Abstract
The aim was to assess whether subject’s participation in early-phase chemoprevention trials was satisfactory and to identify features associated with subjects’ satisfaction. Thirteen trials that investigated a range of candidate agents from 2006–2021 by the Cancer Prevention Network were included. The 5-item “Was It Worth It?” (WIWI) questionnaire was administered to all subjects at the end of each trial’s intervention or at early termination. Satisfied overall was defined as a participant response of “yes” to the first three questions. Six hundred ninety-one participants from the United States, Canada, Puerto Rico, and Honduras enrolled on a trial. Six hundred fifty-two (94.4%) completed the WIWI. Of these, 493 (75.6%) were White, non-Hispanic/Latino; 39 (6.0%) Black, non-Hispanic/Latino; 98 (15.0%) Hispanic/Latino; and 8 (1.2%) of another race/ethnicity. One hundred ninety-three were women (29.6%), 121 (17.5%) were ≥65 years, and 517 (79.3%) participated in a placebo-controlled trial. Eighty-five percent indicated being satisfied overall. Compared with White, non-Hispanic/Latino, the odds of not satisfied overall were 2.96 times higher for Black/Asian/>1race, non-Hispanic/Latino (P<0.001) and 0.40 times lower for Hispanic/Latino (P=0.004). The odds of not satisfied overall was 1.9 times higher when the number of preintervention adverse events (AEs) experienced was ≥1 (P=0.012); 1.8 times higher when the percentage of the intervention duration with AEs was >5% (P=0.024); and 7.4 times higher for subjects who terminated the intervention early (P<0.001). These findings can inform the design of future chemoprevention trials and help investigators improve accrual, retention, adherence, and diversity in this underexplored research setting.
Keywords: Cancer prevention, Chemoprevention, Clinical trials, Questionnaires, Participant experience
Prevention relevance:
The 5-item “Was It Worth It?” (WIWI) questionnaire, which captures the participant-reported experience of trial participation gives the subject a voice in the development of new chemopreventative agents. This study in 652 subjects looked at satisfaction with participation in early-phase chemoprevention trials for higher-risk, cancer-free men and women.
INTRODUCTION
The global incidence of cancer, with its associated morbidity, continues to increase1. Coupled with the accelerated rise of cancer health care costs, cancer chemoprevention is an attractive strategy2. To determine the chemoprevention potential for a growing number of promising new agents, chemoprevention trials are an integral part of cancer control.
For higher-risk, cancer-free participants, participating in a chemoprevention trial is a personal choice and an individual experience. Several barriers to recruitment and retention of higher-risk men and women to chemoprevention trials have been documented3–7. Subject satisfaction in chemoprevention trials has not been sought but is key to the success of these trials. Additionally, any number of demographic and clinical characteristics of the participants as well as trial design features such as trial duration, incorporation of a placebo arm, use of invasive medical procedures, and onerous data collection may be associated with participant satisfaction. Knowing the features associated with participant satisfaction can inform the design of future chemoprevention trials, improve accrual, retention, and adherence and enhance participant diversity.
The 5-item “Was It Worth It?” (WIWI) questionnaire8, 9 was developed to evaluate patients’ perceptions of clinical trial participation and has been used in nationally-conduced cancer treatment trials10, 11. Because chemoprevention trials often recruit healthy individuals whose altruism appears to be a key motivator for their participation, these individuals’ responses to the WIWI questionnaire might be key to sustaining the success of such trials. We performed a pooled analysis of WIWI data from chemoprevention trials conducted by the NCI-funded Cancer Prevention Network (CPN). The aim was to assess participant satisfaction in chemoprevention trials and to identify both subjects’ and trial features associated with greater participant satisfaction.
MATERIALS AND METHODS
Trial and Participant Selection
Participant-level data from 13 early-phase (0, I, II) chemoprevention trials that targeted four disease sites (colorectum; esophagus; liver; lung) and that investigated a range of candidate agents were included. These trials had been conducted from 2006–2021. A synopsis of each trial is provided (Supplementary Table S1) and the primary results from each trial have been previously published12–24. The current pooled analysis was subsumed under the IRB approvals of all these earlier-approved trials and included de-identified datasets.
Primary Outcome – Satisfied Overall
The five-item “Was It Worth It” (WIWI) questionnaire (Box 1) was used to assess whether participation in chemoprevention trials was satisfactory for higher-risk, cancer-free participants, who had met the eligibility criteria for each trial. The WIWI questionnaire was administered at the end of each trial’s intervention or at early termination. Furthermore, the WIWI questionnaire was completed prior to any unblinding of randomization assignment in the double-blinded, placebo-controlled, chemoprevention trials. Participants did not have knowledge of the trial results at the time of WIWI questionnaire completion.
Box 1. Five-item “Was It Worth It” Questionnaire Systematically Administered at the End of the Intervention Period in each Early-phase (0, I, II) Chemoprevention Trial Conducted by the Cancer Prevention Network.
| Participating in a clinical trial / research study is a personal choice and an individual experience. We would like to get your feedback on your experience in this research study. Please respond to the following questions as indicated. | ||||
| Directions: Please answer each question by circling Y (for yes), N (for no), or U (for uncertain). | ||||
| Was it worthwhile for you to participate in this research study? | Y | N | U | |
| If you had to do it over, would you participate in this research study again? | Y | N | U | |
| Would you recommend participating in this research study to others? | Y | N | U | |
| Directions: Circle one response | ||||
| Overall, did your quality of life change by participating in this research study? | ||||
| It improved | It stayed the same | It got worse | ||
| Overall, how was your experience of participating in this research study? | ||||
| Better than I expected | The same as I expected | Worse than I expected | ||
The binary outcome, satisfied overall, was defined as a participant response of “yes“ to the first three questions on the WIWI questionnaire: Q1=“Was it worthwhile for you to participate in this research study;” Q2=“If you had to do it over, would you participate in this research study again;” and Q3=”Would you recommend participating in this research study to others.” Participants who answered no or uncertain to any of these three questions were categorized as not satisfied overall.
Secondary Outcomes
Secondary outcomes were responses to: Q4=”Overall, did your quality of life change by participating in this research study” and Q5=”Overall, how was your experience of participating in this research study.” The three-level response categories for Q4 comprised “It improved”, “It stayed the same”, and “It got worse”, while for Q5 it consisted of “Better than I expected”, “The same as I expected”, and “Worse than I expected.”
Biostatistical Analysis
Participant baseline, trial-design, and early-termination features were summarized and compared between the participants who did and did not complete the WIWI questionnaire using the Fisher’s exact test for categorical variables and the Wilcoxon rank-sum test for continuous variables. The same statistical tests were used to assess the association between each participant baseline, trial-design, and on-study experience feature and being satisfied overall in the participants who completed Q1, Q2, and Q3 on the WIWI questionnaire.
The Kappa statistic25 and 95% confidence interval (CI) were used to quantify the pairwise agreement in response categories (yes, no, uncertain) between the first three WIWI questions Q1, Q2, and Q3 (Supplementary Table S2). The remaining two WIWI questions Q4 and Q5 were summarized descriptively and according to trial, age, sex, race/ethnicity, and whether the trial site was within the United States (US) or international.
A diverse set of 17 participant baseline, trial-design, and on-study experience features were identified based on subject matter knowledge (Figure 1), and were interrogated with the random forests algorithm, a machine learning method26, to triage the features. A hierarchy of features was constructed based on quantification of the importance of their effects on being satisfied overall. Features were ranked, from most important to least important, based on the mean decrease in accuracy, a permutation-based variable importance approach, and plotted. Random forests algorithm contains internal cross-validation; consequently, the method avoids the need for a test sample or an external validation data set, which is unavailable in this case. The randomForest R package was used to implement random forests27.
Figure 1.
Participant flow diagram including a list of the prespecified trial-design, on-study, and participant baseline features.
Random forests algorithm was initially applied to participants who had complete data on all 17 features. Because masking effects can appear in the presence of highly associated or correlated features, potentially diluting the importance of key features28, 29, a filtered set of 12 features were identified (Figure 1) based on a supervised approach to removing a feature that exhibited a large pairwise association (or correlation) with another feature. Cramer’s V (or Phi coefficient for 2-by-2 contingency tables)30 was used to identify large pairwise associations between categorical features and Spearman’s rank correlation coefficient31 was used to identify large pairwise correlations between continuous features (Supplementary Table S3: A–E). Given a large pairwise association or correlation, subject matter expertise was used to keep the feature that we considered potentially actionable to improving participant satisfaction and less tethered to the trial’s research question. Random forests algorithm was then applied three more times: (1) complete case data on the filtered set of 12 features; (2) complete case data on the filtered set excluding the single feature, smoking status, which exhibited a missing rate of 20%; and (3) all available data on the filtered set (excluding smoking status) with missing feature values imputed using the rfImpute function of the randomForest R package. To achieve a stabilization effect of variable importance with each application of the random forests algorithm, the optimal number of variables randomly selected at each node was used and the number of trees in the forest was set high at 10,000.
A multiple logistic regression model was run based on the intersection of the topmost important features identified across the four applications of random forests algorithm and with those features identified as being associated with participant satisfaction status in the univariate analysis. This model was used to facilitate the interpretation of the relationships of the features with the probability of not satisfied overall, adjusted for potential confounding variables of age, sex, and duration of intervention. Because there were <100 events in our sample, to mitigate the risk of overfitting, which can lead to problems of bias in the regression coefficients and under/overestimation of variances of the estimators, at least 10 events per factor/feature were ensured32. Adjusted odds ratios and 95% CIs and two-sided Wald-based P values are reported. No adjustment for multiple testing was made.
Data availability statement
Data and materials are available in the NCI’s Cancer Data Acquisition System and may be requested using the standardized process (https://cdas.cancer.gov/).
RESULTS
Characteristics based on Trial Initiation
A total of 691 (of 706) participants started the trial-specific intervention with 652 (94.4%) completing the WIWI questionnaire and 39 not (Figure 1). A higher percentage of the participants who did not complete the WIWI questionnaire indicated that they currently smoked (38.5% vs 21.2%; P=0.038); the distribution of the remaining baseline and trial-design features was similar (all P>0.05; Table 1).
Table 1.
Baseline participant, trial design, and early termination features presented overall (N=691) and by completion of the WIWI questionnaire (completed [N=652] vs not completed [N=39]).
| Participant Baseline Features | Total (N=691) |
WIWI Questionnaire | ||
|---|---|---|---|---|
| Completed (N=652) |
Not Completed (N=39) |
P | ||
| Age, in years | ||||
| n | 691 | 652 | 39 | 0.672 |
| Median [Range] | 48 [18, 81] | 48 [18, 81] | 51 [21, 77] | |
| Mean (Standard Deviation) | 52 (11.0) | 51 (11.0) | 53 (12.2) | |
| Age, n (%) | ||||
| <65 years old | 570 (82.5%) | 537 (82.4%) | 33 (84.6%) | 0.831 |
| ≥65 years old | 121 (17.5%) | 115 (17.6%) | 6 (15.4%) | |
| Sex, n (%) | ||||
| Female | 206 (29.8%) | 193 (29.6%) | 13 (33.3%) | 0.594 |
| Male | 485 (70.2%) | 459 (70.4%) | 26 (66.7%) | |
| Ethnicity, n (%) | ||||
| Not Hispanic or Latino | 574 (83.1%) | 540 (82.8%) | 34 (87.2%) | 0.494 |
| Hispanic or Latino | 110 (15.9%) | 106 (16.3%) | 4 (10.3%) | |
| Not reported/Unknown | 7 (1.0%) | 6 (0.9%) | 1 (2.6%) | |
| Race, n (%) | ||||
| White | 598 (86.5%) | 565 (86.7%) | 33 (84.6%) | 0.137 |
| Black or African American | 51 (7.4%) | 50 (7.7%) | 1 (2.6%) | |
| Asian | 9 (1.3%) | 7 (1.1%) | 2 (5.1%) | |
| >1 Race | 23 (3.3%) | 22 (3.4%) | 1 (2.6%) | |
| Not Reported/Unknown | 10 (1.4%) | 8 (1.2%) | 2 (5.1%) | |
| Race/Ethnicity, n (%) | ||||
| White, Not Hispanic or Latino | 524 (75.8%) | 493 (75.6%) | 31 (79.5%) | 0.131 |
| Black or AA, Not Hispanic or Latino | 40 (5.8%) | 39 (6.0%) | 1 (2.6%) | |
| Hispanic or Latino | 101 (14.6%) | 98 (15.0%) | 3 (7.7%) | |
| Other | 10 (1.5%) | 8 (1.2%) | 2 (5.1%) | |
| Not Reported/Unknown | 16 (2.3%) | 14 (2.2%) | 2 (5.1%) | |
| Two-level Race/Ethnicity, n (%) | ||||
| White, Not Hispanic or Latino | 524 (75.8%) | 493 (75.6%) | 31 (79.5%) | 0.422 |
| Other | 151 (21.9%) | 145 (22.2%) | 6 (15.4%) | |
| Not Reported/Unknown | 16 (2.3%) | 14 (2.2%) | 2 (5.1%) | |
| Smoking History, n (%) | ||||
| Current Smoker | 153 (22.1%) | 138 (21.2%) | 15 (38.5%) | 0.038 |
| Former Smoker | 210 (30.4%) | 201 (30.8%) | 9 (23.1%) | |
| Never Smoked | 190 (27.5%) | 183 (28.1%) | 7 (18.0%) | |
| Not Reported | 138 (20.0%) | 130 (19.9%) | 8 (20.5%) | |
| Number of Comorbidities, n (%) | ||||
| None | 57 (8.3%) | 53 (8.1%) | 4 (10.3%) | 0.603 |
| 1–3 | 210 (30.4%) | 210 (30.8%) | 9 (23.1%) | |
| 4–9 | 242 (35.0%) | 225 (34.5%) | 17 (43.6%) | |
| 10+ | 158 (22.9%) | 149 (22.9%) | 9 (23.1%) | |
| Not Reported | 24 (3.5%) | 24 (3.7%) | 0 (0.0%) | |
| Trial Design Features | ||||
| Disease Site, n (%) | ||||
| Lung | 199 (28.8%) | 184 (28.2%) | 15 (38.5%) | 0.084 |
| Esophageal | 196 (28.4%) | 190 (29.1%) | 6 (15.4%) | |
| Colorectum | 250 (36.2%) | 237 (36.4%) | 13 (33.3%) | |
| Liver | 46 (6.7%) | 41 (6.3%) | 5 (12.8%) | |
| Placebo-controlled Trial, n (%) | ||||
| Yes | 545 (78.9%) | 517 (79.3%) | 28 (71.8%) | 0.311 |
| No | 146 (21.1%) | 135 (20.7%) | 11 (28.2%) | |
| Agent Class, n (%) | ||||
| Injection | 175 (25.3%) | 163 (25.0%) | 12 (30.8%) | 0.449 |
| Oral | 516 (74.7%) | 489 (75.0%) | 27 (69.2%) | |
| U.S.A. Site, n (%) | ||||
| Yes | 467 (67.6%) | 440 (67.5%) | 27 (69.2%) | 0.999 |
| No | 224 (32.4%) | 212 (32.5%) | 12 (30.8%) | |
| Distance to Clinic (Miles) * | ||||
| n | 609 | 573 | 36 | 0.625 |
| Median [Range] | 16.5 [0.004, 1784.7] | 16.5 [0.004, 1784.7] | 18.1 [0.004, 564.5] | |
| Mean (Standard Deviation) | 56.7 (138.2) | 56.0 (139.0) | 67.4 (125.6) | |
| Early Termination Features | ||||
| Completed Trial, n (%) | ||||
| Yes | 644 (93.2%) | 635 (97.4%) | 9 (23.1%) | <0.001 |
| No – Early Termination | 47 (6.8%) | 17 (2.6%) | 30 (76.9%) | |
| Reason | ||||
| Adverse Event | 19 / 47 (40.4%) | 10 / 17 (58.8%) | 9 / 30 (30%) | 0.069 |
| Participant Withdrawal | 12 / 47 (25.5%) | 4 / 17 (23.5%) | 8 / 30 (26.7%) | 0.999 |
| Lost to Follow-up | 8 / 47 (17.0%) | 1 / 17 (5.9%) | 7 / 30 (23.3%) | 0.228 |
| Physician Decision | 3 / 47 (6.4%) | 1 / 17 (5.9%) | 2 / 30 (6.7%) | 0.999 |
| Other | 5 / 47 (10.6%) | 1 / 17 (5.9%) | 4 / 30 (13.3%) | 0.640 |
Note. P values are two-sided and based on the Fisher’s exact test for categorical features and the Wilcoxon rank-sum test for continuous features. Not reported or unknown were excluded from statistical tests.
A participant-level five-digit zip code was needed to approximate the distance from the centroid of the zip code to the clinic. Distance could not be calculated for N=60 participants in Canada, Puerto Rico, and USA, and none of the N=22 participants in Honduras.
After initiating the intervention, 47 of 691 (6.8%) participants terminated the intervention early, with a higher early termination percentage observed in the group of participants who did not complete the WIWI questionnaire (30/39 [76.9%] vs 17/652 [2.6%]; P<0.001; Table 1). While the reasons for early termination were not significantly different between the group of participants who did and did not complete the WIWI questionnaire, early termination due to adverse events (AEs) occurred in 10/17 (58.8%) in the group who completed the WIWI questionnaire vs 9/30 (30%) in the group who did not; a higher percentage in the group who did not complete the WIWI questionnaire were lost to follow-up (1/17 [23.3%] vs 7/30 [5.9%]) and early termination due to participant withdrawal was similar between the groups (4/17 [23.5%] vs 8/30 [26.7%]).
Characteristics based on WIWI Completion
The 652 participants who completed the WIWI questionnaire were enrolled across geographically diverse participating sites, with representation in large and small communities throughout the US, Canada, Puerto Rico, and Honduras (Supplementary Figure S1). Of these 652 participants, 493 (75.6%) were White, non-Hispanic or Latino; 39 (6.0%) Black, non-Hispanic or Latino; 98 (15.0%) Hispanic or Latino; and 8 (1.2%) of another race/ethnicity. There were 193 women (29.6%), 121 (17.5%) were ≥65 years old, and 517 (79.3%) participated in a placebo-controlled trial. One-third were enrolled outside the US (Table 1).
Overall Satisfaction
Five hundred fifty-seven (85.4%) subjects indicated being satisfied overall. Table 2 shows the association between each baseline, trial-design, and on-study feature with being satisfied overall. There was a significant association between race/ethnicity and overall satisfaction (P<0.001). The proportion of Hispanic or Latinos who were satisfied overall was 94.9%, while the proportion of White, non-Hispanic or Latino and Black, non-Hispanic or Latino was 85.0% and 74.4%, respectively. The participants who were not satisfied overall with their participation reported a higher cumulative number of preintervention AEs (P=0.001), a higher cumulative number of intervention AEs (P=0.018), and a higher percentage of the intervention duration with AEs (P=0.005). Furthermore, a higher percentage of participants who terminated early reported not satisfied overall compared with participants who did not terminate early (52.9% vs 13.5%; P<0.001).
Table 2.
Satisfied overall (yes vs no) by participant baseline features, trial design features, and on-study experience.
| Participant Baseline Features | Satisfied Overall | |||
|---|---|---|---|---|
| Yes | No | TOTAL | P | |
| Age, in years | ||||
| n | 557 | 95 | 652 | 0.685 |
| Median [Range] | 48 [18, 81] | 48 [27, 78] | ||
| Mean (Standard Deviation) | 52 (11.1) | 51 (10.5) | ||
| Age, n (%) | ||||
| <65 years old | 456 (84.9%) | 81 (15.1%) | 537 | 0.470 |
| ≥65 years old | 101 (87.8%) | 14 (12.2%) | 115 | |
| TOTAL | 557 | 95 | 652 | |
| Sex, n (%) | ||||
| Female | 167 (86.5%) | 26 (13.5%) | 193 | 0.715 |
| Male | 390 (85.0%) | 69 (15.0%) | 459 | |
| TOTAL | 557 | 95 | 652 | |
| Race/Ethnicity, n (%) | ||||
| White, Not Hispanic or Latino | 419 (85.0%) | 74 (15.0%) | 493 | <0.001 |
| Black or AA, Not Hispanic or Latino | 29 (74.4%) | 10 (25.6%) | 39 | |
| Hispanic or Latino | 93 (94.9%) | 5 (5.1%) | 98 | |
| Other | 4 (50.0%) | 4 (50.0%) | 8 | |
| TOTAL | 545 | 93 | 638 | |
| Two-level Race/Ethnicity, n (%) | ||||
| White, Not Hispanic or Latino | 419 (85.0%) | 74 (15.0%) | 493 | 0.688 |
| Other Race/Ethnicity | 126 (86.9%) | 19 (13.1%) | 145 | |
| TOTAL | 545 | 93 | 638 | |
| Smoking History, n (%) | ||||
| Current Smoker | 115 (83.3%) | 23 (16.7%) | 138 | 0.579 |
| Former Smoker | 175 (87.1%) | 26 (12.9%) | 201 | |
| Never Smoked | 159 (86.9%) | 24 (13.1%) | 183 | |
| TOTAL | 449 | 73 | 522 | |
| Number of Comorbidities, n (%) | ||||
| None | 47 (88.7%) | 6 (11.3%) | 53 | 0.870 |
| 1–3 | 173 (86.1%) | 28 (13.9%) | 201 | |
| 4–9 | 193 (85.8%) | 32 (14.2%) | 225 | |
| 10+ | 125 (83.9%) | 24 (16.1%) | 149 | |
| TOTAL | 538 | 90 | 628 | |
| Trial Design Features | ||||
| Disease Site, n (%) | ||||
| Lung | 151 (82.1%) | 33 (17.9%) | 184 | 0.112 |
| Esophageal | 171 (90.0%) | 19 (10.0%) | 190 | |
| Colorectum | 202 (85.2%) | 35 (14.8%) | 237 | |
| Liver | 33 (80.5%) | 8 (19.5%) | 41 | |
| TOTAL | 557 | 95 | 652 | |
| Placebo-controlled Trial, n (%) | ||||
| Yes | 443 (85.7%) | 74 (14.3%) | 517 | 0.684 |
| No | 114 (84.4%) | 21 (15.5%) | 135 | |
| TOTAL | 557 | 95 | 652 | |
| Agent Class, n (%) | ||||
| Injection | 139 (85.3%) | 24 (14.7%) | 163 | 0.999 |
| Oral | 418 (85.5%) | 71 (14.5%) | 489 | |
| TOTAL | 557 | 95 | 652 | |
| U.S.A. Site, n (%) | ||||
| Yes | 379 (86.1%) | 61 (13.9%) | 440 | 0.478 |
| No | 178 (84.0%) | 34 (16.0%) | 212 | |
| TOTAL | 557 | 95 | 652 | |
| Distance to Clinic (Miles) * | ||||
| n | 489 | 84 | 573 | 0.477 |
| Median [Range] | 16.9 [0.004, 1784.7] | 16.0 [0.004, 405.9] | ||
| Mean (Standard Deviation) | 57.4 (147.2) | 47.7 (76.0) | ||
| On-study Experience | ||||
| Completed Trial, n (%) | ||||
| Yes | 549 (86.5%) | 86 (13.5%) | 635 | <0.001 |
| No – Early Termination | 8 (47.1%) | 9 (52.9%) | 17 | |
| TOTAL | 557 | 95 | 652 | |
| Duration of Intervention, Days | ||||
| n | 557 | 95 | 652 | 0.473 |
| Median [Range] | 87 [1, 579] | 107 [2, 420] | ||
| Mean (Standard Deviation) | 145.4 (115.3) | 140.3 (118.6) | ||
| % of Intervention Duration Experiencing Adverse Events | ||||
| n | 555 | 95 | 650 | 0.005 |
| Median [Range] | 6.9 [0, 100] | 17.3 [0, 100] | ||
| Mean (Standard Deviation) | 26.9 (34.6) | 34.5 (37.2) | ||
| Cumulative Number of Intervention Adverse Events | ||||
| n | 557 | 95 | 652 | 0.018 |
| Median [Range] | 1 [0, 76] | 2 [0, 33] | ||
| Mean (Standard Deviation) | 2.7 (5.0) | 3.4 (4.8) | ||
| Cumulative Number of Preintervention Adverse Events | ||||
| n | 557 | 95 | 652 | 0.001 |
| Median [Range] | 0 [0, 15] | 1 [0, 14] | ||
| Mean (Standard Deviation) | 1.3 (2.3) | 2.2 (3.0) | ||
| Experienced ≥ Grade 3 Adverse Event | ||||
| Yes | 30 (79.0%) | 8 (21.1%) | 38 | 0.238 |
| No | 527 (85.8%) | 87 (14.2%) | 614 | |
| TOTAL | 557 | 95 | 652 | |
Note. P values are two-sided and based on the Fisher’s exact test for categorical features and the Wilcoxon rank-sum test for continuous features. Not reported or unknown were excluded from statistical tests.
A participant-level five-digit zip code was needed to approximate the distance from the centroid of the zip code to the clinic. Distance could not be calculated for N=58 participants in Canada, Puerto Rico, and USA, and none of the N=21 participants in Honduras.
The top four ranked features associated with participant satisfaction status were race/ethnicity, percentage of the intervention duration experiencing AEs, early termination and either the cumulative number of preintervention AEs or distance to the clinic (in miles). Distance to the clinic (in miles) was derived from the participant zip code and missing in ~12% of participants; furthermore, it was only selected in two of the four analysis sets and, therefore, was not explored in the multiple logistic regression models. The two least important ranked features were consistently sex at birth and the number of comorbidities (Supplementary Figure S2: A, B, C, D).
Table 3 presents the results from the multiple logistic regression adjusting for age, sex, and participant-experienced intervention duration. Compared with White, non-Hispanic or Latino, the odds of not satisfied overall were 2.96 times higher for Black/Asian/>1 race, non-Hispanic or Latino (P<0.001) and 0.40 times lower for Hispanic or Latino (P=0.004). The odds of not satisfied overall was 1.9 times higher when the cumulative number of preintervention AEs experienced was ≥1 (P=0.012); 1.8 times higher when the percentage of the intervention duration with AEs was >5% (P=0.024); and 7.4 times higher for participants who terminated the intervention early (P<0.001).
Table 3.
Final model: Multiple logistic regression adjusted for age (<65; ≥65 years), sex (Female; Male), and participant-experienced intervention duration (in weeks).
| Variable | Modelinga,b the Probability of “Not Satisfied Overall” | ||
|---|---|---|---|
| Odds Ratio | 95% Confidence Interval | P | |
| Black/Asian/>1 Race c , Not Hispanic or Latino vs White, Not Hispanic or Latino | 2.96 | 1.43, 6.11 | <0.001 |
| Hispanic or Latino vs White, Not Hispanic or Latino | 0.40 | 0.16, 1.05 | 0.004 |
| Cumulative number of preintervention adverse events (≥1 vs None) | 1.88 | 1.15, 3.07 | 0.012 |
| % intervention duration experiencing adverse events (>5% vs ≤5%) | 1.78 | 1.08, 2.94 | 0.024 |
| Early termination (yes vs no) | 7.42 | 2.60, 21.17 | <0.001 |
The model includes complete data on all covariates from the 636 participants (of the 652 who completed the WIWI questionnaire) with 93 participants categorized as “Not Satisfied Overall” and 543 participants categorized as “Yes Satisfied Overall”.
The odds ratios shown are adjusted for age (P = 0.419), sex (P = 0.654), and participant-experienced intervention duration (P = 0.128). Duration of intervention serves as a potential confounding variable that represents the trial, disease group, and the length of exposure to the intervention for each participant. Put simply, because intervention duration is tethered to the scientific question driving the trial design, it makes sense to adjust for this factor when interpreting the results of the other factors. The model also controls for the two standard demographic variables age and sex that the scientific community would want us to control for when interpreting the other factors.
8 participants who self-reported as Asian or >1 Race (Not Hispanic or Latino) were combined with the 39 Black, Not Hispanic or Latino participants.
Secondary Outcomes
Overall, 24% of participants indicated that their quality of life improved from participating in the research study, while 7% indicated that it got worse; the remaining 69% indicated that it stayed the same (Figure 2A). There were no notable differences in responses by age and sex; however, relative to White, non-Hispanic or Latino, more Hispanic or Latinos reported that their quality of life improved in both US and non-US sites, but the difference was more pronounced in non-US sites where the majority of Hispanic or Latinos were enrolled (88 of 98 were enrolled in Puerto Rico or Honduras; Figure 2B).
Figure 2A. Response percentages for the fourth question (Q4) on the WIWI questionnaire by chemoprevention trial. Of the 652 participants who completed the first three questions on the WIWI questionnaire, 649 also responded to Q4.
Figure 2B. Response percentages for the fourth question (Q4) on the WIWI questionnaire by race/ethnicity and according to sites in the United States (US) versus Non-US sites.
Fifty-two percent of participants indicated their overall experience with participating in the research study was better than expected and 4% indicated that their overall experience was worse than expected; in contrast with Q4, the response percentages for Q5 did not vary substantially across trials (Figure 3).
Figure 3.
Response percentages for the fifth question (Q5) on the WIWI questionnaire by chemoprevention trial. Of the 652 participants who completed the first three questions on the WIWI questionnaire, 597 also responded to Q5.
DISCUSSION
To the best of our knowledge, this study is the first to look at satisfaction with participation in early-phase chemoprevention trials for higher-risk men and women. Among the hundreds of subjects who participated in CPN trials, 85% were satisfied overall. Features that were independently associated with not satisfied overall included race/ethnicity, the number of preintervention AEs, the percentage of the intervention duration with AEs, and early termination. In chemoprevention trials there is not a pressing need for an individual to participate in contrast to treatment trials for patients diagnosed with cancer. From a practical standpoint, to improve accrual and retention, investigators want to aim high to achieve a high overall satisfaction as opposed to some or a little satisfaction being enough.
Knowing which features are associated with not satisfied overall can be used to help investigators develop strategies to improve accrual, retention, and adherence as well as enhance participant diversity in future chemoprevention trials. The most frequently reported preintervention AEs varied considerably by trial (Supplementary Table S4). Oftentimes, chemoprevention trials have preregistration inclusion/exclusion criteria to facilitate screening evaluations. Taking measures to modify the preintervention requirements with the aim of minimizing the risk of participants experiencing preintervention AEs may lead to increased satisfaction with trial participation. Additionally, diverse clinical trial participation is important33 and low participation of Blacks in cancer clinical trials is well-known. Six percent (or 39) of the trial participants were Black, non-Hispanic or Latino and this racial/ethnic group reported lower overall satisfaction compared with White, non-Hispanics or Latinos and compared with Hispanics or Latinos. Approaches to address the needs and participation concerns of Black, non-Hispanics or Latinos can have the potential to lead to increased participation of patients within this race/ethnicity in clinical trials.
The association observed between early termination and overall satisfaction, and between increased time with AEs while receiving the trial intervention and overall satisfaction, was not surprising. Given that this series represented the first instance of the WIWI questionnaire applied systematically to chemoprevention trials, these results were reassuring and helps support the utility of the tool in measuring participant satisfaction in men and women at increased cancer-risk, but otherwise healthy. Incorporating the WIWI questionnaire in chemoprevention trials can facilitate our understanding of whether novel retention methods to minimize the time a participant spends experiencing intervention AEs and the risk of early termination are effective.
Knowing the intervention AEs experienced by the trial participants can be beneficial to investigators planning chemoprevention trials so that the impact of certain AEs on the targeted study populations can be weighed against the potential benefit of the investigational agents. Supplementary Table S4 shows the top five most frequently reported intervention AEs overall and by trial. The two most frequently reported AEs during the intervention period were injection site reaction and diarrhea, followed by headache, fatigue, and dyspepsia. The 13 chemoprevention trials were investigating an array of oral- and injection-class candidate agents. Injection site reaction was the most frequent intervention AE in each of the three trials evaluating an injection agent, while diarrhea was the most frequently reported intervention AE in eight of the ten trials evaluating an oral agent. The specific AEs that led to treatment discontinuation were not recorded in 5 subjects, and for the other 14 subjects, the reported verbatim AE terms varied.
Rigorous thought and methods guided this study. For example, to classify a participant as being satisfied overall with their participation in the chemoprevention trial, the participant needed to answer “yes” to Q1, Q2, and Q3. The WIWI questionnaire and, particularly, Q1, Q2, and Q3 were all measured in one sitting and, therefore, represented a single experience for each participant. The pairwise agreement in response categories (yes, no, uncertain) between Q1, Q2, and Q3 was not strong. Collapsing over Q1, Q2, and Q3 to obtain an indicator for overall satisfaction was not arbitrary. The decision was based on statistical criteria (lack of strong agreement) and from reading the individual questions; on face value the first three questions are different.
The fact that greater than one fifth of the participants indicated that their quality of life improved from participation in the research study was unexpected. Relative to White, non-Hispanics or Latinos, more Hispanics or Latinos reported that their quality of life improved in both US and international sites, but the difference was more pronounced in international sites where the majority of Hispanic/Latino participants were enrolled. In MAY2015-05-01, for example, greater than three fifths of the participants indicated that their quality-of-life improved from trial participation. Of the 47 participants in MAY2015-05-01 with a median age of 62 years, 45 were Hispanic or Latino with the remaining two participants self-reporting as White, Not Hispanic or Latino. MAY2015-05-01 was conducted at sites in rural western Honduras and Puerto Rico, which were representative of Caribbean and Mesoamerican populations and linked to large U.S. immigrant populations34. Notably, western Honduras is representative of The Central America Four region (Guatemala, Honduras, El Salvador, and Nicaragua) and is the largest low- and middle-income country region in the Western Hemisphere35. The concept of perceived quality of life is influenced by many factors, including by culture36, and culture may have shaped, in part, the favorable subjective experiences of these trial participants.
The subjects’ overall experience with trial participation was generally aligned with or exceeded their expectations. Based on Q5, between 4% to 17% of trial participants reported that their experience was worse than they had expected across the 13 chemoprevention trials. This highly favorable feedback with respect to the subjects’ overall study experience does suggest that the recruitment and retention methods employed were facilitating success.
One limitation is that this study is comprised of secondary analyses. Admittedly, these secondary analyses were carefully planned prior to embarking on the current study, but the chemoprevention trials themselves were not prospectively designed with the primary goals of understanding participants’ satisfaction with trial participation. Furthermore, there are likely features associated with participant’s satisfaction that were not part of the data collection or could not be derived based on the available data. Nonetheless, within this large and diverse cohort we were able to identify potentially actionable features associated with participation satisfaction that should be considered when developing future chemoprevention trials.
Conclusion
In this large and, demographically and geographically diverse cohort of 652 early-phase chemoprevention trial participants, 85% of the participants indicated satisfaction with their participation in the trial. Race/ethnicity, the number of preintervention AEs, the percentage of the intervention duration with AEs, and early termination were the features that were independently associated with not satisfied overall. These findings may inform design of future chemoprevention trials and help investigators improve accrual, retention, adherence, and diversity in this underexplored research setting.
Supplementary Material
ACKNOWLEDGEMENTS
The authors would like to thank the Cancer Prevention Network and especially the participants and sites that enrolled to the trials. We also thank the National Cancer Institute, Division of Cancer Prevention, for funding the Cancer Prevention Network (P.J. Limburg, S.J. Mandrekar, 2003–2012: N01-CN35000 and P.J. Limburg, S.J. Mandrekar, D. Zahrieh, C.A. Strand, 2012–2019: HHSN2612012000421).
Declaration of any potential conflicts of Interest:
Dr. Paul J. Limburg serves as Chief Medical Officer for Screening at Exact Sciences and holds stock in the company. Sumithra J. Mandrekar reports consulting fees and stock options from Harbinger Health Inc. outside the submitted work. The other authors do not have any potential conflicts of interest.
REFERENCES
- 1.Briercheck E, Pyle D, Adams C, Atun R, Booth C, Dent J, et al. , Unification of Efforts to Improve Global Access to Cancer Therapeutics: Report From the 2022/2023 Access to Essential Cancer Medicines Stakeholder Summit. JCO Glob Oncol, 2024. 10: p. e2300256. [DOI] [PubMed] [Google Scholar]
- 2.Steward WP and Brown K, Cancer chemoprevention: a rapidly evolving field. Br J Cancer, 2013. 109(1): p. 1–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Chlebowski RT, Menon R, Chaisanguanthum RM, and Jackson DM, Prospective evaluation of two recruitment strategies for a randomized controlled cancer prevention trial. Clin Trials, 2010. 7(6): p. 744–748. [DOI] [PubMed] [Google Scholar]
- 4.Hudmon KS, Love RR, and Chamberlain RM, Perceived benefits of and barriers to participation in a phase I/II colon cancer chemoprevention trial. J Cancer Educ, 1999. 14(2): p. 83–87. [DOI] [PubMed] [Google Scholar]
- 5.Hudmon KS, Stoltzfus C, Chamberlain RM, Lorimor RJ, Steinbach G, and Winn RJ, Participants’ perceptions of a phase I colon cancer chemoprevention trial. Control Clin Trials, 1996. 17(6): p. 494–508. [DOI] [PubMed] [Google Scholar]
- 6.Kumar NB, Quinn GP, Alexandrow MG, Gray J, Schell M, Sutton S, et al. , Chemoprevention Trial Feasibility Using Botanicals in Exceptionally High Risk Populations for Lung Cancer. J Clin Trials, 2014. 4(4). [Google Scholar]
- 7.Tangrea JA, Adrianza ME, and Helsel WE, Patients’ perceptions on participation in a cancer chemoprevention trial. Cancer Epidemiol Biomarkers Prev, 1992. 1(4): p. 325–330. [Google Scholar]
- 8.Atherton PJ, Szydlo DW, Erlichman C, and Sloan JA, What can Phase I clinical trials tell us about quality of life? A pilot study (MC0115). Clinical Research and Trials, 2015. 1(1): p. 11–14. [Google Scholar]
- 9.Sloan JA, Mahoney MR, and Sargent DJ, Was it worth it (WIWI)? Patient satisfaction with clinical trial participation: results from North Central Cancer Treatment Group (NCCTG) phase III trial N0147. J Clin Oncol, 2011. 29((suppl 15)): p. 6122–6122. [Google Scholar]
- 10.Thanarajasingam G, Basch B, Mead-Harvey C, Bennett AV, Mazza GL, Schwab G, et al. , An Exploratory Analysis of the “Was It Worth It?” Questionnaire as a Novel Metric to Capture Patient Perceptions of Cancer Treatment. VALUE HEALTH, 2022. 25(7): p. 1081–1086. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Fain R, Chen YY, Rana SR, Dang JL, Haly A, Franklin A, et al. , Was it worth it (WIWI)? An OHSU Knight Cancer Institute retrospective analysis of patient satisfaction following radiotherapy. Journal of Clinical Oncology, 2016. 34(3). [Google Scholar]
- 12.Tanabe KK, Zahrieh D, Strand CA, Hoshida Y, Flotte TJ, Della’Zanna G, et al. , Epidermal Growth Factor Receptor Inhibition With Erlotinib in Liver: Dose De-Escalation Pilot Trial as an Initial Step in a Chemoprevention Strategy. Gastro Hep Adv, 2024. 3(3): p. 426–439. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Schoen RE, Boardman LA, Cruz-Correa M, Bansal A, Kastenberg D, Hur C, et al. , Randomized, Double-Blind, Placebo-Controlled Trial of MUC1 Peptide Vaccine for Prevention of Recurrent Colorectal Adenoma. Clin Cancer Res, 2023. 29(9): p. 1678–1688. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Jacobson JM, Zahrieh D, Strand CA, Cruz-Correa M, Pungpapong S, Roberts LR, et al. , Phase I Trial of a Therapeutic DNA Vaccine for Preventing Hepatocellular Carcinoma from Chronic Hepatitis C Virus (HCV) Infection. Cancer Prev Res (Phila), 2023. 16(3): p. 163–173. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Samadder NJ, Foster N, McMurray RP, Burke CA, Stoffel E, Kanth P, et al. , Phase II trial of weekly erlotinib dosing to reduce duodenal polyp burden associated with familial adenomatous polyposis. Gut, 2023. 72(2): p. 256–263. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Weinberg DS, Foster NR, Della’Zanna G, McMurray RP, Kraft WK, Pallotto A, et al. , Phase I double-blind, placebo-controlled trial of dolcanatide (SP-333) 27 mg to explore colorectal bioactivity in healthy volunteers. Cancer Biol Ther, 2021. 22(10–12): p. 544–553. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Lam S, Mandrekar SJ, Gesthalter Y, Allen Ziegler KL, Seisler DK, Midthun DE, et al. , A Randomized Phase IIb Trial of myo-Inositol in Smokers with Bronchial Dysplasia. Cancer Prev Res (Phila), 2016. 9(12): p. 906–914. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Chak A, Buttar NS, Foster NR, Seisler DK, Marcon NE, Schoen R, et al. , Metformin does not reduce markers of cell proliferation in esophageal tissues of patients with Barrett’s esophagus. Clin Gastroenterol Hepatol, 2015. 13(4): p. 665–672 e661–664. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Limburg PJ, Mandrekar SJ, Aubry MC, Ziegler KL, Zhang J, Yi JE, et al. , Randomized phase II trial of sulindac for lung cancer chemoprevention. Lung Cancer, 2013. 79(3): p. 254–261. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Falk GW, Buttar NS, Foster NR, Ziegler KL, Demars CJ, Romero Y, et al. , A combination of esomeprazole and aspirin reduces tissue concentrations of prostaglandin E(2) in patients with Barrett’s esophagus. Gastroenterology, 2012. 143(4): p. 917–926 e911. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Reid JM, Walden CA, Qin R, Ziegler KL, Haslam JL, Rajewski RA, et al. , Phase 0 clinical chemoprevention trial of the Akt inhibitor SR13668. Cancer Prev Res (Phila), 2011. 4(3): p. 347–353. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Limburg PJ, Mahoney MR, Ziegler KL, Sontag SJ, Schoen RE, Benya R, et al. , Randomized phase II trial of sulindac, atorvastatin, and prebiotic dietary fiber for colorectal cancer chemoprevention. Cancer Prev Res (Phila), 2011. 4(2): p. 259–269. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Wigle DA, Mandrekar SJ, Allen-Ziegler K, Gesthalter Y, Holland P, Aubry M, et al. , Pioglitazone as a candidate chemoprevention agent for lung cancer: A pilot window trial in early stage NSCLC. Journal of Clinical Oncology, 2014. 32(15): p. 1581–1581. [Google Scholar]
- 24.Weinberg DS, Lin JE, Foster NR, Della’Zanna G, Umar A, Seisler D, et al. , Bioactivity of Oral Linaclotide in Human Colorectum for Cancer Chemoprevention. Cancer Prev Res (Phila), 2017. 10(6): p. 345–354. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Banerjee M, Capozzoli M, McSweeney L, and Sinha D, Beyond kappa: A review of interratter agreement measures. The Canadian Journal of Statistics, 1999. 27(1): p. 3–23. [Google Scholar]
- 26.Breiman L, Random forests. Machine Learning, 2001. 45(1): p. 5–32. [Google Scholar]
- 27.Liaw A and Wiener M, Classification and regression by randomForest. R News, 2002. 2(3): p. 18–22. [Google Scholar]
- 28.Genuer R and Poggi J, Random forests with R. 2020, Switerland: Springer. p. 59. [Google Scholar]
- 29.Kuhn M and Johnson K, Applied Predictive Modeling. 2016, New York: Springer. p. 43–47. [Google Scholar]
- 30.Cramér H, The two-dimensional case, in Mathematical Methods of Statistics. 1946, Princeton: Princeton University Press. p. 282. [Google Scholar]
- 31.Spearman C, The Proof and Measurement of Association between Two Things. The American Journal of Psychology, 1904. 15(1): p. 72–101. [Google Scholar]
- 32.Peduzzi P, Concato J, Kemper E, Holford TR, and Feinstein AR, A simulation study of the number of events per variable in logistic regression analysis. J Clin Epidemiol, 1996. 49(12): p. 1373–1379. [DOI] [PubMed] [Google Scholar]
- 33.Schwartz AL, Alsan M, Morris AA, and Halpern SD, Why Diverse Clinical Trial Participation Matters. N Engl J Med, 2023. 388(14): p. 1252–1254. [DOI] [PubMed] [Google Scholar]
- 34.Montalvan-Sanchez EE, Hernandez-Marrero J, Norwood DA, Gonzalez-Pons M, Dominguez RL, Rodriguez LM, et al. , Establishment of a Mesoamerican-Caribbean South-South Research Platform: Challenges in the Meriva (Curcuminoids) Gastric Cancer Chemoprevention Trial. Cancer Prev Res (Phila), 2024. 17(12): p. 549–555. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Pineros M, Frech S, Frazier L, Laversanne M, Barnoya J, Garrido C, et al. , Advancing Reliable Data for Cancer Control in the Central America Four Region. J Glob Oncol, 2018. 4: p. 1–11. [Google Scholar]
- 36.Marshall PA, Cultural influences on perceived quality of life. Semin Oncol Nurs, 1990. 6(4): p. 278–284. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
Data and materials are available in the NCI’s Cancer Data Acquisition System and may be requested using the standardized process (https://cdas.cancer.gov/).