Despite advances in the clinical trial space, sponsors and CROs still struggle to retain participants through the end of a study. This is evident through persistent high study discontinuation rates, i.e., when a participant does not remain on a study protocol and “drops out” of a trial. Discontinuation can have devastating implications for a clinical trial: For the sponsor, participant discontinuation often results in increased budgets and delayed timelines — in efforts to replace participants, there are more recruitment costs, CRO billable hours, and site monitoring costs. In more severe cases, the impact can be existential; because some discontinued participants are not evaluable, sponsors risk losing power on endpoints which impacts a trial’s chances of getting through regulatory submission. Discontinuation can have even more severe effects on participants. Participants may have been unnecessarily exposed to a drug for which there is not yet an established safety profile, or, in some cases, participants lose access to meaningful therapies and interventions that are otherwise unavailable. As an industry that focuses on patient health and safety, discontinuation can have unforeseen risks and consequences.
While conventional wisdom and surveys seek to rationalize discontinuation rates anecdotally, we took a data-driven approach to understand key factors that increase or decrease risk of participant discontinuation using Lokavant’s proprietary data and analytics. In the face of the recent pandemic and its wide-ranging effects on the industry, we also wanted to investigate COVID-19’s potential impact on discontinuation rates.
We analyzed participant discontinuation across a sample of recent trial data to understand the effects of COVID-19 and other factors to surface the following takeaways:
- COVID-19 was not a leading predictor of risk for discontinuation. While counterintuitive, this may suggest that the perceived risk around participants missing or unable to make site visits is higher than the actual risk. We believe that historical trial data should play a bigger role in understanding and mitigating high participant discontinuation rates
- A specific sponsor's effect on discontinuation is more meaningful than that of COVID-19, which could translate to certain sponsors being able to minimize the pandemic's impact on discontinuation through tight operational management of the trial
- Different therapy areas have varying participant discontinuation risks that should be considered in clinical operations strategy
- Differences in study discontinuation across regions are statistically significant and should be considered in study feasibility vs. enrollment rates alone
To conduct this analysis, we analyzed the Participant Discontinuation Key Risk Indicator (KRI). Given the dynamism of clinical trial operations throughout this pandemic, we tracked this metric across time – before, at the onset, and further along the pandemic. We investigated the direct external effects of COVID-19, region, therapeutic area, and sponsors on risk of discontinuation.
For the underlying data, we collected information on the life cycle of a study's participants from a proprietary dataset and qualified the data across time. The data was divided into three periods with respect to COVID-19 based on the first stay-at-home order at the state (if available) or country of the study site:
- Pre-COVID: Before stay-at-home order.
- Peri-COVID: The 6 months following the order.
- New-normal: More than 6 months after the order.
We further qualified the data by mapping against sponsor and region. Our dataset included over 95 studies at over 8,600 sites with over 89,000 participants. Our data included small biotech and mid-large cap pharma sponsors and was qualified across geographical regions and major therapeutic areas. Regions were kept large, i.e., North America, South America, Eastern Europe, Western Europe, Africa, Middle East and Central Asia, and Asia Pacific, to capture global trends rather than country-level or inner-country variances.
At the outset of this analysis, we hypothesized that COVID-19 alone would have an overwhelming effect on participant discontinuation. Notwithstanding COVID, there should be differences in clinical care and operations across sponsor, therapeutic area, and region that affect rates of discontinuation. The potential effect of COVID itself was modeled as a controlled variable over time to test what industry experts likely expected to see: COVID-19 had increased discontinuation.
What determines risk of participant discontinuation?
Summary of Risk Factors
Measuring different clinical operations factors, we see the full spectrum of risk variation. Therapy areas like Rheumatology and Orthopedics have increased risks of discontinuation, in comparison to Sponsor F, whose trials have decreased risks. These results indicate a high variance in discontinuation risk dependent on these three factors: therapeutic areas, sponsors, and regions. As we explore each factor's effects on discontinuation, they can inform the optimal operational oversight of a trial.
COVID-19 was not a leading predictor of risk for discontinuation:
The effect of COVID on discontinuation was modeled as a comparison of the time to a participant's discontinuation (if it occurs at all) among different COVID periods while controlling for other variables like study characteristics and participant duration. COVID did not seem to significantly affect the risk of discontinuation by itself. Coefficient values at the start of COVID and in the "New-Normal" period decreased risk of discontinuation in our data. This also suggests that perceived COVID-19-related factors affecting the ability of a participant to attend site visits-COVID-19 restrictions or concerns, scheduling convenience, or proximity to site-may have less significance on participant discontinuation than previously assumed by industry experts. While this refuted our hypothesis, our analysis indicates that, ultimately, other non-COVID factors are more correlated to discontinuation risk.
A specific sponsor’s effect on discontinuation is more meaningful than that of COVID-19
Sponsors themselves significantly affect discontinuation risk. The rank of the sponsors' effect corresponds to that of the discontinuation proportion, with Sponsor B having the largest discontinued population. As part of our internal Lokavant analysis, we explored the underlying sponsor-specific data further and found that Sponsor B's studies had a high degree of protocol complexity in comparison to other sponsors, as it pursued multiple exploratory endpoints in its programs. This may lead to higher discontinuation due to increased participant burden (outside of confirming the primary or secondary trial endpoints) and increased site burden, which could result in process or data collection errors and subsequent re-work. These findings may also suggest a meaningful margin for improvement through technology-enhanced oversight of clinical operations. With near real-time views and granular understanding of the enrollment data, sponsors can be preemptively warned of the high discontinuation risk caused by protocol complexity. The range of sponsor effect on discontinuation shows how sponsors have more agency on their trials' presumed uncertainties.
Therapeutic areas have varying participant discontinuation risks that should be considered for overall portfolio and clinical operations strategy
It is largely understood that different therapeutic areas involve varied clinical challenges, which require different approaches to patient monitoring. With these different clinical difficulties, it should therefore not be surprising that there are varying participant discontinuation risks across therapy areas. Among the therapeutic areas examined in our analysis, Rheumatology and Orthopaedics, Neurology and Psychiatry, and Dermatology showed particularly strong effects on increased risk of participant discontinuation. Other therapy areas showed less significant effects on discontinuation, indicating that a small subset of therapy areas with established clinical difficulties determine discontinuation risks. Causes of discontinuation are different for each participant, and likely nuanced by disease.
Discontinuation can result from difficulty attending in-person visits, perceived lack of efficacy, adverse events, protocol violations, and general loss of responsiveness.
Several strategies can be employed to mitigate discontinuation risk when it poses a challenge for participants to attend in-person visits. Utilization of decentralized data capture technologies (e.g., eConsent, eCOA, ePRO, etc.), as well as decentralized tools (home health care, direct-to-patient services, etc.), can further enable participation. Utilizing such tools would not only hasten the data capture of the study but also lower the threshold of time and effort for the participant to engage in the trial. Depending on the sponsor's available resources and current strategy, increasing enrollment targets and activating more sites in more countries could be an appropriate hedge to perceived risks of discontinuation. In therapeutic areas where patient burden is high or proving efficacy is more challenging, training sites on participant retention might also be key. Site operators and investigators can then communicate effectively to participants the benefits of the trial and the minimum visit requirements. These tactical solutions, however, can only be taken once an accurate, single source of data intelligence captures where the challenges lie and what path to take.
As sponsors and CROs test different tactical solutions to clinical monitoring difficulties, data intelligence that accurately captures the specific roadblocks and informs where to deploy solutions are crucial.
Differences in study discontinuation across regions are statistically significant
Based on the analysis, we posit that the industry's understanding of where risk of discontinuation is highest may be inaccurate or biased for different geographies. Eastern Europe, for instance, is correlated with lower risk than Asia Pacific in our findings. What might explain the differences in risk between geographies? Although in an internal Lokavant analysis we have seen varying performance across a range of quality metrics in Eastern Europe, we hypothesize that discontinuation, on the other hand, is low because participants may be particularly motivated to stay in a study where there is a smaller concentration of trials in their region?
Additionally, these regions tend to have limited access to therapies on the market, signaled by much lower pharmaceutical spends per country. This may explain retention in places like Eastern Europe, South America, or Africa, where our industry tends to de-prioritize research and trials. According to the WHO, between 1999 and 2021, only around 10% of trials were held in Eastern Europe, 5% in South America, and 4% in Africa. In comparison, almost one-fourth of trials were conducted in the United States.
In regions like Asia Pacific, clinical research and clinical care are closely related, garnering a relationship of trust and compliance between patients and providers. Our partners at CMIC, the largest CRO in Japan, have noted this as well:
There is a strong sense of responsibility and respect, both from the patient who is eager to trust and follow the doctor's protocols and recommendations, and the doctor who is conscientious of their patient's needs. This forms a reliable relationship between doctor and patients which strengthens our research abilities.
After quantifying the cultural dynamics supporting low discontinuation, CMIC was able to understand what "good" vs. "great" looked like in their region. The company adopted technology such as direct data capture to reduce patient and site burden, and Lokavant's clinical trial intelligence tools to proactively identify patients and sites at risk. This allowed CMIC more effectively conduct clinical trials and manage discontinuation risks relative to others in their region.
This type of analysis should inform trial planning and feasibility decisions. Sites in regions such as Eastern Europe may prove feasible, while prior assumptions would deter research. Instead of focusing solely on enrollment rates at the regional level, clinical operators should also consider risks around discontinuation. To observe both, intelligence platforms can provide data-driven, unbiased insights and derive the risks and benefits of opening sites in one region over another.
Our investigation challenges the narrative that COVID-19 would add unprecedented challenges to trial execution due to participant discontinuation. In isolation, COVID-19 was not an important factor in increasing participant discontinuation risks.