What is the risk of extraspinal hyperostosis in a patient with psoriasis treated for several months with acitretin? Does PUVA therapy increase the risk of non-melanoma skin cancer in a patient being treated for mycosis fungoides? What is the chance of severe depression in an adolescent taking 13-c/s-retinoic acid for acne? To address these questions, physicians must effectively search for evidence and must be able to assess the validity of the available data, and consider the strength of any documented association, and the relevance when the issue is applied back to an individual patient.21
We have already considered that many different sources of information should be sought and the search should not be limited to RCTs.22 When data from RCTs are scrutinised, the statistical power to detect an important adverse event should be taken into account. When dealing with observational studies, it should be carefully considered if the studies provide reliable quantitative risk estimates or simply generate signals needing further validation. The optimal study design should be one assuring unbiased comparison between exposed and unexposed groups. Comparison groups should be similar with respect to important determinants of outcome. Outcome and exposure should be measured in the same way in the groups being compared, and the exposure should clearly precede the adverse outcome. In addition, follow up should be sufficiently long and complete and the study should have enough statistical power to document the association of interest.
When risk estimates from several studies are available, one should evaluate whether these are roughly in the same direction or if there are discrepancies between the studies. If there are discrepancies between studies, reasons for the discrepancies should be considered. Systematic reviews may help in summarising the study results.23 Unfortunately, the criteria for assessing the quality of meta-analyses of observational studies are not as well established as those concerning RCTs.24
Once an association has been established, the magnitude of the risk should be taken into account and expressed in understandable terms if it is to be of practical use to clinicians. We have already considered that, from the perspective of a physician deciding about the risk of prescribing a particular drug, the excess risk (or risk difference) is a more informative measure than is the relative risk. In the context of RCTs, Sackett et al. proposed a method for converting risk differences into a more intuitive quantity. This quantity was named the number needed to treat (NNT = 1/excess risk).25 It is the number of people who must be treated in order that one clinical event is prevented by the treatment at issue (for example the number of people to be treated to avoid one patient experiencing a relapse of psoriasis) or one additional beneficial outcome is achieved. By analogy, the "number needed to harm" (NNH) or "number of patients needed to be treated for one additional patient to be harmed"26 is the number of people exposed to a given treatment such that on average and over a given follow up period, one additional person experiences an adverse effect of interest because of the treatment. In RCTs and cohort studies, NNH is directly calculated as the reciprocal of the excess risk. Recently, a formula has been proposed using odds ratios from case-control studies and data on the event rate in the unexposed population (Table 10.1).27 According to the formula, given an odds ratio of 3 and unexposed event rate of 1 per 1 000 000 people, the NNTH can be calculated as 500 000 (i.e. 500 000 people to be treated to experience one additional adverse effect with the treatment).
After deriving estimates for the potential harm of an intervention, the estimates should be weighted against the expected benefits of the same intervention. The adverse consequence of withholding the intervention should be carefully considered. A final decision should try to integrate probability issues with the patient's values and preferences about therapy. This requires patient education about the benefits and risks of alternatives, tailored to the particular patient's risk profile.
To conclude, not only should physicians be able to retrieve and critically assess the evidence concerning the safety of any given intervention, they should also take an active part in promoting safety by contributing to surveillance programmes once an intervention is proposed to the medical community.
Was this article helpful?
Complete Guide to Preventing Skin Cancer. We all know enough to fear the name, just as we do the words tumor and malignant. But apart from that, most of us know very little at all about cancer, especially skin cancer in itself. If I were to ask you to tell me about skin cancer right now, what would you say? Apart from the fact that its a cancer on the skin, that is.