have assumed that the analyses were by per-protocol because in table 2 of our manuscript, the sample sizes for the analyses at 3-4 month were less than those for the analyses at 1 week. The former comprised subjects for whom outcome data were available, not subjects who adhered to the intent of treatment. Indeed, there was no 'cross-over' of subjects in our study, which means the results of analyses by intention-to-treat would have been the same as by results per-protocol anyway.An important point was made regarding vitamin D levels increasing in the daily group at 4 months despite a poor adherence rate of 33%. There are plausible explanations for this. Following production in skin from ultraviolet B light exposure, vitamin D undergoes hydroxylation, to form 25OHD (the form measured in routine vitamin D assays to assess vitamin D status) and then to the biologically active form of 1,25(OH) 2 D. Both 25OHD and 1,25(OH) 2 D are then inactivated by 24 hydroxylase to form 24,25(OH) 2 D and 1,24,25(OH) 3 D, respectively. We hypothesise that when this system is exposed to a high bolus dose of vitamin D, metabolism to its inactive form is upregulated, thereby resulting in a substantial decline and lower mean 25OHD at 4 months.In addition, the vitamin D supplement is unhydroxylated and, hence, is not measured in the assay used. It is unlikely that taking vitamin D supplements just prior to the 4-month blood test would explain why vitamin D levels in the daily group are increasing despite a 33% non-adherence rate. However, most poorly adherent infants (9/12) were on mixed feeding regimens and consequently received some vitamin D in their formula feeds. We also reiterate that while there was statistical significance between mean 25OHD levels between both groups at 4 months, this was not clinically significant as these values were greater than 50 nmol/L and therefore considered replete for clinical purposes.