Challenges may result from issues related to the protocol itself and impact recruitment and implementation. A good contingency plan will articulate as many possible situations that might arise and be as clear as possible about how one would deal with them. Plans can be developed for recruitment challenges, implementation strategies that are not achieving the intended goals (lack of fidelity and resulting lack of exposure to the intervention), and interventions that do not achieve their intended effect.

Recruitment challenges

Recruiting enough participants to achieve the study power is a common challenge among clinical trials. Thoughtful pilot work in ePCTs can help prevent unrealistic recruitment expectations. However, there are other recruitment challenges besides enrolling a sufficient number of participants. For example, HiLo originally used cluster-randomization combined with post-randomization consent of patients with end-stage renal disease. Clusters were assigned to either standard-of-care phosphate targets (Low) or less strict phosphate targets (High). At 10% enrollment, investigators noticed an imbalance between the 2 arms in willingness to consent to participation, causing concerns about biased enrollment. Although there were already more than 500 patients involved in the trial, the researchers decided to pivot from a cluster-randomized trial to individual-level randomization, which made enrollment in the arms equal. This ability to pivot enabled the trial to continue to try to answer the important question about phosphate targets in patients with end-stage renal disease.

Implementation strategies that are not achieving the intended goals (lack of fidelity and resulting lack of exposure to the intervention)

The way investigators understand if their intervention is being implemented is through careful monitoring and repeated and rigorous checks to ensure fidelity. A priori, we recommend that investigators identify the core intervention components that will be monitored and develop an action plan should implementation of components be inadequate. Remedies may involve more interaction with partners, including site visits and/or phone calls with study champions.

These 3 examples from the NIH Pragmatic Trials Collaboratory trials highlight the importance of monitoring both the intervention delivery and the data collection process.

• The Pragmatic Trial of Video Education in Nursing Homes (PROVEN) trial was designed to test a video-assisted decision support intervention in advance care planning for patients in nursing homes. Staff at nursing homes were to offer the videos to the participant and subsequently document that the video was offered in the EHR (DeBar et al. 2022). Monitoring reports revealed considerable variation in offer rates across the different nursing home facilities, and much effort was put into contacting the sites and training the personnel on the use of the videos.
• The EMBED researchers found that coding practice related to patient race and ethnicity varied across and within systems, which was a challenge to reconcile. They recommend monthly data pulls during the trial, rather than one single data pull at the end.
• The ACP PEACE investigators planned to use structured data elements from the EHR to examine documentation of an advance care plan (Volandes et al. 2025). However, a chart review of data elements during the planning phase revealed that these were inaccurate—fewer than 60% had the correct documents. ACP PEACE had the advantage of having a laboratory that does natural language processing, which made it possible to extract the same data with far more accuracy.

Interventions that do not achieve their intended effect

A typical ePCT can last for 5 to 7 years, yet through monitoring and DSMBs, trialists may know earlier that their intervention is not achieving its intended effect. In this situation, instead of considering the endeavor a failure, investigators can use this information to pivot and still gather information relevant to solving the problem at hand.

Some questions to consider when designing a trial to prepare for potential signals that the intervention is not having the intended effect:

• Is this an appropriate question to be addressed with a PCT?
• Is this question being framed in the right way?
• Is this the right setting to answer this question?
• Are these the right people to conduct the trial? Are these people qualified?
• Is this the best solution or best thing to do based on what we know right now?
• What would be the backup plan? If this intervention doesn't work, is there something else that we might test?

For more on unanticipated challenges and how they might impact the analysis plan, see the Living Textbook Section on Unanticipated Changes.