Response: Re: Tandem vs Single Autologous Hematopoietic Cell Transplantation for the Treatment of Multiple Myeloma: A Systematic Review and Meta-analysis

Extract

Dr Mehta's critique reflects several methodological errors that are commonly raised by clinicians and non-statisticians who review meta-analyses. The first error reflects a substitution of dichotomous data analysis for time-dependent analysis. The dichotomous data analysis is based only on the number of events (summarized as odds or risk ratio) ( 1 ), but the time-dependent analysis also takes into account the duration to a given event and “censoring,” which is best summarized as hazard ratios ( 2 ). Therefore, it is incorrect to use time-independent analyses, as reported by Dr Mehta, when the key outcome is time dependent.

Furthermore, Dr Mehta refers to a shift in the overall survival point estimate to justify the use of tandem transplant. However, the point estimate has no relevance without also considering confidence intervals ( 3 ). In addition, he interprets research evidence using the “vote counting” method in which the superiority of tandem over single transplant is determined by the number of individual studies showing some advantage for tandem transplantation. Clearly, this method is inappropriate because it does not take into account effect sizes, number of patients, time to event, or the results that are consistent both with “no evidence of effect” and with “evidence of no effect” ( 4 ). For this reason, meta-analysis is required to take all of these factors into account in summarizing data related to treatment effects ( 4 ).