Article Text
PDF
Statistics from Altmetric.com
Objective
To establish whether measurement error introduced by imprecise measurement of alcohol intake introduces may lead to incorrect guidance on safe levels of alcohol consumption.
Design and Setting
Using repeat alcohol intake from the United Kingdom Women’s Cohort Study, a prospective, longitudinal cohort of UK women recruited in 1995.
Participants
33 732 women reporting alcohol intake using both frequency and quantity of specific drinks flagged with the Office for National Statistics. Repeat questionnaires with were available on 1918 women (5%).
Main Outcome Measures
Death from ischaemic heart disease, fitting a fractional polynomial logistic regression model to the expected nonlinear curve, adjusting just for age at baseline.
Results
The mean age of the women at baseline (in 1995) was 52 (SD = 9). Mean alcohol intake was 9 (SD = 11) g/day, i.e. a mean of approximately one unit of alcohol per day, or seven units per week. The intraclass correlation between repeat measures of alcohol intake was 0.79 (95% CI 0.77 to 0.80). 133 women died from ischaemic heart disease. Without correction for measurement error, the estimated range of protective effect of alcohol consumption (compared to non-consumers) was from 0 to 89 g/day (11 units/day), statistically significant (p<0.05) up to 46 g/day (6 units/day), and with the lowest point of the curve at 13 g/day (1.6 units/day). With correction for measurement error, the estimated range of protective effect of alcohol consumption (compared to non-consumers) was lower than without adjustment for measurement error, from 0 to 55 g/day (7 units/day), statistically significant up to 37 g/day (5 units/day), and with the lowest point of the curve at 8 g/day (1 unit/day).
Conclusions
Adjusting for measurement error led to protective effects at substantially lower intakes than ignoring measurement error. Current guidelines recommend limiting alcohol intake for women to less than 16 to 24 g/day, stating that 1 to 2 units/day can help protect against coronary heart disease, but these ignore measurement error. Correction for imprecise measures of long term alcohol intake may lead to substantially reduced recommended limits.