Using ‘number needed to treat’ to express the magnitudes of benefit of ultraviolet B phototherapy and of antitumour necrosis factor‐α therapies for psoriasis

MADAM, The British Journal of Dermatology recently published a review of studies of tumour necrosis factor (TNF)-a inhibitor therapy for psoriasis, expressing the findings of key studies in terms of ‘number needed to treat to benefit’ (NNTB). Use of NNTBs is a useful way to express study results, but it is important to be clear that the NNTBs presented relate to comparisons with placebo. Many studies of TNF-a inhibitor therapies have allowed inclusion of patients who had not received other therapies beyond topical therapies alone, for example, an infliximab randomized comparison with placebo allowed patients to enter if they were ‘considered for phototherapy or systemic therapy’. That is, many patients entering ‘biological’ TNF-a inhibitor therapy studies have not first been treated with phototherapy, psoralen plus ultraviolet A or any other systemic therapy. When interpreting these NNTBs it is important for dermatologists to have an idea what the NNTBs for other therapies are. It is impossible to estimate directly comparable NNTBs for treatments other than the ‘biologicals’ because most earlier studies did not select reduction in Psoriasis Area and Severity Index (PASI) by 75% (PASI 75) as an outcome measure, and because most established treatments have not been compared with placebo. Only one study compared another treatment modality (methotrexate) with ‘biological’ therapy. The only ways, apart from direct comparative studies, to estimate reliably the differences between treatments either (i) involve indirect comparison using a single common comparator (often placebo) or (ii) use a mixed treatment comparison (alternatively called ‘network meta-analysis’) methodology. A mixed treatment comparison method has now been used in comparing various ‘biological’ and other systemic therapies for psoriasis. However, such a method is not possible for comparing outcomes of biological and phototherapy interventions as none of the phototherapies have been directly compared with interventions with which the ‘biologicals’ have been compared. However, I have calculated NNTBs from a placebo-controlled study which included methotrexate as well as adalimumab (Table 1), and estimated NNTBs based on two narrowband ultraviolet B (NB-UVB) phototherapy studies which reported PASI 75 outcomes (Table 2). One of these NB-UVB studies reported unexpectedly (according to our experience of phototherapy) low percentages of participants reaching PASI 75 (possibly because the mean baseline PASI was low at 9Æ15, and it is harder to reduce a low PASI by 75% than it is similarly to reduce a higher PASI); the other reported a higher percentage achieving PASI 75. In the absence of placebo arms to any recent phototherapy studies, these estimates are based (as indicated in Table 2) on assuming various PASI 75 rates for placebo of 3% (as found in one etanercept study), 19% as found in an adalimumab and methotrexate study and 10% as an intermediate estimated value. These estimated NNTBs (Table 2) are similar to those for TNF-a inhibitor therapies (Table 1). These NNTBs are of some value in deciding that a treatment is active. However, when deciding on which particular treatment to use it would be more useful if we could calculate num-