To the Editor: In the absence of randomized clinical trials, comparative effectiveness research plays a critical role in evaluating new technologies. However, because these studies invariably rely on observational data, the devil is in the details.
The recent study by Dr Sheets and colleagues1 related to the comparative effectiveness of radiation therapy modalities for prostate cancer highlights this concern in 2 ways. The first relates to treatment assignment, or in this case, the type of radiation. The authors broadly assigned patients to conformal therapy (the predecessor to intensity-modulated radiation therapy [IMRT] and proton beam), invariably including patients with conventional, stereotactic, and neutron beam therapies. It is established in randomized trials that conventional therapy is associated with greater morbidity and is less effective than conformal radiation for prostate cancer.2 In fact, the use of conventional therapy during the period of study was significant.3 Given the relatively small effect size described in the article, such misclassification might explain the observed differences. An alternative approach would have been to incorporate treatment planning codes specific to conformal therapy to obtain a more homogenous comparison group.3,4