Survival in cancer patients hospitalized for psoriasis: a commentary

Abstract

Conflicts of interest None declared. Cohort studies should include in their reporting (and so in the design) items of the STROBE statement (http://www.strobe‐statement.org) in order to avoid bias. This study fulfils many items but not all.1 The bias included should be taken into account when interpreting the results. The study is an extension of Scandinavian linkage studies from the 1990s demonstrating that patients hospitalized with psoriasis were at a higher risk of developing (smoking‐ and alcohol‐related) cancers compared with controls.2 In the introduction of this study, it is hypothesized that cancer survival is negatively affected by the systemic dysregulation of the immune system of patients with psoriasis. The main advantage of this linkage study of the Swedish Hospital Discharge Registry and the Swedish Cancer Registry is that it is population based and generates a relatively large number of patients with psoriasis. However, studying very specific patient populations such as hospitalized patients with cancer reduces the sample size rapidly. Unfortunately, no sample‐size calculations were performed to estimate the number of patients needed to achieve reliable and valid risk estimates. The authors have restricted the analysis to cancer sites that included at least 10 cases who died during the observation period, but this was not based on formal sample‐size calculations. Moreover, investigating different cancer sites may provide interesting information, but it does introduce the issue of multiple testing (> 150 survival analyses were performed implying that by chance at least seven analyses have a P‐value < 0·05, and P‐values of < 0·002 were required to be significant after Bonferroni adjustment). Therefore, a clear definition of the primary outcome is essential and subgroup analysis should be interpreted with caution.3