Cerebrospinal fluid cytotoxicity in amyotrophic lateral sclerosis and sample size
To our knowledge, our study is the first to analyze the effect of cytotoxicity in the outcome amyotrophic lateral sclerosis (ALS). Thus, it is not possible to calculate the sample size needed accurately prior to the onset of the research, as Dr. Pakzad and Dr. Safiri request. Our study included a cohort of patients with ALS in which a lumbar puncture was performed at the moment of diagnosis. Cytotoxicity effect was analyzed in motor neuron primary cultures, and all patients were followed up until death. The absence of an effect on the outcome suggests that the factors present in CSF involved in cytotoxicity do not modify the prognosis of the disease. Considering a type 1 error of 0.05 and a type 2 error of 0.2, our sample had allowed finding significant differences using Cox regression of at least 2.9 HR. Furthermore, we also compared the mean survivals (diagnosis to death, onset of symptoms to death, diagnosis to gastrostomy, diagnosis to respiratory assistance) using Mann‐Whitney U test, with a very similar results and with no significant differences. If CSF cytotoxicity were involved into the outcome, the differences should probably be large due to a constant effect during the course of the disease, and sample size of our study should be enough. Another interesting issue, beyond the statistical significance, is the clinical significance of the results. Larger samples could detect minimal differences between groups, but these differences may not be clinically significant. Moreover, ALS is an infrequent disease, and the aim of the study was to analyze the capacity of an examination of CSF (collected by a lumbar puncture, an invasive test) to predict the outcome. In this regard, most studies in the literature using CSF in ALS include similar sample sizes to our study.
In addition, other factors are very relevant when a clinical study is designed. Increasing sample size may be associated with an increase in heterogeneity, and this is a source of biases in the selection or during the follow‐up. In this regard, in a mortality study, heterogeneity in the clinical decision‐making may impact on the outcome. Our study was performed in the setting of a strict clinical pathway, where there is homogeneity in the clinical decisions and the possibility of biases is lower. This provides more power to the sample. Strict criteria in the selection and follow‐up may be more relevant than sample size, as has been previously suggested.2
In conclusion, we appreciate the comments and suggestions of the authors and we are aware of the importance of sample size in the design of any study. However, it is necessary to remind that studies must be designed to answer specific research questions and taking into account the specific disease to study.