We reanalysed a large phase I trial of Irinotecan that included 80 children (32 heavily pretreated patients and 48 less heavily pretreated). An extended simulation study was performed to investigate the robustness of the conclusions in the context of small sample sizes. Dose recommendations were studied according to scenarios with group differences, as measured by odds ratio (OR), ranging from 1 (no difference) to 10 (large difference) and sample sizes increasing from 20 × 2 to 60 × 2 patients.
This study shows a high risk of misidentification of the MTD in each of the two groups, regardless of the group difference. With a group difference corresponding to OR = 8 and balanced sample sizes (20 × 2 patients), the same MTD was identified in 11 % of the simulations. Even with larger sample sizes (40 × 2 patients), this figure reached 24 % for OR = 3. There is also a very high risk of identifying two different MTD (52 % for 40 × 2 patients) although the risk is similar in both groups.
Two independent sample designs in paediatric phase I trials should be avoided or reserved to limited situations when there is a strong rationale possibly based on adult data.