Can we inflate effect size and thus increase chances of producing "positive" results if we raise the baseline threshold in schizophrenia trials?

The standardized mean difference (SMD), also referred to simply as effect size, is often used to summarize the results of a clinical trial when the outcome measure is continuous. SMD is calculated by dividing the difference in the mean scores of the experimental and control groups by their standard deviation (SD). One of the major arguments against SMD is that, if the studied sample is chosen to be artificially homogeneous and thus have a small SD, SMD can be overestimated. On the other hand, smaller SDs raise the chances of finding a statistically significant difference. This study examined whether we can increase sample homogeneity and decrease SD by raising the severity threshold to enter a clinical trial in secondary analyses of individual patient data from three large acute phase schizophrenia trials. Raising the baseline threshold on PANSS and BPRS did reduce the SDs at baseline but SMDs at endpoint remained by and large constant. It is concluded that restricting the entry criteria into schizophrenia trials cannot lead to larger SMDs or to smaller sample size necessary to detect an efficacy signal.     «

The standardized mean difference (SMD), also referred to simply as effect size, is often used to summarize the results of a clinical trial when the outcome measure is continuous. SMD is calculated by dividing the difference in the mean scores of the experimental and control groups by their standard deviation (SD). One of the major arguments against SMD is that, if the studied sample is chosen to be artificially homogeneous and thus have a small SD, SMD can be overestimated. On the other hand, sm...     »