Clinical Trials Comparing Surgical vs Nonsurgical Therapy

Publisher Website
Google Scholar
Abstract
1 correctly empha- size that studies comparing surgical vs nonsurgical therapy pose distinctive statistical challenges because the complications of surgery usually occur early, whereas the complications of non- surgical therapy generally occur at a constant rate over time. However, other inherent differences between surgical and non- surgical therapies can invalidate comparisons between them despite the most careful statistical adjustments. 2 Drug therapy is constant throughout the duration of a study since the chemical compound does not change, whereas surgical technique evolves continuously. The operation being used by the end of a study can differ markedly from the one with which the study began. Thus, increased experience with surgery im- proves results and decreases complication rates, whereas in- creased use of drugs usually unveils more complications. In one study, the mortality of coronary bypass surgery decreased from 3 (12%) of 25 patients in 1968 to 2 (1.5%) of 134 patients in 1973. 3 Furthermore, drug therapy in collaborative studies is standard- ized among all participants and is unrelated to physician skill, whereas the quality of surgical therapy varies unavoidably. Crossovers occur from medical to surgical therapy, but the re- verse cannot occur. Drugs also usually have an indistinguish- able placebo. Moreover, when surgery is an alternative, imper- ceptible bias often occurs before randomization. Physicians are likely to refer their sickest patients for the therapy they feel is best, while allowing lower-risk patients to be randomized. 4 This tendency probably explains why randomized studies that com- pare nonsurgical with surgical therapy frequently enroll low- risk subsets that usually fare better than historical controls in both arms of treatment. For these reasons, it is often impossible to make enough statistical adjustments in studies that compare surgical with nonsurgical therapy. The implications are substantial because an increasing number of invasive procedures (comparable to surgical therapy) are being carried out by nonsurgeons, which leads to the possibility of more randomized studies that are unavoidably flawed.

This publication has 0 references indexed in Scilit: