Assessing differences in clinical trials comparing surgical vs nonsurgical therapy: using common (statistical) sense.

Abstract

The statement of hypotheses and choice of statistical tests in clinical trials that compare surgical with nonsurgical treatment are complicated by the likelihood of excess risk in the surgical group during the perioperative period but lower risk after that compared with the more uniform risk in the nonsurgical group. Commonly used statistical survival analyses implicitly assume a constant ratio of risks in the 2 groups during the follow-up period. However, the changing pattern of risk for one treatment but not the other implies that the assessment of the relative efficacy of the treatments varies with the length of the follow-up. As such, determining whether survival curves for the 2 groups are different may not translate easily into selecting the best treatment. Alternative statements of the hypothesis based on consideration of the time horizon of patients and on clinical judgment may be more consistent with the goals of the study. Regardless of the choice of a statistical test, the choice of treatment is a decision specific to the individual patient and should be influenced by the patient's life expectancy, attitude toward taking risks, quality of life, and cost considerations. When the survival curves cross, there is a trade-off between the risk of surgery and the increase in life expectancy among the survivors of surgery. Accordingly, assessment of differences in outcomes in clinical trials comparing surgical vs nonsurgical therapy should provide both a conclusion about whether 1 treatment can reasonably be considered best for most patients and should provide information to the individual patient and physician on the expected outcome to aid in the decision-making process.