less with glargine†
No difference
15
Bedtime glargine ⫹ OHA*
Morning NPH ⫹ MET
⫹ SU
Morning NPH ⫹ SU
214
32
34
Bedtime NPH ⫹ OHA*
Bedtime NPH ⫹ MET
⫹ SU
Bedtime NPH ⫹ SU
14
No difference
No difference
33
31
32
Bedtime NPH ⫹ SU
Bedtime NPH ⫹ SU
Bedtime NPH ⫹ SU
24
39
135
Morning NPH ⫹ SU
3 ⫻ regular ⫹ SU
3 ⫻ lispro ⫹ SU
24
41
139
No difference
No difference
No difference
No difference
Less with bedtime NPH
Less with bedtime NPH
Less with bedtime
NPH
No difference
No difference
No difference
40
15
*OHA, oral hypoglycemic agents, 59% SU ⫹ MET, no differences in OHA between groups using NPH versus insulin glargine; †statistically significant difference
between regimen 1 and regimen 2.
clinicians would be interested in. Despite
these deficiencies, some conclusions regarding the role of insulin combination
therapy in the treatment of type 2 diabetic
patients seem justified.
No study reported worse glycemic
control with insulin combination therapy
than with insulin alone. Glycemic control
was better with insulin combination therapy than with insulin alone in most studies of previously insulin-treated patients,
but this could be explained by a smaller
difference (⬃20% for metformin or sulfonylureas) (Table 2) in the insulin dose between the two modes of treatment than in
studies performed in insulin-naı̈ve patients (30 – 40%, Table 1). Combination
regimens allow use of less insulin injections, which may ease titration of the insulin dose and compliance (12,15,41).
These benefits must be balanced against
the side effects of oral drugs and, in some
countries, their cost. Abnormal renal or
liver function also limits the use of many
oral agents. Weight gain seems proportional to the number of insulin injections
used (12,15,31,32) and can be counteracted by inclusion of metformin in the
treatment regimen. Metformin also seems
to reduce the incidence of hypoglycemias
(12), as does the use of the peakless longacting insulin analog insulin glargine
compared with NPH (40). These considerations and the need to treat not only
hyperglycemia but also other risk factors
in type 2 diabetes support the use of simple insulin combination regimens such as
insulin glargine and metformin and or a
sulfonylurea (40). The prevailing view
that patients who are poorly responsive to
such a regimen benefit from adding additional insulin injections is not supported
by existing data. Instead, special emphasis should be placed on increasing the
dose of the single long-acting insulin to a
Table 6—Simple algorithm to start insulin combination therapy in an insulin-naive patient treated with oral combination therapy
Objectives
Visit–1, before start of insulin therapy
䡠 Teach home-glucose monitoring
䡠 Correct gross errors in diet
Visit 0, initiation of insulin therapy
䡠 Stop sulfonylurea, continue metformin 2 g/day†
䡠 Teach insulin injection technique
䡠 Define initial dose of insulin (glargine, NPH or 30/70 at
6:00 P.M. or later)
䡠 Give written instructions regarding self-adjustment of the
insulin dose
䡠 Teach symptoms of hypoglycemias
䡠 Schedule a phone call after 1 week and visit after 2–4 weeks
Subsequent visits
䡠 Individualize frequency—consider electronic transfer of home
glucose–monitoring
䡠 Results and phone calls instead of outpatient visits
Details
Home glucose monitoring
䡠 Measure fasting glucose daily during first weeks or months; after
reaching target frequency can be even once a week
Initial dose of insulin (insulin glargine, NPH, or ultralente)
䡠 Irrelevant if adjusted by patient
䡠 Safe starting dose ⫽ fasting glucose (mmol/l). i.e., 10 IU if fasting
glucose is 10 mmol/l
Self-adjustment of insulin doses
䡠 If fasting glucose exceeds 5.5 mmol/l (100 mg/dl) on three consecutive measurements, increase bedtime insulin dose by 2 IU
䡠 During combination therapy with NPH and oral agents
(ref. FINFAT), or fasting glucose of ⱕ6 mmol/l corresponds to
ⱕ7.5% HbA1c
*There are no data on use of glitazones in combination therapy with insulin in insulin-naive patients; †based on the FINFAT study, in which glyburide and NPH
insulin were used and use of this combination prevented adequate titration of the insulin dose (12); a higher incidence of symptoms of mild hypoglycemia was found
using glimepiride combined with 30/70 insulin given at 6:00 P.M. Similar problems were not reported in another study in which glimepiride was combined with
30 –70 insulin at 6:00 P.M. (25) and may not be a problem with insulin glargine (40). Note that stopping a sulfonylurea necessitates a rapid increase in the insulin
dose, which can be performed by teaching the patient self-adjustment of the insulin dose.
DIABETES CARE, VOLUME 24, NUMBER 4, APRIL 2001
765
Combination therapies with insulin in type 2 diabetes
dose that normalizes the fasting glucose
concentration.
References
1. U.K. Prospective Diabetes Study (UKPDS)
Group: Intensive blood-glucose control
with sulphonylureas or insulin compared
with conventional treatment and risk of
complications in patients with type 2 diabetes (UKPDS 33). Lancet 352:837– 852,
1998
2. Stratton IM, Adler AI, Neil HA, Matthews
DR, Manley SE, Cull CA, Hadden D,
Turner RC, Holman RR: Association of
glycaemia with macrovascular and microvascular complications of type 2 diabetes
(UKPDS 35): prospective observational
study. BMJ 321:405– 412, 2000
3. Ryysy L, Hakkinen AM, Goto T, Vehkavaara S, Westerbacka J, Halavaara J,
Yki-Jarvinen H: Hepatic fat content and
insulin action on free fatty acids and glucose metabolism rather than insulin absorption are associated with insulin requirements during insulin therapy in type
2 diabetic patients. Diabetes 49:749 –758,
2000
4. Peters AL, Davidson MB: Insulin plus a
sulfonylurea agent for treating type 2 diabetes. Ann Intern Med 115:45–53, 1991
5. Pugh JA, Wagner ML, Sawyer J, Ramirez
G, Tuley M, Friedberg SJ: Is combination
sulfonylurea and insulin therapy useful in
NIDDM patients? Diabetes Care 15:953–
959, 1992
6. Yu JG, Kruszynska YT, Mulford MI, Olefsky JM: A comparison of troglitazone and
metformin on insulin requirements in euglycemic intensively insulin-treated type
2 diabetic patients. Diabetes 48:2414 –
2421, 1999
7. Knowler WC, Pettitt DJ, Savage RJ, Bennett PH: Diabetes incidence in Pima Indians: contributions of obesity and parental
diabetes. Am J Epidemiol 113:144 –156,
1981
8. Bogardus C, Taskinen M-R, Zawadzki J,
Lillioja S, Mott D, Howard BV: Increased
resting metabolic rates in obese subjects
with non-insulin-dependent diabetes mellitus and the effect of sulfonylurea therapy.
Diabetes 35:1–5, 1986
9. Welle S, Nair KS, Lockwood D: Effect of a
sulfonylurea and insulin on energy expenditure in type II diabetes mellitus.
J Clin Endocrinol Metab 66:593–597, 1988
10. Franssila-Kallunki A, Groop L: Factors
associated with basal metabolic rate in patients with type 2 (non-insulin-dependent) diabetes mellitus. Diabetologia 35:
962–966, 1992
11. Mäkimattila S, Nikkilä K, Yki-Järvinen H:
Causes of weight gain during insulin therapy with and without metformin in pa-
766
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
tients with non-insulin-dependent diabetes mellitus. Diabetologia 42:406–412, 1999
Yki-Jarvinen H, Ryysy L, Nikkila K, Tulokas T, Vanamo R, Heikkila M: Comparison of bedtime insulin regimens in
patients with type 2 diabetes mellitus: a
randomized, controlled trial. Ann Intern
Med 130:389 –396, 1999
O’Brien SV, Mangnall M: Metformin superior to gliclazide when used in combination with NPH insulin in type 2 diabetic
patients inadequately controlled on oral
hypoglycemic therapy: median follow-up
13 months (Abstract). Diabetes 48:A115,
1999
Schernthaner G, Biesenbach G, Kacerovsky G, Mihaljevic R, Pieber TR: Evaluation of combination therapy in NIDDM
patients with sulfonylurea failure: the
Austrian intervention study (Abstract).
Diabetes 46 (Suppl. 1):166A, 1997
Yki-Järvinen H, Kauppila M, Kujansuu E,
Lahti J, Marjanen T, Niskanen L, Rajala
S, Ryysy L, Salo S, Seppälä P, Tulokas
T, Viikari J, Karjalainen J, Taskinen M-R:
Comparison of insulin regimens in patients with non-insulin-dependent diabetes mellitus. N Engl J Med 327:1426 –
1433, 1992
Chow C-C, Tsang LWW, Sorensen JP,
Cockram CS: Comparison of insulin with
or without continuation of oral hypoglycemic agents in the treatment of secondary failure in NIDDM patients. Diabetes
Care 18:307–314, 1995
Akazawa S, Sun F, Ito M, Kawasaki E,
Eguchi K: Efficacy of troglitazone on body
fat distribution in type 2 diabetes. Diabetes Care 23:1067–1071, 2000
Kawai T, Takei I, Oguma Y, Ohashi N,
Tokui M, Oguchi S, Katsukawa F, Hirose
H, Shimada A, Watanabe K, Saruta T: Effects of troglitazone on fat distribution in
the treatment of male type 2 diabetes. Metabolism 48:1102–1107, 1999
Mori Y, Murakawa Y, Okada K, Horikoshi
H, Yokoyama J, Tajima N, Ikeda Y: Effect
of troglitazone on body fat distribution in
type 2 diabetic patients. Diabetes Care 22:
908 –912, 1999
Balfour JA, Plosker GL: Rosiglitazone.
Drugs 57:921–930, 1999
Gillies PS, Dunn CJ: Pioglitazone. Drugs
60:333–343, 2000
The Diabetes Control and Complications
Trial Research Group: The effect of intensive treatment of diabetes on the development and progression of long-term
complications in insulin-dependent diabetes mellitus. N Engl J Med 329:977–986,
1993
Ohkubo Y, Kishikawa H, Araki E, Miyta
T, Isami S, Motoyoshi S, Kojima Y, Furuyoshi N, Shichiri M: Intensive insulin
therapy prevents the progression of diabetic microvascular complications in
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
Japanese patients with non-insulin-dependent mellitus: a randomized prospective 6-year study. Diabetes Res Clin Pract
28:103–117, 1995
Aviles-Santa L, Sinding J, Raskin P: Effects
of metformin in patients with poorly controlled, insulin-treated type 2 diabetes
mellitus: a randomized, double-blind,
placebo-controlled trial. Ann Intern Med
131:182–188, 1999
Riddle MC, Schneider J: Beginning insulin treatment of obese patients with
evening 70/30 insulin plus glimepiride
versus insulin alone: Glimepiride Combination Group. Diabetes Care 21:1052–
1057, 1998
Stenman S, Groop P-H, Saloranta C, Tötterman KJ, Fyhrqvist F, Groop L: Effects
of the combination of insulin and glibenclamide in type 2 (non-insulin-dependent) diabetic patients with secondary
failure to oral hypoglycaemic agents. Diabetologia 31:206 –213, 1988
Riddle MC, Hart JS, Bouma DJ, Phillipson
BE, Youker G: Efficacy of bedtime NPH
insulin with daytime sulfonylurea for subpopulation of type II diabetic subjects. Diabetes Care 12:623– 629, 1989
Riddle M, Hart J, Bingham P, Garrison C,
McDaniel P: Combined therapy for obese
type 2 diabetes: suppertime mixed insulin
with daytime sulfonylurea. Am J Med Sci
303:151–156, 1992
Wolffenbuttel BH, Sels JP, Rondas-Colbers GJ, Menheere PP, Nieuwenhuijzen
KA: Comparison of different insulin regimens in elderly patients with NIDDM. Diabetes Care 19:1326 –1332, 1996
Yki-Järvinen H, Ryysy L, Kauppila M, Kujansuu M, Lahti J, Marjanen T, Niskanen
L, Rajala S, Salo S, Seppälä P, Tulokas
T, Viikari J, Taskinen M-R: Effect of
obesity on the response to insulin therapy
in noninsulin-dependent diabetes mellitus. J Clin Endocrinol Metab 82:4037–
4043, 1997
Landstedt-Hallin L, Adamson U, Arner P,
Bolinder J, Lins P-E: Comparison of bedtime NPH or preprandial regular insulin
combined with glibenclamide in secondary sulfonylurea failure. Diabetes Care 18:
1183–1186, 1995
Bastyr EJ, Johnson ME, Trautmann ME,
Anderson JHJ, Vignati L: Insulin lispro in
the treatment of patients with type 2 diabetes mellitus after oral agent failure. Clin
Ther 21:1703–1714, 1999
Groop LC, Widen E, Ekstrand A, Saloranta C, Franssila-Kallunki A, Eriksson
JGS-J: Morning or bedtime NPH insulin
combined with sulfonylurea in treatment
of NIDDM. Diabetes Care 15:831–834, 1992
Soneru IL, Agrawal L, Murphy JC, Lawrence AM, Abraira C: Comparison of morning or bedtime insulin with and without
glyburide in secondary sulfonylurea fail-
DIABETES CARE, VOLUME 24, NUMBER 4, APRIL 2001
Yki-Järvinen
ure. Diabetes Care 16:896 –901, 1993
35. Turner R, Stratton I, Horton V, Manley S,
Zimmet P, Mackay IR, Shattock M, Bottazzo GF, Holman R: UKPDS 25: autoantibodies to islet-cell cytoplasm and glutamic acid decarboxylase for prediction of
insulin requirement in type 2 diabetes:
UK Prospective Diabetes Study Group.
Lancet 350:1288 –1293, 1997
36. World Health Organization: Report of a
WHO Consultation: Definition, Diagnosis
and Classification of Diabetes Mellitus and
its Complications. Part 1: Diagnosis and
Classification of Diabetes Mellitus. Geneva,
World Health Org., 1999
37. Landstedt-Hallin L, Arner P, Lins PE,
Bolinder J, Olsen H, Groop L: The role of
sulphonylurea in combination therapy
assessed in a trial of sulphonylurea withdrawal: Scandinavian Insulin-Sulphonylurea Study Group Research Team. Diabet Med 16:827– 834, 1999
38. Wolffenbuttel BH, Sels JP, Rondas-Colbers GJ, Menheere PP: Prognostic factors
for successful insulin therapy in subjects
with type 2 diabetes. Neth J Med 54:63–
69, 1999
39. Muller UA, Muller R, Starrach A, HungerDathe W, Schiel R, Jorgens V, Grusser M:
Should insulin therapy in type 2 diabetic
patients be started on an out- or inpatient
basis? Results of a prospective controlled
trial using the same treatment and teaching programme in ambulatory care and a
university hospital. Diabete Metab 24:251–
255, 1998
40. Yki-Jarvinen H, Dressler A, Ziemen M:
Less nocturnal hypoglycemia and better
post-dinner glucose control with bedtime
insulin glargine compared with bedtime
NPH insulin during insulin combination therapy in type 2 diabetes: HOE 901/
3002 Study Group. Diabetes Care 23:
1130 –1136, 2000
41. Bergenstal RJM, Whipple D, Noller D,
Boyce K, Roth L, Upham P, Fish L, Debold
R: Advantages of adding metformin to
multiple dose insulin therapy in type 2
diabetes (Abstract). Diabetes 47:A89, 1998
42. Relimpio F, Pumar A, Losada F, Mangas
MA, Acosta D, Astorga R: Adding metformin versus insulin dose increase in insulin-treated but poorly controlled type 2
diabetes mellitus: an open-label randomized trial. Diabet Med 15:997–1002, 1998
43. Robinson AC, Burke J, Robinson S, Johnston DG, Elkeles RS: The effects of metformin on glycemic control and serum
lipids in insulin-treated NIDDM patients
with suboptimal metabolic control. Diabetes Care 21:701–705, 1998
44. Kyllastinen M, Groop L: Combination of
insulin and glibenclamide in the treatment of elderly non-insulin dependent
DIABETES CARE, VOLUME 24, NUMBER 4, APRIL 2001
45.
46.
47.
48.
49.
50.
51.
52.
53.
54.
55.
(type 2) diabetic patients. Ann Clin Res 17:
100 –104, 1985
Lindstrom T, Nystrom FH, Olsson AG,
Ottosson AM, Arnqvist HJ: The lipoprotein profile differs during insulin treatment alone and combination therapy with
insulin and sulphonylureas in patients
with type 2 diabetes mellitus. Diabet Med
16:820 – 826, 1999
Clauson PG, Linde B: Absorption of rapid-acting insulin in obese and nonobese
NIDDM patients. Diabetes Care 18:986 –
991, 1995
Kitabchi AE, Soria AG, Radparvar A, Lawson-Grant V: Combined therapy of insulin and tolazamide decreases insulin requirement and serum triglycerides in obese
patients with noninsulin-dependent diabetes mellitus. Am J Med Sci 294:10 –14,
1987
Gutniak M, Karlander S-G, Efendic S:
Glyburide decreases insulin requirement,
increases b-cell response to mixed meal,
and does not affect insulin sensitivity: effects of short- and long-term combined
treatment in secondary failure to sulfonylurea. Diabetes Care 10:545–554, 1987
Schade DS, Mitchell WJ, Griego G: Addition of sulfonylurea to insulin treatment
in poorly controlled type II diabetes: a
double-blind, randomized clinical trial.
JAMA 257:2441–2445, 1987
Groop L, Harno K, Tolppanen EM: The
combination of insulin and sulphonylurea in the treatment of secondary drug
failure in patients with type II diabetes.
Acta Endocrinol (Copenh) 106:97–101, 1984
Groop L, Harno K, Nikkila EA, Pelkonen
R, Tolppanen EM: Transient effect of the
combination of insulin and sulfonylurea
(glibenclamide) on glycemic control in
non-insulin dependent diabetics poorly
controlled with insulin alone. Acta Med
Scand 217:33–39, 1985
Schwartz S, Raskin P, Fonseca V, Graveline JF: Effect of troglitazone in insulintreated patients with type II diabetes mellitus: Troglitazone and Exogenous Insulin
Study Group. N Engl J Med 338:861– 866,
1998
Rubin C, Egan J, Schneider R, Pioglitazone 014 Study Group: Combination
therapy with pioglitazone and insulin in
patients with type 2 diabetes (Abstract).
Diabetes 48:A110, 1999
Raskin P, Dole JF, Rappaport EB: Rosiglitazone (RSG) improves glycemic control
in poorly controlled, insulin-treated type
2 diabetes (Abstract). Diabetes 48:A94, 1999
Bachmann W, Lotz N, Mehnert H, Rosak
C, Schoffling K: Effectiveness of combined treatment with glibenclamide and
insulin in secondary sulfonylurea failure:
a controlled multicenter double-blind
56.
57.
58.
59.
60.
61.
62.
63.
64.
65.
66.
clinical trial. Dtsch Med Wochenschr 113:
631– 636, 1988
Quatraro A, Consoli G, Ceriello A, Giugliano D: Combined insulin and sulfonylurea therapy in non-insulin-dependent diabetics with secondary failure to oral drugs: a
one year follow-up. Diabete Metab 12:315–
318, 1986
Giugliano D, Quatraro A, Consoli G, Minei A, Ceriello A, DeRosa N, D’Onofrio F:
Metformin for obese, insulin-treated diabetic patients: improvement in glycaemic
control and reduction of metabolic risk
factors. Eur J Clin Pharmacol 44:107–112,
1993
Lins PE, Lundblad S, Persson-Trotzig E,
Adamson U: Glibenclamide improves the
response to insulin treatment in non-insulin-dependent diabetics with second
failure to sulfonylurea therapy. Acta Med
Scand 223:171–179, 1988
Reich A, Abraira C, Lawrence AM: Combined glyburide and insulin therapy in type
II diabetes. Diabetes Res 6:99 –104, 1987
Lewitt MS, Yu VK, Rennie GC, Carter JN,
Marel GM, Yue DK, Hooper MJ: Effects of
combined insulin-sulfonylurea therapy in
type II patients. Diabetes Care 12:379 –
383, 1989
Osei K, O’Dorisio TM, Falko JM: Concomitant insulin and sulfonylurea therapy
in patients with type II diabetes: effects
on glucoregulation and lipid metabolism.
Am J Med 77:1002–1009, 1984
Feinglos MN, Thacker CR, Lobaugh B,
DeAtkine DD, McNeill DB, English JS,
Bursey DL: Combination insulin and sulfonylurea therapy in insulin-requiring
type 2 diabetes mellitus. Diabetes Res Clin
Pract 39:193–199, 1998
Longnecker MP, Elsenhans VD, Leiman
SM, Owen OE, Boden G: Insulin and a
sulfonylurea agent in non-insulin-dependent diabetes mellitus. Arch Intern Med
146:673– 676, 1986
Casner PR: Insulin-glyburide combination therapy for non-insulin-dependent
diabetes mellitus: a long-term doubleblind, placebo-controlled trial. Clin Pharmacol Ther 44:594 – 603, 1988
Kelley DE, Bidot P, Freedman Z, Haag B,
Podlecki D, Rendell M, Schimel D, Weiss
S, Taylor T, Krol A, Magner J: Efficacy and
safety of acarbose in insulin-treated patients with type 2 diabetes. Diabetes Care
21:2056 –2061, 1998
Chiasson J-L, Josse RG, Hunt JA, Palmason C, Rodger NW, Ross SA, Ryan EA,
Tan MH, Wolever TMS: The efficacy of
acarbose in the treatment of patients with
non-insulin-dependent diabetes mellitus:
a multicenter controlled clinical trial. Ann
Intern Med 121:928 –935, 1994
767