Dear Editor: Further data regarding the use of specific insulins for continuous subcutaneous insulin infusion (CSII) are always of interest because of the relative paucity of research in this area. As such, the recent article by van Bon et al.1 is of particular value. However, we consider that some of the conclusions, as presented by the authors, deserve greater scrutiny. First, according to the methods, this was a randomized, controlled, open-label, crossover clinical trial designed to show the superiority of insulin glulisine over both insulin aspart and insulin lispro. For the primary outcome (unexplained hyperglycemia and/or perceived perfusion set occlusion), a numerically greater percentage of patients reported this outcome with glulisine (68.4%) compared with either aspart (62.1%) or lispro (67.3%). The authors refer to a lack of statistical significance (P=0.04 for glulisine vs. aspart, P=0.03 for glulisine vs. lispro) for this outcome, implying little to no difference among the three insulins, rather than as a failure to show superiority (which was what the study was powered to detect). Furthermore, there were significant differences among the three insulins in the secondary outcomes of monthly rate of unexplained hyperglycemia or perceived infusion set occlusion, rate of significant hyperketonemia and/or hyperketonemia at risk of ketosis, and rate of symptomatic and nocturnal hypoglycemia. These results seem to belie the conclusions of the authors that there were no significant differences among the three insulins when used in CSII. I would also question the authors' suggestions that a trend toward greater occurrence of unexplained hyperglycemia and perceived infusion set occlusion seen in this study with insulin glulisine compared with insulin aspart and lispro should be taken as balancing the results from the previous study by Hoogma and Schumicki2 in which the opposite trend was seen. As the previous study was not statistically powered to detect superiority for this outcome, whereas the study by van Bon et al.1 was explicitly designed to do so, these results should instead override the previous findings. Second, the article describes various post hoc analyses that were carried out on the final data set but that had not previously been planned or described in the study design. There appears to be no well-defined reason for these additional statistical analyses, and I would query their relevance and utility. The relative value of any results produced by such analyses must be considered of lesser weight than the main study as cohort numbers were not chosen with these in mind, and thus the statistical power may not be sufficient to show a true difference. Third, the authors place a large amount of emphasis on the results relating to time to change infusion set, stressing the point that less difference was seen in occurrence of occlusion between glulisine and aspart or lispro when sets were changed more frequently. This finding is of questionable clinical relevance when compared with the observation that the timings seen in this study, with patients changing catheters much less frequently than advised, are much more likely to be applicable to a real-world scenario than a rigidly enforced limit of every 2 days. Following the results of this study, it seems that patients using glulisine in infusion sets should be instructed to change their infusion sets every 2 days. This requirement for more frequent changes could be due to a lower chemical and physical stability of glulisine compared with lispro and aspart, presenting challenges in the pump setting. Finally, the authors suggest that the increased frequency of hypoglycemia seen with glulisine could be the result of slight overdosing of glulisine throughout this study. They refer to previous studies in which patients have required a lower basal dose of glulisine than other insulins. In this current study, although initial dosage was based on the patient's dose at trial commencement, dosing adjustments for each of the study insulins were based on each patient's glucose control in line with recommendations from the American Diabetes Association. If the glulisine dose was high enough to lead to increased hypoglycemic events, it could be expected that average glycosylated hemoglobin levels in these patients would be decreased compared with aspart and lispro. Throughout the study, glycemic control remained stable across all three treatment groups with no differences observed between glulisine and aspart or glulisine and lispro. Instead of stating in the conclusion that “all three short-acting insulin analogs could be used in CSII as no difference was seen among GLU, ASP, and LIS on the primary outcome measure..,” the authors should have reported that glulisine failed to show superiority over either aspart or lispro, countering the trend seen in the previous trial of Hoogma and Schumicki,2 and was inferior with regard to several of the outcomes investigated.