• oral contraceptives
• rheumatoid arthritis
• androgens

In 1993 I reviewed the evidence that rheumatoid arthritis (RA) is more common in women than men; that remissions commonly occur in pregnancy; and that relapses commonly occur post partum.¹ These features suggested involvement of the sex hormones, and in view of the lack of persuasive evidence relating high oestrogen levels to RA, I suggested that one cause might be low androgen levels. Cross sectional studies relating oral contraceptives (OCs) to RA (despite large numbers of cases and carefully selected controls) had proved inconclusive (not to say contradictory). So I reconciled these data with my hypothesis by offering grounds for supposing that OC use is (initially) a marker for high steroid hormone levels. It is worth repeating the argument because, in principle, it might explain contradictory findings relating RA to behavioural risk factors other than OC use (for example, smoking and alcohol consumption). I cited evidence for three propositions—namely, that

• Low androgen concentrations are a cause of RA

• Women who choose OCs initially have higher androgen concentrations than other women

• A pharmacological effect of OCs is to lower women’s androgen concentrations.

The suggestion is that in cross sectional studies, androgens act as confounders between OC use (treated as a risk factor) and RA. Indeed, the principle has been generalised to suggest that gonadal hormones confound between a number of behavioural risk factors (OC use, smoking, alcohol consumption, vasectomy, induced abortion) and a number of disease conditions that are known or suspected to be caused by high or low hormone levels.²

The point of these comments will now become clear. In a prospective study, Drossaers-Bakker et al found that despite the number of reports from cross sectional studies of a negative (protective?) association between OC use and RA, OC use does not significantly influence outcome in long term RA.³ I suggest that this is as would have been predicted on the basis of my hypothesis.

In my original paper¹ I cited persuasive evidence that both male and female patients with RA have lower androgen levels than healthy controls; and this conclusion has been corroborated by subsequent evidence.^4,5 However, the possibility may be acknowledged that low androgen levels are markers (rather than causes) of RA. But it is now becoming clear that (counter to this possibility), administration of androgens to patients with RA ameliorates the condition.^6,7 There may be several agents which cause or exacerbate RA: I suggest that in some cases, low androgen levels are one of them.0