Experimental vs control interventions in cancer therapy: Which is better? - The Children’s Oncology Group experience.
Kumar A., Soares HP., Wells RJ., Khayat A., Clarke M., Hills RK., Bleyer A., Reaman G., Djulbegovic B.
Background: The outcomes in childhood cancers have dramatically improved since 1950. However, it is unclear to what extent this improvement has been a result of continuous progress with experimental treatments, or a consequence of variable successes and failures of the new treatments according to the equipoise hypothesis (Cur Oncol Rep 2001;3:389). Methods: To accurately describe the outcomes of innovations, tested in randomized controlled trials (RCTs), three factors are paramount: publication rate, quality of trials and the choice of control intervention. We reviewed all completed phase III RCTs conducted by the Children’s Oncology Group (1972 to 2000). Published and unpublished data were used. Seventy three (104 comparisons) RCTs were identified. The success rate of experimental vs. standard treatment was evaluated both semi-quantitatively (on a scale 1?6 describing the authors’ conclusions about treatment success) and quantitatively using a standard meta-analytic technique. Results: Overall, the trials were of high methodological quality. Publication rate was 97% (72/74; 4 abstracts, 68 papers). Innovative treatment was favored in 45% of trials, with a hazard ratio (HR) of 0.91 (95% CI 0.86, 0.96) for event-free survival (p=0.0001), and a HR 0.96 (95% CI 0.89, 1.05) for overall survival (p=0.3). There was no difference in results when restricting analyses to primary outcome measures or according to the type of treatment, disease or comparator. Conclusions: There is no predictable pattern of innovative successes in pediatric oncology. Even in situations where in retrospect, tremendous improvements in therapeutic outcomes occur, there is no evidence that the outcomes of RCTs, in aggregate, consistently and predictably favor experimental over control interventions, in spite of the trialists’ hopes that the new interventions are better. Nevertheless, individual experimental treatments can be more or less successful. However, identifying which experimental treatment is a success is only known after a RCT is conducted. Improvement in clinical outcomes can only come from continuing empirical testing of new treatments in RCTs