To the Editor.
—Dr Greenfield and colleagues1 concluded that there were no meaningful differences in health outcomes for patients with hypertension or non—insulin-dependent diabetes mellitus (NIDDM) in the Medical Outcomes Study (MOS), whether their patients were treated by different systems of care or by different physician specialists. While the authors acknowledge the study's limitations regarding "nonrandomized circumstances" and "[n]o attempt... to equalize important variables," their conclusions seem inconsistent with the statistical implications of such limitations.First, conclusions of equivalence or no meaningful differences require designs that can estimate with high statistical power the effects of systems of care and physician specialists independent of extraneous confounders such as age, severity of disease, and income. Because this was a nonrandomized observational study requiring extensive covariate adjustment, NIDDM comparison sample sizes of 77, 75, and 18 (Table 3) and 109,27, and 31 (Table 4) offer extremely low power, which practically guarantees that
Testa MA. The Medical Outcomes Study. JAMA. 1996;275(14):1083. doi:10.1001/jama.1996.03530380025014