select search filters
briefings
roundups & rapid reactions
before the headlines
Fiona fox's blog
This analysis accompanied a roundup which can be viewed here.
Long term alcohol intake and risk of rheumatoid arthritis in women: a population based cohort study, 10 July 2012, BMJ
The paper provides weak evidence of a relationship between long-term increased alcohol consumption and decreased risk of Rheumatoid Arthritis (RA) in women.
Observational study; highly possible that any observed associations due confounding effects of unmeasured variables (see below).
The paper’s main conclusion is that “moderate consumption of alcohol is associated with reduced risk of rheumatoid arthritis.”
The paper does not adequately differentiate between moderate and high alcohol consumption. These groups are lumped together.
The study uses inconsistent definitions of moderate/high consumption for different analyses that may potentially affect conclusions. The finding of 52% decreased risk for women drinking more than 3 glasses of alcohol per week compared with never-drinkers (based on both 1987 and 1997 assessments) was not mentioned in the statistical analysis section of the methods, arguably weakening any resultant claims. The result is only just statistically significant (confidence interval just excluding 1) and is based on just 14 cases (>3 glasses/week in both 1987 and 1997) vs. just 21 cases (never drinkers). Other relative risk estimates plus confidence intervals from Table 4 do not appear to provide additional evidence to support this headline figure.
No statistically significant association was found between increased consumption of any of wine, beer or liquor (when analysed separately) and incidence of RA, though the results seem to trend in that direction.
Analysis of headline result (52% decrease for >3 glasses/week vs. never drinker) was not at all clearly mentioned in methods section. In other words, why was 3 drinks a week used as a cut-off? We cannot know from the data whether 4, 5 or more drinks a week had a similar effect.
A large, prospective population-based cohort design does provide some strength, as ascertainment of outcome and potential risk factor were independent.
One strength is the attempt to examine alcohol consumption in different ways, for example by splitting into type of alcohol.
The authors acknowledge that they could not evaluate the effect of high doses of alcohol on risk of RA because of lack of heavy drinkers in the cohort.
Potential confounders were not adequately adjusted for, especially co-morbidities, which is a major limitation. Women who already suffered from another disease may have been resultantly unable to drink, with the co-morbidity perhaps putting them at higher risk of developing RA. Conversely it could be argued that only healthier women were able to drink, and so may have been inherently less likely to develop RA.
They also point out that they did not measure family history of RA, a proxy for genetic/environmental risk factors. Additionally they say that alcohol consumption may have been under reported, as it could be seen as a socially undesirable behaviour.
When people didn’t report their drinking frequency (10.3% for beer, 3.9% for wine, and 12.1% for liquor) this was assumed to mean none or seldom consumption. This assumption should be strongly questioned and its sensitivity assessed by statistical methods. Whether this was done was unclear.
Women who completed the second questionnaire in 1997 but were diagnosed with RA (or a non-RA joint condition, or died) before study start in 2003 were excluded as they would not have been incident cases for the study. In the study there were 197 new cases of RA. There were 443 cases of RA before study start, as well as 1559 deaths. It is highly plausible that women who were ‘going to get sick’, were already sick, and thus excluded from the study. This could have influenced the results in either direction.
Cohort study: a way of analysing risk factors by following a group (cohort) of people over time, and observing who experiences the outcome(s) of interest; no interventions take place, unlike a clinical trial.
No
‘Before the headlines’ is a service provided to the SMC by volunteer statisticians: members of the Royal Statistical Society (RSS), Statisticians in the Pharmaceutical Industry (PSI) and experienced statisticians in academia and research. A list of contributors, including affiliations, is available here.