Ricerc@Sapienza

To test the validity of the model and answer the second question by Wang et al.,2 considering the small sample of the population, we implemented the analysis including the cross-validation with bootstrapping. To quantify the discrimination performance of the model, Harrell’s C-index was measured. The model was subjected to bootstrapping validation (1000 bootstrap resamples) to calculate the optimistic corrected estimate of C-index. The C-index for OS models was 0.70 while the corrected C-index was 0.67. Despite the small sample, the conclusions of our model seem to be robust. The potential for significant bias has already been mentioned in our report1 and the results must be considered cautiously. Large-scale studies are warranted.