Determining causation--a case study: adrenocorticosteroids and osteoporosis. Should the fear of inducing clinically important osteoporosis influence the decision to prescribe adrenocorticosteroids?

It is generally accepted that exogenous adrenocorticosteroids cause clinically important osteoporosis. We have reviewed the evidence regarding causation in two stages: an examination of the strength of the methods used, and an application of five "diagnostic tests" for causation. The methods that have been used to investigate the association are weak: there have been no randomized clinical trials or prospective cohort studies. The measures of bone density used to quantify osteoporosis do not bear a close relation to clinically important outcomes. Nine analytic surveys and two before-after studies have examined the relation between steroids and osteoporosis. Although some have shown a strong relation, this finding has not been consistent. Evidence regarding temporality , dose-response gradient, and underlying mechanisms are conflicting. The available evidence does not substantiate a causal role of exogenous adrenocorticosteroids in producing clinically important osteoporosis, and does not support withholding steroid therapy on the basis of fears of osteoporosis induced pain and disability.