Association of Perceived Benefit or Burden of Research Participation With Participants’ Withdrawal From Cancer Clinical Trials

This survey study analyzes the association of patients' perceived benefits and burdens of research participation with retention in cancer clinical trials.


Introduction
2][3] For patients with cancer, a trial may be the only option for aggressive or refractory disease.Cancer is expected to take the lives of more than 605 000 people in the US in 2022. 4The COVID-19 pandemic has heightened treatment concerns for many patients with cancer and initially placed a strain on the research enterprise. 5,6Moreover, pandemic-related delays in diagnosis and treatment increased late-stage diagnoses.Participation in cancer clinical trials (CCTs) is an option for patients with cancer, albeit a challenging one given that patients may encounter difficulties in both enrollment and remaining in the trial.Issues related to recruitment, enrollment, and retention are costly for the trial and for participants who lose potential benefits. 7Trial withdrawal, although every participant's right, can thwart study goals and hamper advancing novel treatments. 7Participant withdrawal has scientific and statistical consequences and can itself be meaningful if withdrawal suggests a lack of novel-treatment tolerance in the intervention group.
Researchers are expected to present information to interested persons to assist them in enrollment decisions.Such communication should continue throughout the trial because participants' and researchers' perceptions of benefits and burdens may change depending on participant response to experimental drugs, procedures, or tests and the course of illness.The Belmont Report has guided researchers on the ethical conduct of research for more than 40 years. 8is seminal document calls on researchers to respect human participants' ability to make autonomous decisions and protect participants from harms while equitably distributing benefits and burdens.The Belmont Report emphasizes the systematic assessment of both risks and benefits of research participation by researchers, institutional review boards (IRBs), and prospective participants.For prospective participants, the report simply states, "the assessment will assist the determination whether or not to participate." 8 Little attention has focused on the weight that CCT participants give to the benefits and burdens of research participation and whether perceived benefits are a larger factor than burdens in withdrawal decisions.This study aimed to examine the association between patients' perceived benefits and burdens of research participation and CCT retention.We were particularly interested in identifying the association of patient-reported benefit and burden scores and the difference between these scores with trial withdrawal.

Data and Sample
We used survey data from a mixed-methods study focused on CCT retention.The University of Pennsylvania IRB and Abramson Cancer Center approved the study.Written or oral informed consent was provided by each survey participant.We followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline. 9,10tients were recruited from a National Cancer Institute-designated comprehensive cancer center in an urban Northeast region.Eligible patients participated in a CCT (phase 1, 2, or 3) and/or

Outcomes
We examined 2 trial withdrawal outcomes: (1) primary outcome of actual withdrawal from the CCT defined by 2 variables: known trial dropout at time of survey consent and/or indication on the survey that withdrawal had already occurred, and (2) a composite outcome that included both actual withdrawal and reported thoughts of withdrawing.Participants were asked if they "ever thought of dropping out of their study," with a response option of yes or no.If the answer was yes to either variable (already withdrew or thought of withdrawing), the composite outcome was yes.[13][14] Primary Exposure Using a previously developed survey instrument, 1,2 participants rated items on a list of 22 benefits and 23 burdens of research participation on a 5-point Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree).Higher scores indicated greater perceived burdens and benefits.The mean benefit and mean burden scores were calculated to yield the scale score.The instrument development study 2 reported face and content validity and good internal consistency reliability scores for the benefit scale (0.90) and the burden scale (0.87).In this study, the internal consistency reliability scores of both scales were 0.91.

Statistical Analysis
Demographic data and scores were summarized overall.Mean benefit and burden scores were analyzed both as continuous and categorical variables using tertiles.The difference between the benefit and burden scores, which was calculated using mean benefit score minus mean burden score, was dichotomized to ease interpretation as follows: benefits as being greater than or equal to the burdens vs benefits as being less than the burdens.Bivariate associations between each participant characteristic (demographic and clinical) and each outcome were assessed.Characteristics with P < 0.2 were included in the adjusted logistic regression models as covariates.Logistic regression models examined the unadjusted associations between benefit score, burden score, and benefitburden difference and retention outcomes.Logistic regression was repeated, adjusting for important and number of days to survey completion).We evaluated 2 multivariable models.Model 1 included the benefit and burden scores, and model 2 assessed the difference between benefits and burdens.
We visually assessed how the probabilities of each outcome changed at different values of the 2 scores and the difference between them.
Missing data were less than 6% for variables in the analysis.Twenty participants (approximately 6% of the sample) had missing data for variables used in the adjusted models.We compared distributions of variables for participants with or without missing data and found no significant differences.Because the level of missing data was relatively small and there were no group differences, we used participants with complete data for regression.
Analyses were performed with SAS, version 9.4, (SAS Institute Inc).Significance was set at the α = .05level.Analysis of study data occurred from November 2019 through October 2022.

Characteristics
A total of 334 adult patients with cancer participated in the survey.Of these patients, 174 (52.1%) were women and 160 (47.9%) were men, with a mean (SD) age of 61.9 (11.5) years.

Perceived Benefits and Burdens
Table 2 summarizes, both as continuous and categorical variables using tertiles, the benefit and burden scores overall and within the 2 outcomes.The mean (SD) benefit item score was 3.75 (0.61), and the mean (SD) burden item score was 2.45 (0.67).Individual survey items are listed in eTables 1 and 2 in the Supplement.Most patients wanted to help others (94.2%), even though the trial might not benefit them; wanted to contribute to society (90.3%); and help their children or other family members who might acquire the disease in the future (81.5%).Most patients (86.0%) hoped for a cure, and nearly two-thirds (63.6%) reported the benefit of access to unavailable medications and treatments.Being treated respectfully or like a person was also an important benefit (86.2%).
Participants ranked their top 3 benefits: hoping for a cure (46.3%), helping future patients (32.3%), and actively treating their disease (27.5%).Worrying over receiving a placebo was the most agreed-on burden (61.3%), with some patients (29.1%) ranking it in their top 3 burdens.Other agreed-on burdens included, for example, needing to rearrange one's life (41.9%),experiencing adverse effects (41.6%), realizing the seriousness of one's cancer (35.4%), and paying for trial-related expenses (26.9%).Patients also ranked not deriving an individual benefit (37.1%) and experiencing adverse effects (24.5%), along with receiving a placebo, among their top 3 burdens.

Association Between Perceived Benefit vs Burden and Withdrawal
A higher burden score was associated with a higher actual withdrawal probability (adjusted OR, 1.86; 95% CI, 1.1-3.17;P = .02)and composite withdrawal (adjusted OR, 3.44; 95% CI, 2.09-5.67;P < .001)(Table 3).Higher benefit scores were associated with a lower probability of composite withdrawal (adjusted OR, 0.40; 95% CI, 0.24-0.66;P < .001)but did not reach significance for actual  withdrawal.Results of adjusted regression analyses are in Table 3, and the full model output is provided in eTables 3 and 4 in the Supplement.Because there were no significant interactions between benefit and burden scores, we excluded the interaction term from the models.
Table 4 shows the association between the benefit-burden difference score and the 2 withdrawal outcomes (full model output is provided in eTables 5 and 6 in the Supplement).Most participants (93.4%) perceived the benefits as being equal to or greater than the burdens.For those patients, 13.4% withdrew from trials compared with the 33.3% who reported the burdens as being greater than the benefits (adjusted OR, 3.38; 95% CI, 1. 13-10.14).The risk of withdrawal was even higher for the composite outcome (adjusted OR, 7.70; 95% CI, 2.76-21.48;P < .001).Participants who perceived the burdens as being greater were 3 times more likely to withdraw (unadjusted OR, 3.24; 95% CI, 1.23-8.51).For the composite withdrawal, there was a 43.9% difference between those who perceived the benefits as being equal to or greater than the burdens (22.8%) and those who perceived the burdens as being greater than the benefits (66.7%); the latter group was nearly 7 times more likely to report composite withdrawal (unadjusted OR, 6.78; 95% CI, 2.63-17.47).These ORs were even greater in multivariable logistic regression adjusting for sociodemographic and clinical characteristics (Table 4).The greater the difference between the benefits and burdens (ie, benefit score higher than burden score), the lower the probability of each withdrawal outcome.

Estimated Probabilities
Figure, A shows probabilities of actual withdrawal at different benefit and burden scores and difference scores.The probability was low when the benefit score was high or when it was much higher than the burden score.The probability increased as the burden score increased.Figure , B shows probabilities of the composite withdrawal at different score values, with patterns similar to those shown in Figure , A. Specifically, as the benefit-burden difference became greater (ie, when the benefit score was higher than the burden score), patients had a significantly lower probability of each outcome.

Discussion
Risk-benefit assessments are essential in human participant protection.The general assumption is that perceived burdens of research participation and exposure to experimental drugs, tests, and procedures weigh heavily in enrollment and withdrawal decisions. 1Findings of the present study confirmed this hypothesis.Those who perceived more burden than benefit were more likely to withdraw or think about withdrawing.Similarly, higher perceived benefits reduced the likelihood of A, The graphs show how the probability of withdrawal changed with the benefit score, burden score, and the difference between the benefit and burden scores.The probability of withdrawal increased with the burden score and decreased with the benefit score and the benefit-burden difference.B, The graphs show the probability of thought of withdrawing or actually withdrawing.There was a sharp decline in the probability of thought of withdrawing or actually withdrawing as the benefit score became greater than the burden score.The shading indicates the 95% CIs.
How individuals make this decision, what factors might alter the decision, and what elements are important for individuals to consider throughout the trial are not well known.

Figure .
Figure.Estimated Probabilities and 95% CIs of Each Outcome by Changes in the Benefit and Burden Scores and Changes in the Benefit-Burden Difference Probability plots from model for actual withdrawal A

JAMA Network Open | Ethics Perceived
Benefit or Burden of Research Participation and Withdrawal From Cancer Trials Patients were 18 years or older, spoke English or Spanish, and were able to provide informed consent.Patients were contacted by phone or at their clinic site using an IRB-approved script to obtain their oral or written informed JAMA Network Open.2022;5(11):e2244412. doi:10.1001/jamanetworkopen.2022.44412(Reprinted) November 30, 2022 2/13 Downloaded From: https://jamanetwork.com/ on 09/18/2023 other clinical research treatment trial and had a cancer diagnosis (gastrointestinal, genitourinary, hematologic-lymphatic, lung, or breast or gynecological cancer).

Table 1
summarizes demographic and clinical factors and their association with withdrawal outcomes.Most patients were of White race and ethnicity (86.2%) and married or partnered (77.2%), 38.9% were working, and 41.6% were retired.More than half of patients had some or completed college education (54.8%), with 31.3% reporting postgraduate work.More than a third of the sample (37.2%) were receiving public insurance and supplemental benefits through Medicaid, Medicare, or US Department of Veterans Affairs; 48.3% had private insurance.The number of participants who withdrew from the trial was 47 (14.1%), and the number who withdrew or thought of withdrawing was 83 (24.9%).Those with stage IV disease at trial enrollment (35.6%) had a higher withdrawal rate compared with those at other disease stages (stage IV, 27[23.1%];stage not available, 7 [10.8%];stages I-III, 13 [8.8%];P = .003).Those with breast or gynecological cancers had higher rates of composite withdrawal than participants with other cancer types (breast or gynecological cancer, 22 [36.1%];gastrointestinal or genitourinary cancer, 23 [19.7%]; hematologic-lymphatic cancer, 21 [20.0%];lung cancer, 17 [33.3%];P = .03).No other demographic variables were associated with withdrawal outcomes.

Table 1 .
Demographic Characteristics Overall and by Outcome (continued) b Race and ethnicity were self-reported.c

Table 2 .
Summary of Benefit and Burden Scores P values were from the χ 2 test for categorical variables or the Wilcoxon rank sum test for continuous variables.P values only took into account nonmissing data. a

Table 3 .
Adjusted Logistic Regressions Assessing the Association Between Benefit and Burden Scores and the 2 Outcomes

Table 4 .
Association Between the Benefit-Burden Difference Score and the Two Outcomes a 10 variables were adjusted: age (continuous), sex (2 categories), marital status (2 categories), employment status (3 categories), importance of spiritual-religious beliefs (3 categories), health insurance status (4 categories), cancer type (4 categories), Eastern Cooperative Oncology Group Performance Status score (2 categories), advanced disease stage (3 categories), and number of days to survey completion (2 categories).The detailed adjusted model is provided in eTables 5 and 6 in the Supplement.