Meta-analyses of data from randomized clinical trials (RCTs) are often used by hematologists to compare the efficacy of
therapies of
blood diseases. This is especially so when results of RCTs are not decisive. This situation in RCTs arises when the magnitude of differences in treatment outcomes between
therapies tested is small, when trials are unpowered to detect differences (these are confounded) and/or when RCTs reach, or seem to reach, contradictory conclusions. Contributing to these limitations of RCTs are the relative rarity of many
blood diseases, poor recruitment into RCTs and the greater interest of many hematologists in
therapy strategy than in a direct comparison of alternate
therapies. These limitations of RCTs are solvable, but only in part, by meta-analyses. Adding data from high-quality observational database studies(ODBs) to meta-analyses is sometimes useful in resolving controversies, but this approach also has limitations: biases may be difficult or impossible to identify and/or to adjust for. However, ODBs have large numbers of diverse subjects receiving diverse
therapies and adding these data to meta-analyses sometimes gives answers more useful to clinicians than meta-analyses of RCTs alone. Side-by-side comparisons suggest analyses from high-quality ODBs often give similar conclusions as meta-analyses of high-quality RCTs. Quantification of expert opinion of high quality is also sometimes useful: experts rarely disagree under precisely defined circumstances and their consensus conclusions are often concordant with results of meta-analyses of high-quality RCTs with and without ODBs. We conclude that meta-analyses are often helpful to determine the best
therapy of
blood diseases. Accuracy can be improved by including data from high-quality ODBs, when appropriate, and by resolving discordances, if any, with conclusions from high-quality ODBs and from quantification of expert opinion.