Methods to measure salt intake at
individual level: gold standard, proxy and
calibration needs
Francesco P Cappuccio MD MSc DSc FRCP FFPH FBHS FAHA
Professor of Cardiovascular Medicine & Epidemiology
Head, WHO Collaborating Centre
University of Warwick, Coventry, UK
Disclosures: Technical Advisor to the World Health Organization, the Pan American Health Organization,
Member of C.A.S.H., W.A.S.H., UK Health Forum and Trustee of the Student Heart Health Trust
Vice-President, British Hypertension Society – all unpaid.
How much sodium are we consuming?
Methods to estimate
sodium intake
300
Men
6 x 24-h urine collection
(n=159)
250
6,000
7-day diary
(n=3,423)
5,000
– Spot (casual)
– Overnight
– 24-hour
Norfolk-EPIC Study (Khaw et al. 2004)
Sodium (mg/d)
3,000
2,000
50
Food frequency
7-day weighed records
Food diary
24-hour recall
• Urinary collection
150
100
1,000
0
0
1
2
3
4
5
Quintiles of sodium intake
300
6 x 24-h urine collection
(n=181)
250
Women
7-day diary
(n=3,833)
6,000
5,000
200
Sodium (mmol/d)
–
–
–
–
4,000
4,000
150
3,000
100
2,000
50
1,000
0
0
1
2
3
Quintiles of sodium intake
4
5
Sodium (mg/d)
• Duplicate diets
• Dietary surveys
Sodium (mmol/d)
200
Background
• In steady state, most of Na eaten in a day is excreted through
urine in the following 24-h.
• 24-h urinary Na excretion ‘gold standard’ for measuring
sodium (salt) intake both in individuals and in populations.
• 24-h collections often deemed inconvenient, hence alternative
methods suggested (e.g. spot and timed urines).
• Their reliability and reproducibility often disappointing.
• However, in recent years a confusion has arisen as to what
extent ‘proxy’ measures can be used to assess ‘salt’ intake.
• ‘Population’ estimates (to monitor population policies) vs
‘Individual’ estimate (to use as exposure in research studies).
Background
•
•
•
•
Gold standard: 24h urine collection
High participation burden
Completeness
Need to explore alternatives
– Casual spot
– Timed (Day, Eve, Night)
– Estimation v Extrapolation
– Individual v Group
• Their reliability and reproducibility often disappointing
“Timed” urines
• Less participant burden than 24 hour collections
• More variable at individual level but may give good estimate of
group mean
• Desirable alternative for monitoring program effects over time
(validation necessary)
• Need baseline 24 hour urine assays to compare between
populations or time points
• Role of urinary creatinine
• Should learn from monitoring program for iodisation
• Could be used for ‘effective’ monitoring of iodine (ideally in
adult populations)
‘Spot’ urines
• Less participant burden than 24 hour collections
• Highly variable at individual level but may give estimate of group
mean
• Less desirable for monitoring program effects over time
• Highly dependent on hydration, volume, residual bladder
volume
• Currently used for monitoring iodine (mainly children and
women of childbearing age)
• Extrapolated to 24h (validation studies published –
interpretations vary - population specific ?)
Overnight collections
• Less participant burden than 24 hour
collections
• May give biased estimates of sodium excretion
– (greater % excretion overnight in hypertensive
individuals…)
• Undesirable for monitoring program effects
over time
Measurement of sodium intake
Sodium intake highly variable day to day
 Misclassification of individual intake and increase in
variance of group (mean unbiased)
Between individuals
Within individuals
s 2B
s 2W
Liu K et al. Hypertension 1979; 1: 529-36
s 2B
Reliability =
s 2B + s 2W
How many 24h collections are needed to characterize
an individual’s sodium excretion?
The variability of 24-hour urinary sodium (…) was studied in a
sample of 22 Neapolitan men with mild blood pressure elevation.
On 5 days within a 1-month period, 24-hour urine specimens were
collected by each subject. The estimated ratio of intraindividual-tointerindividual variance was 1.12 for urinary sodium, (…). Based on
these values, five 24-hour urine collections are necessary to reduce
to less than 10% the diminution of the correlation coefficient
between urinary sodium and another related variable; this number
is substantially lower than that found in previous studies in a North
American population sample, but similar to the one reported for
Chinese population samples (…).
Siani A et al. Hypertension 1989; 13: 38-42
•
•
•
•
•
•
•
•
We performed 2 independent ultralong-term salt balance studies lasting 105 (4
men) and 205 (6 men) days in 10 men
We controlled dietary intake of all constituents for months at salt intakes of 12, 9,
and 6 g/d and collected all urine.
Urinary recovery of dietary salt was 92% of recorded intake.
Even at fixed salt intake, 24-hour urine collection for sodium excretion (UNaV)
showed infradian rhythmicity.
Collecting seven 24-hour urines and sodium intake samples improved classification
accuracy to 92%.
Single 24-hour urine collections at intakes ranging from 6 to 12 g salt per day were
not suitable to detect a 3-g difference in individual salt intake.
Repeated measurements of 24-hour UNaV improve precision.
This knowledge could be relevant to patient care and the conduct of intervention
trials.
Lerchl K et al. Hypertension. 2015; 66: 850-7
Multiple collections reduce the variability and
improve the predictive value
Lerchl K et al. Hypertension 2015; 66: 850-7
Data
Study 1: Discovery sample
915 British men & women aged 40-59
(297 W, 326 B, 292 SA)
Study 2: Validation sample
148 Italian White men aged 3275
Na, Cr in 24-h and timed spot urines
Reliability and Reproducibility
• Tanaka’s method†
• Arithmetic extrapolation method
†: Tanaka T et al. J Hum Hyper,2002;16, 97-103.
Ji C et al. Nutr Metab Cardiovasc Dis 2014; 24: 140-7
Bland-Altman plot comparing estimated 24h UNa
by Tanaka’s method and actually measured 24h UNa
95% CI
Overestimation
Mean
diff
Underestimation
Ji C et al. Nutr Metab Cardiovasc Dis 2014; 24: 140-7
95% CI
Bland-Altman plot
Tanaka’s method
Ji C et al. Nutr Metab Cardiovasc Dis 2014; 24: 140-7
The Prospective Urban Rural
Epidemiological (PURE) Study
The Prospective Urban Rural Epidemiological (PURE) Study:
SUMMARY
To assess the validity of three methods (Kawasaki, INTERSALT, Tanaka) for estimating 24h
urinary sodium excretion using spot urine samples against measured 24h urinary sodium
excretion in a Chinese sample.
Sub-study of the PURE study (120 participants 35-70 yrs; 4 excluded – final n=116)
Morning fasting time urine and 24h urine
Mean bias for the Kawasaki method was the lowest among the three methods, that for the
INTERSALT method was the highest.
Bland-Altman plots indicated that all three methods underestimated 24-h urinary sodium
excretion.
Among the three methods, the Kawasaki method was least biased, but was still relatively
inaccurate.
A more accurate method is needed to estimate the 24-h urinary sodium excretion from spot
urine.
Peng Y et al. PLoS ONE 2016; DOI: 10.1371/journal.pone.0149655
Peng Y et al. PLoS ONE 2016; DOI: 10.1371/journal.pone.0149655
The PURE Study
•
We compared estimates of 24-h sodium excretion from a single morning fasting urine (MFU) using
three different formulae in healthy individuals.
•
We studied 1083 individuals aged 35-70 years from the general population in 11 countries. A 24-h
urine and MFU specimen were obtained from each individual. A subset of 448 individuals repeated
the measures after 30-90 days. The Kawasaki, Tanaka, and INTERSALT formulae were used to
estimate urinary excretion from a MFU specimen.
•
The intraclass correlation coefficient (ICC) between estimated and measured sodium excretion was
higher with Kawasaki compared with INTERSALT and Tanaka formulae (P <0.001). The degree of
bias (vs. the 24-h urine) for sodium was smaller with Kawasaki compared with INTERSALT and
Tanaka formulae (P <0.001 and P = 0.02, respectively).
•
In a diverse population, the Kawasaki formula is the most valid and least biased method of
estimating 24-h sodium excretion from a single MFU and is suitable for population studies.
•
‘First void’ (overnight) urine sample
•
Kawasaki equation designed to assess ‘second’ void morning samples
•
Very high rate of incomplete collections (>50%)
•
Bias described higher at higher levels of 24h urinary sodium, likely due to incompleteness
Campbell N. J Hypertens 2014; 32: 2499-503
Mente A et al. J Hypertens 2014; 32: 1005-14
Mean bias in predicted minus measured 24h urinary
sodium excretion on the same day by prediction model and
time of day*
* Timed urine samples not independent of 24h sample
Cogswell ME et al. Am J Clin Nutr 2013; 98: 1502-13
Bland-Altman plots of the relative bias between measured and
predicted 24h urinary sodium excretion based on INTERSALT
equation
Morning*
Afternoon*
Evening*
Overnight*
* Timed urine samples not independent of 24h sample
Cogswell ME et al. Am J Clin Nutr 2013; 98: 1502-13
Conclusions
• Spot timed urinary Na does not provide reliable
and reproducible estimates of 24-h urinary Na
excretion in an individual.
• Spot urines should not be used in cohort studies
to estimate individuals’ exposures
• 24-h urinary collection for the measurement of
urinary Na excretion remains the preferred tool
for assessing salt intake.
• Multiple collections are necessary to increase the
prediction.
1.
2.
………………………………………………………
3.
4.
24-hour urine is the most accurate method for measuring
population salt intake if carried out properly and measures
are taken to ensure that the 24-hour urine collections are
complete. Experts with experience in 24 hour urine
collection e.g. endocrinologists, should be engaged to coordinate the surveys.
If 24-hour urine collection is not feasible, there is some
evidence that in a generally healthy adult (18-69) population,
spot urine could approximate mean sodium intake and
identify populations where salt intake is above the 5gram
target. It is noted that this is a subject of ongoing research.
To detect change over time in population salt intake, 24-hour
urine is the most accurate method for repeat measurement
but if not feasible, spot urine could approximate the mean
sodium intake with some significant degree of
underestimation of the change over time to take into
consideration
24-hour urine is the most accurate method of assessing
individual sodium intake and single spot or short duration
timed urine samples are inappropriate to use for this
purpose