br Authors contributions br Funding br Conflict of interest
Conflict of interest
Introduction The cancer clinical trial network groups of the National Cancer Institute (NCI) have a history of successful conduct of large randomized phase III trials of chemoprevention for cancer [, , , , ]. However, such trials are expensive, and only a limited number of research questions have found a consensus around which to conduct such trials. An important research question for federal funding agencies is whether large, randomized phase III chemoprevention trials provide good scientific impact for expenditure, regardless of the trial outcome. To examine this, we compared the scientific impact of four large phase III randomized chemoprevention studies to the impact of a series of randomized phase III treatment trials conducted over the same time period. We used a bibliometric technique, facilitated by the strategic CAY10683 of multiple databases, including citation data and the study and publication databases of multiple NCI network groups.
Discussion The secondary science derived from the four chemoprevention trials was much greater than for most of the treatment trials we examined. Moreover, articles from chemoprevention trials were more often published in non-oncology journals, suggesting that chemoprevention trial findings have a wider scientific impact than oncology alone. Because large chemoprevention trials are rare, individual chemoprevention trials may be more likely to be utilized for secondary examination . After the PCPT results were published, the study was transitioned into a separately funded translational science resource, to allow internal and external investigators to access study samples and correlative data to examine biomarker, risk, and cancer etiology research questions . The SELECT trial was transitioned into an observational cohort, maintains a large biorepository available to researchers, and has multiple embedded ancillary protocols . In both instances, targeted funding allowed investigators to maximize the value of the data and identify new research questions to be examined. The enormous scientific impact from secondary science associated with chemoprevention trials suggests greater potential use of treatment trials as data resources. Randomized phase III treatment trials include large samples of uniformly staged and treated patients, prospective data collection on tumor characteristics and baseline clinical prognostic factors, long term follow-up, biologic sample collection and, increasingly, common quality of life instruments . The linkage of trial data to external data resources (i.e. Medicare claims) provides the opportunity to examine treatment utilization, comorbid conditions and late effects of treatment [24,25]. In both chemoprevention and treatment trial settings, the use of trial repositories and databases as resources for secondary science promises to contribute to a virtuous cycle, in which new research questions are generated, then examined in future trials, which again become new resources for secondary science. No other study in the literature has previously examined differences in scientific impact between chemoprevention and treatment trials. Our analytic approach was facilitated by the strategic linkage of multiple databases, including citation data from Google Scholar and the study and publication databases of a large NCI network group. However, the study had limitations. Treatment trials from only a single network group were used. This enabled thorough examination of the secondary science associated with trials through use of a comprehensive publication database, but also allowed for the potential for selection bias if the relationships we identified are not generalizable to other networks. Additionally, the inclusion of breast and prostate cancer chemoprevention trials only could generate more citations than chemoprevention trials for less common cancers. Citation data provide an empirical metric to assess scientific impact whose magnitude can be reliably and independently ascertained. However, the ultimate goal of clinical research is to change practice for the better, and citation counts do not explicitly represent clinical impact of the intervention. A different approach would be to examine a trial’s effect on clinical practice guidelines, though this approach could be limited, since clinical practice guidelines often do not adequately represent the available scientific evidence [, , , , ]. An approach to calculating value in clinical trials that uses quality-adjusted survival if a practice change occurred would be a useful and complementary alternative in this setting . The use of citation data is also limited by the fact that any particular citation does not represent a uniform level of impact, and citation analysis may not fully represent the scientific impact of a given trial. Also, a fair comparison of trial costs for chemoprevention versus treatment trials is complicated by the different nature of the two research settings with respect to types of patients, interventions, and funding sources. In particular, although chemoprevention trials have much larger sample sizes, the interventions, testing, and monitoring are less expensive, and toxicities and hospitalizations are limited.