n the intervention group), which was statistically significant both in the adjusted and unadjusted analyses (table 1). However, despite the relatively modest reduction for the other components of the primary endpoint the overall infectious score was significantly reduced—mainly as a result of the large effect on the antibiotic parameter—both in the unadjusted and in the adjusted analyses (table 1 and figure 2). It is important to interpret the statistical significance in light of our power calculation, which was based on a significance level of p=0.02. In the power calculation, the significance level was reduced from 0.05 to 0.02 in order to increase the statistical power at the p=0.05 level. This approach was incorrect, and the targeted power (at the p=0.05 level) should instead have been increased without altering the p-value threshold. However, we have used the widely accepted significance level p=0.05 in the statistical analyses for both the primary and secondary endpoints, respectively. Another potential problem was that the patient population was very heterogeneous with regard to immune deficiency and concomitant diseases. We adjusted for these factors in the multivariable analyses of the primary endpoint, but the sample sizes in each subgroup were too small to draw any conclusions of effects in specific disease groups. However, a detailed post hoc analysis of the relation between immunological diagnosis, concomitant lung disease and the secondary endpoints ‘taken bacterial cultures’, ‘positive bacterial’ cultures and ‘microbiological findings’ was performed. There was a clear trend that vitamin D-treated patients with subclass deficiency and/or asthma produced fewer bacterial cultures, fewer positive cultures and fewer fungal cultures (see online supplementary tables S4 and S5 and figure S3). Although this analysis may lack precision by the small number of patients included, it could have clinical implications regarding target groups for vitamin D3