Post hoc subgroup analysis.

I am surprised by the published conclusions of Hung et al 1 in a recent issue of CHEST (August 2013). The authors’ stated meth ods are not adequate to support their conclusion that hyperimmune IV immunoglobulin (H-IVIG) ben efi ts mortality if given within 5 days to patients with severe 2009 infl uenza A(H1N1) infection. The primary outcome analysis of 34 patients presented in Table 1 of the study 1 shows that fi ve patients who received H-IVIG died, and four control subjects died. The authors subsequently performed a subgroup analysis of 22 patients who received treatment within 5 days of symptom onset. All fi ve of the H-IVIG fatalities were excluded from this analysis, but all four of the control subjects’ deaths were retained. This is not explicitly explained by the authors but can be seen by comparing survival data in Tables 1 and 2. 1 There is no mention of any plan to perform this subgroup analysis in the Materials and Methods section; therefore, there is no guarantee that this subgroup was not specifi cally formulated to elicit a presupposed conclusion. This is an excellent example of why unplanned subgroup analysis should not be acceptable as a basis for scientifi c conclusions.