For personal use. Only reproduce with permission from The Lancet Publishing Group. Is there a place for leflunomide in the treatment of rheumatoid arthritis

2children with intussusception were more likely to be boys, Hispanic or black, and have mothers with low levels of education. Among control children factors associated with having had rotavirus vaccine were being white and having a mother with a high level of education. So here is clear evidence of potential negative confounding—children who get vaccine are those with a low risk of disease. Unless this confounder is adjusted for, any relation between vaccine coverage and risk of disease is potentially disguised and certainly the size of any effect is likely to be reduced. The CDC group has also carried out an historical cohort study of children enrolled in Health Maintenance Organisations 3 which confirmed the results of the casecontrol study. There was a striking similarity in the estimate of the relative risk between the studies—the risk of intussusception in the 3–7 days after rotavirus vaccine was 30-fold higher in the cohort study and 37-fold in the casecontrol study. This strikingly high relative risk, which is 3–4 times higher than the association between smoking and lung cancer, is strong evidence of causality. Adjustment for potential confounders in the case-control study changed an unadjusted estimate from 28 to 38, and this clear evidence of confounding has implications for the interpretation of ecological studies. The size of the risk of intussusception is important because policy will be determined by weighing this risk against the benefits of the vaccine. So what does this ecological analysis add to this policy debate? Probably little given the difficulty of interpretation. One of the weaknesses of Simonsen and co-workers’ study is the lack of confirmation of the diagnosis of intussusception, which could have been done by examining medical records. Nor were vaccine records obtained, which would have yielded information on potential confounders. This lack means they could not look at anything more than a wide time window of risk over which they assumed children received the first dose of vaccine. Had they had vaccination dates they could have narrowed this window. The study illustrates another saying of Geoffrey Rose: “An epidemiologist needs dirty hands and a clean mind”—dirty hands from collecting all of those confounding variables and vaccine histories, and clean minds with which to judge the evidence. Good quantitative estimates in epidemiology require precise data.