Improving Subject Recruitment

Slow subject recruitment and poor retention are perpetual problems for the pharmaceutical industry. According to industry estimates, delays caused by recruitment affect most studies and end up costing hundreds of thousands to millions of dollars. Biopharmaceutical companies concerned with achieving their Last Patient In (LPI) targets should focus on making their subject recruitment and retention strategies, planning, and execution far more effective through a combination of technology, data assets, and process changes. Through appropriate use of technology platforms to bring together disparate information from across the organization, biopharmaceutical companies can better predict and manage recruitment and retention through data-driven frameworks.

Access to the right data allows trial sponsors to more accurately address three vital questions about subject recruitment:

• What are the likely enrollment rates for a particular clinical trial?

• How will known development issues, such as staggered study startup and holiday periods, impact my LPI milestones?

• Who are the best investigators to achieve the required recruiting objectives?

Enrollment rates

Information concerning the feasibility or difficulty of subject recruitment for specific types of subjects, indications, and sites directly impacts the accuracy of projected enrollment rates. By identifying appropriate data assets to support feasibility studies and site selection, sponsors can make the right decisions about trial staffing, budget, and timelines and positively impact subject recruitment.

Feasibility should be driven by a triangulation process involving three major elements:

• previous recruitment results from similar trials (internally generated or via literature reviews)

• internal and external experts

• investigator surveys.

A database of previous experiences that includes key information such as eligibility criteria, study synopses, and enrollment rates by country are critical to determining the relevance of a previous trial's results on the current trial. An analysis of the similarities between eligibility criteria, the drug, procedures, and standard of care can be used to modify the enrollment rate of the previous trial to approximate what can be expected in the new study.

This internal data can be augmented with published literature, particularly in areas where there is little relevant experience within the company and/or in indications where a substantial number of similar trials have been published. Most published studies provide enough data to allow for an approximate enrollment rate to be developed, but they tend to play a secondary role due to the lack of detailed information around enrollment time frames at the site level or specific eligibility criteria that is needed to develop a more refined estimate.

Data from previous trials can be used as part of a formal analysis process, conducted by internal and external experts. These experts should have a firm grasp of the standard of care in each relevant country (past, current, and medium-term future) and the clinical trials environment, and they should also have access to epidemiological data. Through the combination of these elements and the historical trial data, a reasonably tight range of most likely enrollment rates can be developed.

Finally, this process should be compared to the results of a survey of potential investigators in each relevant country. This process is typically blinded and asks relevant questions about such topics as likely enrollment rates, IRB/EC issues, and other factors that may influence the conduct of the trial. This process plays a critical role in refining the analysis described earlier and identifying potential high enrollers. A reduction factor, however, must be applied to the enrollment results from the sites to account for the lack of complete data about the trial, lack of time for site staff to carefully think through the survey, and the level of irrational exuberance that's typically contained in the responses—a 50% reduction of the enrollment estimates is often appropriate.

This entire process should be coordinated by a central group that is separate from the trial team. This allows for both process flow management efficiency and a level of dispassionate estimation that is critical to accurate planning.

Proper modeling

Accurate data is absolutely essential to planning and management, but it is only one part of the equation to improve subject recruitment performance. A sponsor must also have the appropriate data modeling and management tools in place to leverage the data. Inappropriate modeling tools will fail to account for enrollment variables that affect every trial.

Sponsors should leverage widely available technology (e.g., Excel) to develop comprehensive analytic tools that allow for realistic predictions to be made and that also enable changes in the environment or the clinical plan to be quickly incorporated into decision making. Combined with accurate historical data, models can be created that recognize important recruiting variables, such as:

• Differences in enrollment rates among countries, sites, and investigators

• Variations in time to first site live by country

• Variations in time to 20%, 40%, 60%, 80%, and 100% of sites live by country

• First "month" of enrollment that is not a full 30 days

• Impact of seasonal variations, holidays, site fatigue, and database effects on recruiting rates.

As the case study in Figure 1 shows, use of a recruitment modeling tool can lead to a much more accurate estimate of enrollment. The assumptions in the Request for Proposal (RFP) included both a straight-line extrapolation of enrollment as well as an unrealistic recruitment rate. When these assumptions were retrospectively run through a recruitment modeling tool with a more realistic recruitment rate, it was determined that half of the six month gap between the desired LPI date and actual LPI date was due to the use of inaccurate recruitment modeling. The other half was due to the use of a less than realistic recruitment rate. Developing an accurate estimate required both improved modeling and a more robust estimate of recruitment rates.

Figure 1. With the right modeling tools in place, sponsors can more accurately identify variances to recruitment goals.

Site selection

To improve subject recruitment performance, the industry must also rethink its use of investigator databases, which typically contain little more than contact information, previous trial participation, and areas of therapeutic specialty. Choosing the right sites is particularly critical given that analyses show 80% of subjects typically come from the top 20% to 30% of investigators across a range of therapeutic areas and countries. By using a more data driven approach to choosing investigators that includes previous enrollment performance (relative to peers), external data, and quality metrics, it is possible to identify high potential investigators that are much more likely to be top sites that will drive recruitment for the study.

For instance, a recent analysis of 371 Phase III studies demonstrates the importance of picking the right sites. For an average size Phase III study (10 months of enrollment and 73 sites), replacing the seven bottom-performing investigators (10%) with investigators that match the performance of the top 10% reduces overall enrollment timelines by 25%—or 2.5 months. The impact that can be generated by a relatively modest change in sites is clearly substantial.

For many sponsors, the data for this type of analysis can often reside in many different systems (e.g., Clinical Trial Management Systems, IVRS, and project manager spreadsheets). A cross functional team may be required to reach out to these various systems to develop a clean and accurate profile of investigator performance on previous studies. While the effort associated with these types of aggregations is substantial, it can generate enormous value in selecting high performers and reducing enrollment timelines.

Contingency planning

Bringing together these various planning and management tools allows sponsors to begin moving away from a reliance on rescue contingency plans that are implemented only after subject recruitment issues develop. By building proactive contingency plans with predefined triggers into core development programs, months of delay and costs can be saved by detecting signs of trouble well before they become major issues.

For contingency planning to be maximally effective, it is critical to develop early warning analytics, ideally connected to the modeling tools described. These analyses should be driven not only by recruitment results but also by variances to planned study startup timelines, screening/enrollment rates, and average speed for sites to begin enrolling their first subjects. These factors can provide early evidence of delays long before they are typically spotted using only recruitment results.

In addition to ongoing analysis of results, effective contingency planning requires implementing small but critical seeds at the outset of a study. These can include identifying/qualifying backup sites, developing explicit escalation plans for underperforming sites, and getting subject outreach materials approved by IRB/EC. By implementing these activities at the beginning of a study, time and money can be saved if enrollment results lag.

The project manager or trial leader for the study—with input from medical and clinical staff—is often best placed to develop the overall subject recruitment plan and to identify the appropriate contingency triggers. The recruitment plan should incorporate both the current situation as well as multiple levels of contingency activities based on the potential need. Resources required, lead times, and expected impact should be part of this plan to allow for thoughtful and timely implementation of contingency activities. Finally, the recruitment plan should be viewed as a "living document" that changes frequently over the life of the study as the environment changes and as more data is available (see Figure 2).

Figure 2. Contingency-based escalation is predicated upon incorporating preparatory activities into the core plan.

Contingency planning: case study

A recent case study of a multiprotocol diabetes program illustrates the contingency management concepts described earlier. To maximize the chances of achieving LPI milestones, the initial steps of a range of recruitment support tactics were implemented, including IRB/EC approval of subject outreach materials, identifying back-up sites, developing/communicating site-level escalation plans for nonperformance, and building site support plans (such as provisioning rapid screening kits and subject travel stipends). These initial activities were inexpensive and were done in parallel with protocol development and study startup activities.

Using the scenario planning capabilities of a recruitment modeling tool, it was determined that regulatory and other delays during the start-up period were generating unacceptable risk to the LPI milestones. Given the value to the sponsor of achieving these timelines, a decision was made to fully implement the contingency activities that had been previously put in place. Because of these early preparations, it was possible to launch backup sites, nonperformer escalation plans, and subject outreach in a fraction of the time it would have taken—leading to a rapid improvement to the recruitment situation.

Overall, with the right data and tools, sponsors can more accurately predict, monitor, and manage subject recruitment, while avoiding expensive and risky trial delays. The result of investments in better data assets, recruitment modeling tools, and feasibility/management processes can be leveraged to reduce costs and bring new products to market faster—improvements that are invaluable in today's highly competitive pharmaceutical marketplace.

Joshua Schultz, is vice president, patient recruitment, clinical research services, Parexel International, 200 West Street, Waltham, MA 02451, email: Joshua.Schultz@PAREXEL.com