In and Out, Good and Bad News, of Generalizability of SWOG Treatment Trial Results

Abstract

The issue of generalizability of clinical trial results has challenged researchers and practitioners of adult cancer patients for years. It is one reason why the adult cancer cooperative group enterprise in the United States is undergoing yet another transformation (1). Figure 1 may capture the essence of the situation. Its foreground shows the most recent data available to this commentator on the estimated proportion of cancer patients in the United States entered onto cooperative group treatment trials as a function of individual year of patient age. The background shows the rest of the United States. With only 2% to 4% of adults aged 20 to 65 years entered onto clinical trials, how can we rely on such a small sample size to be applicable to the general population? In a comparison of survival outcomes among cancer patients treated in and out of clinical trials published in this issue of the Journal, Unger and colleagues analyzed 21 SWOG treatment trials enrolled during the period from 1986 to 2007 with comparable patients in the US Surveillance, Epidemiology and End Results (SEER) database (2). Their primary subset analysis was good- vs poor-prognosis trials defined by a 2-year average study-specific survival rate of greater than 50% or less than or equal to 50%, respectively. The former had 11 trials that included melanoma, myeloma, and carcinoma of the breast, bladder, uterine cervix, and prostate. The latter had 10 trials that included acute myelogenous leukemia, brain tumors, lung cancer, and pancreatic carcinoma. The results may be interpreted as a series of good- vs bad-news scenarios (Table 1), with several advantages as a source of disadvantage and vice versa. Among the best news is that patients in SWOG trials had a survival benefit, albeit only in poor-prognosis trials. The bad news for them was that it “endured for only 1 year” (2). However, according to Supplementary Figure 2 in the Unger et al. report (2), which shows the SWOG and SEER survival curves for each of 10 poor-prognosis trials, the survival benefit is better than the authors concluded, lasting at least 5 years in two trials, 3 years in three trials, and 2 years in two trials. In only two of the 10 trials was it less than 1 year. Whether or not the apparent benefit of the SWOG trial in patients with a poor prognosis is more durable than concluded, the striking difference between the broad categories of good and poor prognosis as defined has major implications for the interpretation of cancer clinical trial outcomes and application to the general population: