Factors Associated with Adherence to an End-of-Study Biopsy: Lessons from the Prostate Cancer Prevention Trial (SWOG-Coordinated Intergroup Study S9217)

Medical professionals and researchers recognize that nonadherence to medical treatment remains a challenge and have devoted significant resources to the development of intervention strategies to improve adherence rates. Despite these efforts, half of the strategies that are developed and tested fail (1). The preponderance of the literature is devoted to medication adherence, particularly in patients with chronic illnesses, such as diabetes, cardiovascular disease, or HIV, but there is a small literature examining adherence to clinical procedures.

The Prostate Cancer Prevention Trial (PCPT) was unusual in that its primary outcome (prevalence of prostate cancer during the 7-year trial) was based on a nonclinically indicated biopsy at the end of the trial (2). A few studies have addressed adherence to a prostate biopsy recommendation and factors associated with agreeing to have the procedure. The Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial (PLCO) offers an opportunity to examine adherence to prostate biopsy screening recommendations. In this trial, across cancer disease sites, adherent participants were younger, had higher levels of education, and did not have a first-degree relative with cancer (3). Pinsky and colleagues examined factors associated with undergoing a prostate biopsy after a positive PLCO screen (4). Participants were more likely to undergo a biopsy after a prostate-specific antigen (PSA) value of >7 ng/mL and after a positive digital rectal exam (DRE); a history of prostate problems and Asian ethnicity were also important correlates with adherence (4). Men over the age of 70 were less likely to obtain a prostate biopsy when presented with a positive screening test in this study (4). Moul noted that 55% of men with a positive PSA (>4 ng/mL) had a biopsy within 3 years of the screening information; even with the highest risk category of PSA levels (PSA level >10 ng/mL), only 75% obtained a biopsy by 3 years (5). In a Veterans Administration clinic–based screening program for prostate cancer, Krongrad and colleagues reported that 57% of men with abnormal PSAs and/or DREs obtained a prostate biopsy (6). These reports suggest that presentation of high-risk status is not sufficient to guarantee that a man will obtain a prostate biopsy.

This analysis expands the potential factors associated with adherence to an invasive end-of-study (EOS) procedure that were previously identified using PCPT data (7, 8) by examining psychosocial outcomes, participant health status, participant adherence, and site characteristics. These factors were drawn from well-known health behavior models such as the Theory of Planned Behavior and the Health Belief model (9, 10). Identifying factors associated with potential nonadherence in a timely manner allows clinical trial researchers to select the population of participants most in need of an intervention to increase adherence. One year before the EOS biopsy coincides with a time when participants are close to the time of the EOS biopsy, yet far enough out that an intervention strategy can be delivered and have sufficient time to work. This approach mimics the real-world experiences of study staff working on clinical trials by only using information site staff have access to at the time the intervention strategy is applied.

This study has two aims. The first is to investigate novel factors associated with adherence through multivariate logistic regression; the second is to use receiver operating characteristic (ROC) curves to compare the specificity and sensitivity of our model among men who were and were not prompted for a biopsy.

The PCPT was a 7-year randomized, double-blind, placebo-controlled trial of the efficacy of finasteride for the prevention of prostate cancer (11, 12). The primary outcome for the trial was prevalence of prostate cancer as measured by a transrectal ultrasonographic–guided biopsy of the prostate (minimum of six cores) at the end of 7 years or an interim diagnosis of prostate cancer. Men randomized to finasteride had a 24.8% reduction in the prevalence of prostate cancer compared with those who were randomized to placebo. Details of the trial design and eligibility criteria have been presented elsewhere (2, 13).

This analysis examines factors associated with EOS biopsy adherence using information obtainable as of year 6, 1 year before the EOS biopsy. Participants who died or became lost to follow-up between years 6 and 7 (n = 431) were included in the primary model and counted as having no EOS biopsy.

Eligibility and exclusion criteria are shown in Fig. 1. Participants were eligible if none of the following events happened as of year 6: prostate cancer diagnosis; prostatectomy or cystoprostatectomy; death; and loss to follow up. Participants randomized after March 26, 1996, were excluded because of early study closure on June 24, 2003; these participants did not complete their EOS windows (7-year anniversary of their randomization + 90 days) and their adherence could have been influenced by the early publication of study results. Missing demographic data on 14 participants led to their exclusion from these analyses.

The analysis is stratified by biopsy prompt history: Those participants who had ever been prompted for a prostate biopsy before year 6 (via elevated PSA or DRE suspicious for cancer) and those who had not, as these groups have potentially different motivations for adherence with the EOS biopsy.

The covariates considered for inclusion in the models are: age at year 6; demographics measured at randomization; comorbidities over the course of the trial; health-related quality of life (HRQL) at year 6; measures of participant adherence at year 6; prior negative biopsy; and site characteristics. See Table 1 for a complete listing of covariates examined in this analysis.

HRQL covariates were assessed using the Short Form-36 (SF-36) Health Survey. SF-36 scales (14–16) are scored on a 0 to 100 scale with higher values reflecting better HRQL. Table 1 lists the four scales selected from the full SF-36 for this analysis, which cover key domains of HRQL. Participants who were off treatment were no longer required to submit HRQL forms. PCPT outcome and covariate measures for the HRQL component of the PCPT are described in Moinpour and colleagues (17) Participants with missing forms and participants with below average scores had similar ORs and were grouped and compared with participants with average or better scores. An “average or better” score is defined as a score that is equal to or greater than the population mean minus half an SD (a moderate-sized effect; ref. 18) for that particular score, based on the SF-36 scales for the U.S. general population for men 55 and older (17); these age-specific norms were provided for the PCPT by Dr. John Ware (personal communication, 1994), based on the normative database for the SF-36 published in the SF-36 manual (19).

General adherence covariates measured how well a participant followed the study protocol. Study drug administration was once daily; participants who stopped taking the study drug were considered off treatment. Adherence to this regimen was measured using pill counts and calculated as a percentage of required pills taken over time; at least 80% was considered adherent. Study contacts were quarterly, with visits every 6 months and phone calls at 3 months between visits. A participant who missed a regularly scheduled visit or call was counted as ever missing a visit. PSA tests and DREs were required annually with abnormal results prompting a biopsy. Participant refusal of a prompted biopsy by year 6 was considered a measure of nonadherence.

PCPT was conducted at 219 sites, with enrollment ranging from 1 to 1,444 randomized participants. Site characteristics included being an NCI-funded Community Clinical Oncology Program (CCOP) site (20), site performance, and receipt of a retention and adherence (R&A) grant. Site performance was measured by submission rate of study forms, with low submission rates indicating poor site performance. R&A grants were intended to support overall site R&A activities. Additional adherence interventions are described in Table 2. Most of these activities were initiated midway through the trial, particularly after the conduct of a series of focus groups in which we examined barriers and reinforcements for complying with all trial requirements.

The probability of participants having an EOS biopsy versus nonadherence was modeled using logistic regression. Analyses used PROC SURVEYLOGISTIC in SAS 9.2 (SAS Institute Inc.) and accounted for participants clustered within sites.

Differences in descriptive study covariates were tested with the t test and the χ² test for continuous and categorical measures, respectively. Collinearity diagnostics for multivariate models were performed on all candidate covariates, looking at variance inflation factor. Finally, potential interactions between site characteristics, general adherence, and age were tested; statistically significant interactions were included in the final model. Multivariate models were conducted separately for men who received a biopsy prompt and men who did not.

Study covariates were chosen based on prior literature (7–10); the final multivariate models included factors with significant bivariate relationships (Tables 1 and 3). In selecting these variables, we were aware of the interrelated nature of the many variables involved in encouraging study adherence (21). For this analysis, therefore, we examined a full range of potential interactions between study participants, study staff, and investigators that might reinforce study bonding and enhance participant adherence to study requirements. Predictive power was measured using the area under the ROC curve, with a value of ≥0.7 considered adequate discrimination (22). All data were used to calculate ROC curves (Fig. 2).

Excluded participants were less likely to be white (88% vs. 93%) or be from a larger site (33% vs. 41%). Differences were greater for participants lost to follow-up than participants excluded because of early study closure. Participants who had ever received a biopsy prompt by year 6 were more likely to have an EOS biopsy than those who had not received a prompt [unadjusted OR, 1.28; 95% confidence interval (CI), 1.17–1.41]. The EOS biopsy rate among men ever and never prompted for a biopsy was 67.4% and 61.7%, respectively. The combined EOS biopsy rate for all participants included in this analysis was 62.8%.

Table 3 provides descriptive information for participants by EOS biopsy adherence status. Participants who had an EOS biopsy were different from participants who did not have the procedure. Participants who obtained biopsies were more likely to be adherent to study drug at year 6 (84.2% vs. 46.9%) or have a DRE or PSA test at year 6 (98.5% vs. 75.2%). They were more likely to come from sites that were larger, received R&A grants, or were performing well. Differences for most characteristics presented in Table 3 are not very sizeable, despite the small P values.

Table 4 presents the final model and ORs for having an EOS biopsy for the subset of men who ever had a biopsy prompt by year 6. Main effects show that participants were more likely to adhere to the EOS biopsy if they were adherent to study visit schedules and procedures (adherent to study drug, had DRE/PSA test, and no missed contacts) and/or were in good health (younger age, high/better SF-36 Physical Functioning scores). The SF-36 Mental Health score was identified as being collinear with the other HRQL measures and was excluded from the analyses presented in Tables 4 and 5. Interaction results showed participants at larger sites that received R&A grants were more likely to have an EOS biopsy.

As shown in Table 5, the subset of participants never prompted for a biopsy by year 6 also were more likely to adhere to the EOS biopsy if they were adherent to study visit schedules and procedures and/or were in good health (high/better SF-36 General Health/Health Perceptions scores). Interaction results showed participants at larger sites that received R&A grants also were more likely to have an EOS biopsy.

ROC curves for each model are shown in Fig. 2. The models have adequate discrimination between men who will and will not receive an EOS biopsy; the area under the ROC curve for men ever and never prompted for a biopsy is 0.74 (95% CI, 0.72–0.76) and 0.77 (95% CI, 0.76–0.78), respectively.

Site size has a modest impact on the variability of EOS biopsy adherence. The intraclass correlation coefficient (ICC) for site size and EOS biopsy adherence among men ever prompted for a biopsy is 0.11 (0.11 for small sites; 0.07 for large sites). Among men never prompted for a biopsy ICC = 0.08 (0.08 for small sites; 0.05 for large sites).

The goal of this research was to identify factors prospectively associated with the EOS biopsy in the PCPT. More specifically, we sought to determine which factors identifiable at year 6 were associated with the willingness of a participant to undergo an invasive procedure 1 year later. The PCPT is unique because it was a prevention, rather than a treatment, trial: The biopsy procedure was critical for evaluating the intervention, but not clinically indicated for most participants. A recent study examining the use of research biopsies in therapeutic clinical trials shows that participants may not have a clear understanding of the risks and benefits of these procedures, suggesting that researchers need to improve study protocols and informed consents to ensure that participants fully understand the study procedures and requirements (23). This is particularly important in trials like the PCPT in which the EOS biopsy happens 7 years after participant enrollment.

Our results show that a variety of participant and site characteristics are associated with participant adherence to the invasive EOS requirement. At the participant level, adherence with study schedule and procedures and general good health is associated with higher EOS biopsy rates; participants at sites that were larger and/or procured R&A grants also had higher EOS biopsy rates.

Probstfield and colleagues have identified failing to adhere to the study agent regimen, missing a study visit, and going off study treatment as “red flags” (24, 25) that should be addressed during the course of the trial to keep participants fully engaged in trial activities and outcomes ascertainment.

These “red flags” coincide with the general adherence factors identified in this analysis as being associated with EOS biopsy adherence. Although the analysis does not indicate that improving adherence to study requirements will improve adherence to the EOS biopsy, it does lend support to their status as “red flags.”

Our study found that older men were less adherent, as did the PLCO for any diagnostic procedure (4). In the PCPT, the lower EOS biopsy adherence rate for older participants could be related to the perceived physical demands of the procedure or age-related medical contraindications. For example, one of the reasons physicians recommended that participants not obtain the EOS biopsy was the need to stop anticoagulation medication for this study-specific procedure. It is important to note that these physicians were not associated with the study, so their primary concern was the well-being of their patients.

We expect larger sites to perform better in nearly all aspects of the study, including having higher rates of EOS biopsies, primarily due to economy of scale. Successful large sites require better management practices to be efficient. Smaller sites can tolerate inefficiencies better and, due to small volume, may lack the motivation to streamline study practices. For example, the top accruing PCPT site had 1,444 participants, nearly three times the next largest site, and was considered one of the best sites of PCPT for data quality. The finding about R&A grants is ambiguous. We cannot determine from this analysis whether site use of R&A grants specifically affected EOS biopsy rates, or whether there is something different about the sites that applied for and received the R&A grants.

The primary concern of this analysis is correctly identifying participants at risk of not completing final study requirements. Identifying these participants shortly before the final study outcome assessment and targeting them with intervention strategies to increase adherence with final study requirements may be a better use of limited staff time than strategies directed at all participants throughout the trial. The cost of any strategies and the cost of identifying the appropriate participants, however, must be weighed against the benefit to the study. If the study design assumptions are not being met and the ability of the study to achieve its objectives is at risk, an intervention strategy may be worthwhile to salvage the study. In the case of the PCPT, we conducted a number of adherence activities to help site staff and participants recognize the importance of the EOS biopsy. These initiatives were not formally evaluated, although we solicited and received feedback from site staff and study participants on their reaction to most of these activities; this feedback was helpful each time and helped inform the next set of activities.

PCPT met its study specified biopsy rate of 60%, which implies reasonable adherence because the study design required a 60% rate ascertainment of prostate cancer status. Interventions were suggested to all PCPT sites for enhancing adherence with the EOS biopsy because this was an invasive trial outcome. As noted in Table 2, we held EOS informational/educational sessions with participants (often funded with an R&A grant), asked sites to show the EOS biopsy video distributed to all PCPT sites at the educational session and making it accessible to participants whenever they were at the site, distributed the EOS Biopsy brochure to all participants, and provided the EOS biopsy manual to be used by site staff to ready each site and its participants for this challenging outcome.

The study population was mostly white, highly educated, and healthy; the results may not be applicable to more diverse and underserved populations. To obtain the PCPT prostate biopsy endpoint, the study required a participant commitment of 7 years, and the models apply only to participants who remained in the study for at least 6 years.

A variety of factors are prospectively associated with complying with the EOS biopsy: adherence to basic requirements for research study participation, good participant health, and certain site characteristics. This model is perhaps best used as a starting point for other studies to consider when trying to increase adherence to invasive, nonclinically indicated procedures. These factors can be used to identify a subgroup of trial participants who are not likely to have a nonclinically indicated procedure. The study can then apply an appropriate intervention strategy to increase adherence to that procedure.

We conclude with a quotation from a PCPT participant who agreed to be in one of the focus groups held midway through the 7-year study period. The question was raised about the impact of any problems experienced with the study drug. One participant who indicated some problems with the drug said, “…but dropping out would be like jumping ship. All that time would be wasted.” The facilitator asked him, “For you or the study?” and the participant responded, “Well, for me and the study because I am the study.” Our ability to carry out successful long-term prevention trials is certainly enhanced if we can create a study environment that generates this level of participant commitment.

C.M. Moinpour has ownership interest in Merck (stock). No potential conflicts of interest were disclosed by the other authors.

The content of this work is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.

Conception and design: E.R. Gritz, K.B. Arnold, C.M. Moinpour, C.M. Tangen, J.F. Probstfield, S. Moody-Thomas, L.M. Minasian

Development of methodology: E.R. Gritz, K.B. Arnold, C.M. Moinpour, A.M. Burton-Chase, C.M. Tangen, M.M. Lieber, S. Moody-Thomas

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C.M. Tangen, J.F. Probstfield, W.A. See, M.M. Lieber, V. Caggiano, I.M. Thompson Jr

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K.B. Arnold, C.M. Moinpour, A.M. Burton-Chase, C.M. Tangen, P.J. Goodman, L.M. Minasian

Writing, review, and/or revision of the manuscript: E.R. Gritz, K.B. Arnold, C.M. Moinpour, A.M. Burton-Chase, C.M. Tangen, J.F. Probstfield, V. Caggiano, S. Moody-Thomas, C. Szczepanek, A. Ryan, S. Hill, P.J. Goodman, L.M. Minasian, F.L. Meyskens

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K.B. Arnold, A. Ryan, S. Carlin, F.L. Meyskens

Study supervision: E.R. Gritz, C.M. Moinpour, C.M. Tangen, J.F. Probstfield, M.M. Lieber, P.J. Goodman, L.M. Minasian, I.M. Thompson Jr

Other (membership on the PCPT Recruitment and Adherence Committee, which developed and provided guidance, logistical suggestions, and tools to support protocol implementation): C. Szczepanek

Other (data management support): S. Carlin

Other (coordinator for PCPT adherence initiatives; prepared reports for these initiatives): S. Hill

Other (worked on the study team, from the government side, for the entire study period, assisted with recruitment and retention, and focused on the end-of-study biopsy): R.M. Padberg

The authors thank the men in the PCPT who participated for the 7 years of the trial. In addition, the authors thank the physicians, nurses, and clinical research associates who recruited participants for the trial and successfully followed them for the 7-year period.

PCPT Recruitment and Adherence Committee Members: Otis Brawley, MD; Susie Carlin, BA; Dee Daniel; Edward DeAntoni, PhD (Co-Chair with Dr. Gritz for several years); Patricia Ganz, MD; Ellen R. Gritz, PhD (Chair); Carolyn Harvey, PHD; M. Shannon Hill, BM; Lori Minasian, MD; Carol M. Moinpour, PhD; Rose Mary Padberg, RN, MA; Jeanne Parzuchowski, RN, MS; Jeffrey Probstfield, MD; Anne Ryan, RN; Connie Szczepanek, RN, BSN; Sarah Moody Thomas, PhD; Ian Thompson, MD; and Clarence Vaughn, MD.

This research was supported by Public Health Service grant number CA37429 (C.D. Blanke; C.A. Coltman) from the Department of Cancer Prevention, National Cancer Institute, NIH, Department of Health and Human Services (Bethesda, MD). The study agents (finasteride and placebo) were provided by Merck, Inc. Merck, Inc. and the National Cancer Institute (NCI), Division of Cancer Prevention (DCP), also provided funding to produce videos and to support projects to enhance trial recruitment and adherence. E. Gritz was supported in part by funding from the National Cancer Institute and P30CA16672 (to R.A. DePinho).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.