• © 2003 by American Society of Clinical Oncology

Researching the Cost of Research

OPTIMIZING PATIENT recruitment to clinical trials is a topic that has received much attention in the medical literature.¹ Nonetheless, the proportion of patients enrolled in trials remains disappointingly low, compared with the proportion of patients eligible to participate.² A number of barriers to recruitment have been identified, with no singular, simple solution forthcoming.^1,2 But if incremental gains are to be made, how could our current models of clinical trials expand to include significantly more patients? There are many issues germane to the conduct of clinical trials that could be better understood through formal research and peer-reviewed communication.

The cost of conducting clinical trials is one area that has not been well researched, and most published articles have been retrospective studies focusing on the costs of clinical care.^3,4 The availability of administrative databases has facilitated these comparisons, which have led to the conclusion that treatment costs may be—at most—slightly higher for patients involved in clinical trials, compared with costs for patients treated in routine practice. These results have driven governmental policies, such as the decision to ensure Medicare coverage for the treatment costs of patients in trials in the United States. Unfortunately, most of these studies have not considered other unique aspects of clinical trials, such as patient screening, the consent process, the collection and submission of research data, and compliance with regulatory guidelines.

A more inclusive retrospective review of clinical trial–related costs was undertaken for two investigational new drug studies conducted by the National Cancer Institute of Canada Clinical Trials Group (NCIC-CTG). Evans et al⁵ found that the major cost drivers were the clinical costs of the laboratory and imaging requirements for the trials. Data management costs were relatively small contributors to total costs but still represented 21% and 13%, respectively, for these two studies. The cost of a trial with an active experimental agent was found to be higher than for an inactive agent because of the greater likelihood of receiving treatment for a longer period of time and the subsequent need for more clinical and radiologic evaluations. The fixed costs associated with the operation of the cancer center, such as heating and lighting costs, were not accounted for in the study. The models did not incorporate costs for toxicity or serious adverse event reporting. Nursing and data management support was determined from a survey administered at 27 of the NCIC-CTG collaborating centers. Large differences were found among the submitted time estimates for various support tasks. In a prospective study, Roche et al⁶ measured the time to conduct 41 identified trial tasks and also found large differences between centers for the same trials and large differences within centers for various trials.

In this issue of the Journal of Clinical Oncology, Emanuel et al⁷ focus on the nontreatment time and costs of clinical trials. Through a combination of both written and verbal surveys, time and funding estimates for 13 identified activities were solicited for the conduct of a mock clinical trial. It is not clear how many centers were approached, but the responses of 21 sites are presented. Just two of these were academic medical centers, three were managed care organizations, and 16 were oncology group practices. Participants were asked to review the sample clinical trial, first under the assumption that it was government-sponsored and then as an industry-sponsored trial. The detailed estimates of time were translated into costs, with the conclusion that the average nontreatment related cost associated with the conduct of the clinical trial was just under $2,000 per subject.

The authors acknowledge the limitations of their study. Only 21 centers participated in the exercise, and only two of these were academic medical or comprehensive cancer centers, despite “substantial efforts” to recruit additional sites. This obviously limits the generalizability of the study results. The survey resulted in estimates for a single study, and it was hypothetical in nature. In terms of estimating resource use, this is not particularly problematic as long as the trial summary presented a level of detail sufficient for estimating resource needs. However, in reality, every study budget negotiated with industry is hypothetical until the trial is locally activated. In terms of the data collection, the authors report that with previous methodology, both written and in-person surveys were administered and were subsequently validated with on-site reviews and “shadowing” of research personnel at four sites. The real limitation of using a hypothetical study in this setting is that real observations were not possible. The many tasks that were reported by the study site representatives and then subsequently observed with the selected methodology were consistently underestimated.

Despite these limitations, the results of this study remain informative and we think quite useful in expanding our limited current knowledge in the area of the total costs of clinical trials. Although presented as justification for the necessary per-case funding that typically sponsors such activity, these data reveal much more. Consistent with the other limited examples in the literature, the range in estimates of time and costs are quite large.^5,6 The submitted estimates available from the two academic medical sites highlight this issue: at one site, the time for data management and analysis is estimated at 24 hours, whereas at the other site, it is 165 hours. Not surprisingly, the range in cost estimates is of similar magnitude, from $2,068 at one center to $4,865 at the other. Examples of similarly large ranges exist within the other types of research sites. The range in total costs from all centers for the study is from $706 to $4,865, an almost seven-fold difference.

If, as the authors suggest, only 22.2% of oncologists or their practice colleagues have actually calculated the costs of participating in clinical research, perhaps the ranges in estimates are not surprising. Minor differences could be expected because of differing models of care and the distribution of responsibility for the various activities within the different centers. More likely, these larger differences are secondary to the fact that most physicians or administrators have had limited opportunity to compare their internal processes of resource estimation. This study provides feedback to the 21 participants in the study and to others who care to complete the exercise, as provided for in the paper. Still, we have neither collectively nor systematically shared our insights or experiences about the process of negotiating appropriate per-case reimbursement. In this study, a number of the large academic centers either did not or would not participate, though they were likely to have had the most experience in, and therefore the greatest opportunity for, leading the way in understanding and resolving these issues.

The conclusion that total nonclinical costs for clinical trials is approximately $2,000 per patient is largely driven by the significantly lower cost estimates provided by the oncology group practices that dominated the sample. Costs do vary from trial to trial, but in most instances, nontreatment funding of $2,000 per patient is likely to be too low. If we want to optimize patient accrual, we need to better understand the resource issues required to properly conduct clinical trials. Developing a structured framework that includes the cost implications of all appropriate research-related demands may lead to more consistently accurate budgets. Such budgets would greatly assist efforts in securing appropriate funding for both industry-sponsored trials and for the conduct of trials supported, directly or indirectly, through government funding agencies.