Coffee Intake and Mortality from Liver Cirrhosis
AAGE TVERDAL, PHD, AND SVETLANA SKURTVEIT, PHD
PURPOSE: The aim of the study was to evaluate the association between coffee consumption and mortality from liver cirrhosis.
METHODS: We conducted a mortality follow-up of 51,306 adults who underwent screening for cardiovascular disease from 1977 to 1983. During the subsequent 17 years, the total number of deaths from all
causes in the studied cohort was 4207. Fifty-three had the diagnosis of cirrhosis mentioned on the death
certificate; of these, 36 had alcoholic cirrhosis.
RESULTS: The relative risk of liver cirrhosis mentioned on the death certificate associated with an increase of two cups of coffee, adjusted for sex, age, alcohol use and other major cardiovascular risk factors
was 0.6 (95% confidence interval, 0.5–0.8). For alcoholic cirrhosis the results were identical. When studying cirrhosis as the underlying cause of death, the inverse relationship becomes somewhat stronger.
CONCLUSIONS: The present study confirms the existence of an inverse association between coffee
consumption and liver cirrhosis.
Ann Epidemiol 2003;13:419–423. © 2003 Elsevier Inc. All rights reserved.
KEY WORDS:
Liver cirrhosis, Alcohol, Coffee consumption.
INTRODUCTION
Alcoholic beverages and coffee have been part of the human diet for centuries. Norway is well known for a restrictive alcohol policy and the total alcohol consumption is
relatively low and liver cirrhosis mortality is not predominant. On the other hand Norwegians are among the highest consumers of coffee in the world (1).
Alcohol has a well-established toxic and graded effect on
liver, including cirrhosis. Since most heavy drinkers do not develop alcoholic cirrhosis, it is widely believed that alcoholic
cirrhosis has multiple causes and that both start and progression
of the process may be dependent upon predisposing factors and
other causes. It has the well known effect of raising blood concentration of the liver enzyme glutamyltransferase (GGT)
which is widely used as a marker of alcohol intake, although it
is documented that other factors also are associated with serum
levels of GGT (2–4).
An inverse association between serum GGT and intake of
coffee has been documented in the study of Arnesen et al. (2) in
population studies in Norway. This finding has been consistently observed in population studies from other countries
(3–6). Attention has recently been drawn to possible beneficial
effects of coffee drinking on liver diseases. A prospective study
of Klatsky et al. (7–8) reported a lower risk of hospitalization,
From the Norwegian Institute of Public Health, Oslo, Norway (A.T., S.S.).
Address correspondences to: Svetlana Skurtveit, Ph.D., Norwegian Institute of Public Health, P.O. Box 4404, Nydalen, 0403 Oslo, Norway.
Tel.: (47) 234-081-72; Fax: (47) 234-082-60. E-mail:svetlana.skurtveit
@fhi.no
Received May 28, 2002; accepted October 4, 2002.
© 2003 Elsevier Inc. All rights reserved.
360 Park Avenue South, New York, NY 10010
and death from alcoholic cirrhosis in association with coffee
drinking. Similar results were also observed in three casecontrol studies published later (9–11).
To provide further information on the issue, we utilized
a cardiovascular survey in a population characterized by
high coffee consumption and low risk of liver cirrhosis and
did mortality follow-up with liver cirrhosis mentioned on
the death certificate as endpoint.
MATERIALS AND METHODS
Between 1977 and 1983, men and women living in three
counties in different parts of Norway (Finmark 1977–78,
Sogn og Fjordane 1980, Oppland 1981–83) were invited to
participate in a cardiovascular screening program organized
by the National Health Screening Service. The screening
procedures have been described in detail elsewhere (1, 12).
The survey included all residents aged 40–55, and an approximately 20% random sample of men and women aged
20–39 years. Response rate was high (87% for men and
93% for women) and it was much the same in the three
counties. On the reverse side of the invitation letter was a
questionnaire including questions on smoking habits, physical activity, and history of cardiovascular diseases.
The information about coffee and alcohol consumption
was obtained from a self-administered food-frequency questionnaire (13). This was handed out to the person who attended the screening, and was to be returned by mail. All
participants, except in 5 municipalities in Sogn Fjordane,
got this additional questionnaire and about 94% returned
it. A small percentage, 4%, of those who responded, did not
answer the question on coffee habits.
1047-2797/03/$–see front matter
doi:10.1016/S1047-2797(02)00462-3
420
Tverdal and Skurtveit
COFFEE AND MORTALITY FROM CIRRHOSIS
Our study population included persons, aged 20–55 years
(median 43.6 years), who answered all questions on life
style habits used in the present study—a total of 25,763
men and 25,543 women. The examination comprised measurements of weight, height, systolic and diastolic blood
pressure, and collection of venous non-fasting blood samples. Serum cholesterol and triglycerides were determined
at the same laboratory—Central Laboratory, Ullevål Hospital, Oslo (1, 12).
The question about coffee read, “How many cups of coffee do you usually drink per day?” with preset categories:
“Do not drink coffee or less than one cup”, “1–2 cups”, “3–4
cups”, “5–6 cups”, “7–8 cups”, and “9 or more cups”. These
categories were coded as 1, 2, 3, 4, 5 and 6 when estimating
linear trend and relative risk. No question about the
method of coffee preparation was asked. Alcohol consumption was recorded using two questions: “Do you usually
drink wine/spirits during one week?” and “Do you usually
drink beer during one week?” Drinkers of alcohol were defined as people who answered yes on one or both alcohol
questions. Information on smoking habits was collected by
the questions: “Do you smoke daily at present?” and “How
many cigarettes do you or did you usually smoke daily?”
The number of observation years was calculated for each
person from the time of examination at screening to the
time of death, time of emigration, or at the end of the follow-up (end of 1997). Mean follow-up was 16.9 years. The
total number of deaths from all causes in the studied cohort
was 4207, and 173 persons emigrated from Norway during
follow-up period.
The follow-up period covered three versions of the International Classification of Diseases (ICD). Liver cirrhosis
was defined as ICD-8, 9: 571 and ICD10: K70 or K73 or
K74. Alcoholic liver cirrhosis was defined as ICD-8, 9:
571.0–571.3 and ICD 10: K70. Up to four causes of death
were recorded on the death certificate in ICD-8 and ICD-9
and up to seven causes of deaths in ICD-10. The first cause
is the underlying; the others are contributory causes of
death. Altogether, 33 deaths were recorded with cirrhosis
as the underlying cause and 20 deaths with cirrhosis as the
contributing cause; i.e. cirrhosis was mentioned on the
death certificate for 53 deaths, 38 of the deaths in men and
15 in women. The corresponding figures for alcoholic cirrhosis were 25 deaths as the underlying cause and 11 deaths
as the contributing cause; i.e. alcoholic cirrhosis was mentioned on the death certificate for 36 deaths, 32 of the
deaths in men and 4 in women. The mean age at death was
46.3 years for cirrhosis and 45.1 years for alcoholic cirrhosis.
Relative risks were estimated by using the Cox proportional hazards regression model in SPSS (14). Interaction
was tested by comparing –2 *(log likelihood) in models
with and without the interaction term. Direct standardization was used to obtain age-adjusted rates/100000 personyears with the study population in five-year age-groups as
AEP Vol. 13, No. 6
July 2003: 419–423
the standard. Linear trend across various variables of the
coffee categories was estimated by linear regression, with
coffee as the dependent variable.
RESULTS
Table 1 shows biologic variables and other lifestyle factors according to different levels of coffee intake. Cholesterol levels and prevalence of the current daily smoking and alcohol
use increased steadily with increasing level of coffee intake.
The mortality rates were distinctly lower among persons
drinking three or more cups of coffee than in those drinking
two or less (Table 2). This applies to both cirrhosis and alcoholic cirrhosis as mentioned and underlying causes. For
above three cups of coffee there was no clear relationship
with liver cirrhosis mortality.
When coffee intake was entered as a continuous variable in
the Cox regression analysis, a distinct inverse relationship with
death from liver-, total- and alcoholic cirrhosis appeared. (Table 3). The relationship persisted or became even stronger after
adjustment for smoking and alcohol use. Additional adjustment for triglycerides, systolic blood pressure, BMI, and cholesterol did not change the relationship. Smoking and alcohol
use were positively related to mortality from cirrhosis and alcoholic cirrhosis. When the analyses were done with cirrhosis as
the underlying cause of death as endpoint, similar relative risks
emerged. When number of cigarettes was used as covariate,
similar relative risk of coffee consumption on death from cirrhosis was observed.
Testing for interaction between coffee and smoking, and
smoking and alcohol in relation to death of liver cirrhosis
produced no significance. However, our study had less than
20% power to reveal the interactions of the magnitudes
that we actually found, with 5% significance level. This is
based on the sample size consideration given by Schmoor et
al. (15) for binary prognostic factors in survival analysis. In
this case, we dichotomized coffee intake into 4 and 4
cups/day. Hence, the existence of interactions cannot be
ruled out by this study. It is mentioned that our study had
close to 80% power to detect the main effect of coffee.
DISCUSSION
This study provides evidence of a favorable role of coffee
intake on the risk of death from liver cirrhosis. To our
knowledge, only the studies of Klatsky et al. (7–8), Corrao
et al. (9–10) and Gallus et al. (11) have related the risk of
cirrhosis to coffee intake. In the prospective study conducted
in Northern California (7), an inverse coffee-cirrhosis relation
was reported for the first time. Coffee drinking, but not tea
drinking, was inversely related to alcoholic cirrhosis risk,
with persons who drank four or more cups per day at onefifth the risk of those who did not drink coffee. These find-
AEP Vol. 13, No. 6
July 2003: 419–423
Tverdal and Skurtveit
COFFEE AND MORTALITY FROM CIRRHOSIS
421
TABLE 1. Baseline characteristics according to coffee consumption in Norwegian men and women 20–55 years
Cups of coffee per day
Baseline characteristics
No. at risk
Median age
Sex (men %)
Alcohol use (%)
Current daily smokers (%)
BMI (kg/m2)
Total cholesterol (mmol/l)
Triglycerides (mmol/l)
Systolic blood pressure (mmHg)
1
1–2
3–4
5–6
7–8
9
p (trend)
2327
40.2
57.1
26.6
17.8
24.4
5.59
1.74
132
4802
43.2
48.0
28.3
23.1
24.7
5.91
1.80
134
15,138
44.6
43.0
26.4
29.0
25.0
6.10
1.75
135
15,598
44.1
48.6
28.4
43.8
25.0
6.19
1.76
134
8059
43.1
56.6
32.3
58.9
25.0
6.27
1.79
134
5382
41.7
64.7
37.0
71.0
25.0
6.37
1.79
133
0.001
0.001
0.001
0.001
0.001
0.001
ns
0.05
ings were later supported by case-control studies with morbidity of liver cirrhosis as endpoint (9–10). In the study of
Gallus et al. (11), an inverse relation was observed across
strata of tobacco and alcohol consumption, also after adjustment to age, sex and other major covariates like body
mass index, fruit or vegetable consumption, and education.
Results from our study are in agreement with these studies.
One strength of our study is the prospective design minimizing the chance of exposure recall bias.
It is unlikely that our findings are due to selection of
people with a favorable life style. In previous studies in almost the same study population, correlation coefficients between coffee consumption and a battery of lifestyle and
biological variables were estimated (13). Coffee drinking
was largely related to factors that are considered to constitute an unfavorable lifestyle. Furthermore, previously discussed studies (7–11) on coffee and liver cirrhosis and our
study are obtained from different patients or populations
who have very different cultural and dietary patterns.
One weakness of the study is that we only have mortality as
the endpoint. People might have liver cirrhosis without dying
from it or without having it mentioned on the death certificate. However, it is less likely that a diagnosis of liver cirrhosis
is passed over in alcohol users. As there is a positive correlation between coffee intake and alcohol intake, it is reasonable
to believe that diagnosis of liver cirrhosis also is less likely to be
missed in coffee users. The corollary is that if a differential misclassification is present, it is more likely that the consequence
has been a dilution of the relationship between coffee intake
and mortality from liver cirrhosis.
The questions on alcohol intake did not concern alcohol
dose. However, it is hard to believe that those who usually did
not consume alcohol during one week should have the highest
intake. Another potential confounder is the smoking. Coffee
consumers smoke more cigarettes (13). The analysis with number of cigarettes as covariate produced similar results. It is also
noted that adjustment for smoking did not alter or strengthen
the inverse relationship between coffee intake and mortality
from liver cirrhosis. Hence, it is not likely that the observed relationship is due to residual confounding of smoking.
Coffee drinking was associated with higher levels of cholesterol. To see if the favorable effect of coffee drinking on
death from liver cirrhosis was mediated through cholesterol
level, we related serum cholesterol to liver cirrhosis mentioned on the death certificate in a Cox analysis. Serum
cholesterol was not related to liver cirrhosis.
What agent in coffee may explain the protective effect of
coffee consumption on liver cirrhosis is not clear today. The
possibility that substances contained in coffee may exert a protective effect on the hepatocyte has been underlined by the
study of Casiglia et al. (16), which found consistently lower
liver enzymes in coffee drinkers than in non-coffee drinkers;
the higher the daily number of cups of coffee, the lower the serum levels of all liver enzymes and bilirubin. Coffee intake was
significantly related to decreased serum concentrations of liver
enzymes also in other previous studies (5, 17).
The study of Corrao et al. (10) concluded that the inverse relationship of coffee and cirrhosis is probably not attributable to
caffeine because no negative liver cirrhosis relationship was
found to the other beverages containing caffeine. Studies of
Urgert et al (18) and Weusten-Van der Wouw et al. (19)
showed in human experimental studies that coffee oils, brews,
and grounds containing cafestol and kahweol increased liver
function enzymes such as alanin aminotransferase, asparate aminotransferase but reduced serum level of GGT and creatinin.
The acute alteration in liver enzyme levels with intake
of cafestol and kahweol as individual agents differ from the
results of epidemiological studies that examined the association between coffee intake and liver enzymes other than
GGT. Such findings could suggest that cafestol and kahweol
are not the only agents in coffee responsible for the epidemiological findings in our study. Other ingredients contained in coffee with various biological actions may well
play a role, as there are three to four-hundred pharmaceutically active substances in coffee (20).
There seems to be a threshold in the relationship; no further decline in mortality when the intake is more than three
to four cups of coffee. The size of the cups varies a lot in Norway, with an average of 1.25 dl (1). If part of the etiology, a
sufficient dose seems to be in order of at least 3.7 dl per day.
5
12
11
15
6
4
Number of
deaths
Mentioned
15.50
15.64
3.89
5.59
4.60
4.54
0.01
0.01
Mortality,
age-adjusted
12.90
15.05
4.35
5.68
4.33
4.21
Mortality,
crude
4
9
7
10
2
1
Number of
deaths
0.001
10.31
11.28
2.77
3.78
1.44
1.05
Mortality,
crude
Underlying
0.001
12.29
11.60
2.57
3.72
1.66
1.09
Mortality,
age-adjusted
3
9
6
10
6
2
Number of
deaths
Mentioned
0.05
7.73
11.28
2.37
3.78
4.33
2.10
Mortality,
crude
0.05
9.04
12.07
2.19
3.76
4.60
2.18
Mortality,
age-adjusted
0.6 (0.5–0.8)
10.1 (3.8–26.6)
2.7 (1.3–5.5)
0.6 (0.5–0.8)
4.8 (2.6–8.9)
2.8 (1.6–5.0)
Coffee, alcohol
and smoking
0.6 (0.5–0.8)
10.1 (3.8–27.1)
2.9 (1.4–6.0)
0.6 (0.5–0.8)
4.9 (2.6–9.1)
3.0 (1.7–5.4)
Coffee, alcohol,
smoking
and othersa
0.6
14.4
2.6
0.6
7.6
2.7
0.5 (0.4–0.7)
12.9 (3.7–44.6)
3.1 (1.3–7.5)
0.5 (0.4–0.7)
6.7 (2.9–15.4)
3.4 (1.6–7.3)
Coffee, alcohol
and smoking
Model with sex, age and
Coffee or
alcohol or
smoking,
one at time
Model with sex, age and
0.01
5.15
10.03
1.98
2.65
1.44
1.05
Underlying
BMI, total cholesterol, systolic blood pressure and triglycerides (log transformed); all variables in the model.
0.7
11.0
2.4
Alcoholic cirrhosis
Coffee drinking (per unit)
Alcohol use (yes/no)
Smoking (yes/no)
a
0.7
5.3
2.4
Cirrhosis
Coffee drinking (per unit)
Alcohol use (yes/no)
Smoking (yes/no)
Coffee or
alcohol or
smoking,
one at time
0.01
5.88
10.42
1.85
2.63
1.66
1.09
Mortality,
age-adjusted
0.5 (0.4–0.7)
13.3 (3.8–46.6)
3.4 (1.4–8.2)
0.5 (0.4–0.7)
6.9 (2.9–16.1)
3.7 (1.7–7.9)
Coffee, alcohol,
smoking
and othersa
Underlying
Mortality,
crude
Mentioned
2
8
5
7
2
1
Number of
deaths
Alcoholic cirrhosis
TABLE 3. Relative risks (95% confidence intervals) estimated by Cox proportional hazards regression for men and women 20–55 years
p (trend)
1
1–2
3–4
5–6
7–8
9
Cups of
coffee/day
Cirrhosis
TABLE 2. Crude and age-adjusted mortality per 100,000 person-years from liver cirrhosis or alcoholic cirrhosis by coffee intake for men and women 20–55 years
422
Tverdal and Skurtveit
COFFEE AND MORTALITY FROM CIRRHOSIS
AEP Vol. 13, No. 6
July 2003: 419–423
AEP Vol. 13, No. 6
July 2003: 419–423
In summary, there was an inverse relationship between
coffee intake and mortality from liver cirrhosis in this middle-aged Norwegian population with a high coffee intake.
The mechanisms behind this relationship remain unclear.
Tverdal and Skurtveit
COFFEE AND MORTALITY FROM CIRRHOSIS
10.
11.
12.
REFERENCES
1. Stensvold I. Coffee and health. Is high consumption a risk factor for
cardiovascular disease or cancer? National Health Screening Service.
1996. Oslo, Norway (ISBN 82-7703-014-2).
2. Arnesen E, Huseby NE, Brenn T, Try K. The Tromsø heart study, distribution of, and determinants for, gamma-glutamyltransferase in a
free-living population. Scand J Clin Lab Invest. 1986;46:63–70.
3. Poikolainen K, Vartiainen E. Determinants of gamma-glutamyltransferase: positive interaction with alcohol and body mass index, negative association with coffee. Am J Epidemiol. 1997;146:1019–1024.
4. Nakanishi N, Nakamura K, Suzuki K, Tatara K. Lifestyle and serum
gamma-glutamyltransferase: a study of middle-aged Japanese men. Occup Med. 2000;50:115–120.
5. Honjo S, Kono S, Coleman MP, Shinchi K, Sakurai Y, Todoroki I, et
al. Coffee drinking and serum gamma-glutamyltransferase: an extended study of self-defense officials of Japan. Ann Epidemiol.
1999;9:325–331.
6. Pintus F, Mascia P. Distribution and population determinants of
gamma-glutamyltransferase in a random sample of Sardinian inhabitants. Eur J Epidemiol. 1996;12:71–76.
7. Klatsky AL, Armstrong MA. Alcohol, smoking, coffee and cirrhosis.
Am J Epidemiol. 1992;136:1248–1257.
8. Klatsky AL, Armstrong MA, Friedman GD. Coffee, tea, and mortality. Ann Epidemiol. 1993;3:375–381.
9. Corrao G, Lepore AR, Torchio P, Valenti M, Galatola G, D’Amicis
A, et al. The effect of drinking coffee and smoking cigarettes on the
13.
14.
15.
16.
17.
18.
19.
20.
423
risk of cirrhosis associated with alcohol consumption. Eur J Epidemiol.
1994;10:657–664.
Corrao G, Zambon A, Bagnardi V, Dàmicis A, Klatsky A, and collaborative SIDECIR group. Coffee, caffeine, and the risk of liver cirrhosis. Ann Epidemiol. 2001;11:458–465.
Gallus S, Tavani A, Negri E, La Vecchia C. Does coffee protect
against liver cirrhosis? Ann Epidemiol. 2002;12:202–205.
Stensvold I, Urdal P, Thurmer H, Tverdal A, Lund-Larsen PG, Foss
OP. High-density lipoprotein cholesterol and coronary, cardiovascular and all cause mortality among middle-aged Norwegian men and
women. Eur Heart J. 1992;13:1155–1163.
Solvoll K, Selmer R, Løken EB, Foss OP, Trygg K. Coffee, dietary
habits, and serum cholesterol among men and women 35–49 years of
age. Am J Epidemiol. 1989;129:1277–1288.
Statistical Package for Social Sciences (SPSS). 10.0 for Windows. 2001.
Schmoor C, Sauerbrei W, Schumacher M. Sample size considerations
for the evaluation of prognostic factors in survival analysis. Stat Med.
2000;19:441–452.
Casiglia E, Spolaore P, Ginocchio G, Ambrosio GB. Unexpected effects of coffee consumption on liver enzymes. Eur J Epidemiol. 1993;9:
293–297.
Tanaka K, Tokunaga S, Kono S, Tokudome S, Akamatsu T,
Moriyama T, et al. Coffee consumption and decreased serum gammaglutamyltransferase and aminotransferase activities among male alcohol drinkers. Int J Epidemiol. 1998;27:438–443.
Urgert R, Shulz AGM, Katan MB. Effects of cafestol and kahweol
from coffee grounds on serum lipids and serum liver enzymes in humans. Am J Clin Nutr. 1995;61:149–154.
Weusten-Van der Wouw MPME, Katan MB, Viani R, Huggett AC,
Liardon R, Lund-Larsen PG, et al. Identity of the cholesterol-raising
factor from boiled coffee and its effects on liver function. J Lipid Res.
1994;35:721–733.
Garattini S. Caffeine, Coffee and Health. Monographs of the Mario
Negri Institute for Pharmacological Research, Milan. New York:
Raven Press; 1993.