Advancing trial recruitment science through enhanced study design

Although clinical trials are key to advancing cancer care, enrolling participants has been a critical barrier to their successful completion.¹ Low enrollment is the most common reason for trial failure, prolonging trial duration and ballooning costs.¹ Despite these substantial issues, there are no reliable, evidence-based strategies to improve trial enrollment.^{2, 3} This critical recruitment science knowledge gap can be attributed to two key issues: trialists rarely describe what they do to improve enrollment, and described enrollment strategies are seldom evaluated with the scientific rigor we expect of other health care interventions.^{2, 3}

With this in mind, we read with great interest the recent article in Cancer by Nathan et al., wherein the authors describe enrollment interventions in the CISTO trial (ClinicalTrials.gov identifier NCT03933826).⁴ Their work emphasizes an association between the rollout of a centralized communication modality (CISTOquestion) and increased screening and enrollment. This type of study helps shrink the aforementioned trial enrollment knowledge gap through detailed description and retrospective analysis of an enrollment intervention, which potentially could be applied in other multicenter trials. Transparent sharing of enrollment interventions is an important first step in developing evidence-based strategies to help our trials run better. Trialists who use enrollment interventions would be well served to follow this example in reporting.

However, the crucial question of whether this intervention caused improved enrollment (rather than merely being associated with enrollment) remains methodologically unanswered. In particular, pre-post retrospective analyses face confounding from external circumstances (coronavirus disease 2019 restrictions and their relaxation), secular trends in trial structure (increasing trial sites over time), and bundled interventions. Indeed, a recent randomized trial of an enrollment intervention highlights the importance of running prospective trials to account for these confounders.⁵ The importance of the causal question is that decisions must be made about whether an intervention such as CISTOquestion is effective enough to be applied at scale or to be recommended by funders or professional societies. Causality and effect size estimates alike are important aspects of this equation.

These types of questions could be answered more rigorously in the future with a manageable amount of preplanning. One strategy is to control the timing of intervention rollout to different sites in a stepped wedge trial.⁶ For example, in this trial with 38 centers, a new group of three centers could receive access to CISTOquestion every month for 12 months, with the order of these groups randomly selected. Akin to the current study, every center would eventually get access, but randomization and stepping-out would allow for comparison with control while adjusting for other confounders (e.g., underlying secular trends, other enrollment interventions). This design also facilitates logistical troubleshooting; intervention adaptation may even be acceptable in some studies. Alternatively, a traditionally randomized trial could allocate one half of the sites to an enrollment intervention for a brief period (e.g., 6 months); and, if it effectively improved trial enrollment, it could then be distributed to all sites. Digital interventions such as CISTOquestion are particularly amenable to this type of prospective evaluation.

Understanding how well enrollment interventions work is critical because trials are run under resource constraints in which cost–benefit tradeoffs must be considered. If something is proven unlikely to be helpful, money and time could be redirected elsewhere. For example, an email eligibility system could be substituted with enrollment reminders or audit and feedback to physicians. With more resource-intensive interventions, like hiring additional trial staff, empirical knowledge of the effect size on enrollment can powerfully inform cost effectiveness and trial budgets.

A countervailing force in these considerations is achieving buy-in to build these types of analyses into the rollout of trial initiatives. It is extremely tempting to avoid the additional effort of measurement and instead look to rapid intervention rollout. Developing streamlined methods to embed trials in existing initiatives could lower the barriers to evaluating these interventions while minimally impeding the effectiveness and pace of their rollout.⁷ Efforts to develop pragmatic studies and toolkits for these types of evaluations are ongoing.

We can design better, more efficient trials by applying randomized controlled trial principles to enrollment interventions. A critical first step is taken by Nathan et al. in describing enrollment interventions, and other study teams should be encouraged to describe their own approaches to enrollment. Going forward, as these types of interventions are planned, we encourage their deliberate design and rollout so that the knowledge generated can be more rigorous and generalizable. With minor tweaks to existing practice, we can build out a more robust and applicable evidence base for how best to design and run our clinical trials.

The authors disclosed no conflicts of interest.