Abstract
Objective
Identify predictors of non-compliance with first round screening exams in the Prostate, Lung, Colorectal and Ovarian (PLCO) Cancer Screening Trial.
Method
PLCO was conducted from 1993–2011 at 10 US institutions. A total of 154,897 healthy men and women ages 55–74 years were randomized. Intervention arm participants were invited to receive gender-appropriate screening exams for prostate, lung, colorectal and ovarian cancer. Using intervention-arm data (73,036 participants), non-compliance percentages for 13 covariates were calculated, as were unadjusted and adjusted odds ratios (ORs), and 95% confidence intervals. Covariates included demographic factors as well as factors specific to PLCO (e.g., method of consent, distance from screening center).
Results
The rate of non-compliance was 11% overall but varied by screening center. Significant associations were observed for most covariates but indicated modest increases or decreases in odds. An exception was use of a two-step consent process (consented intervention arm participants for exams after randomization) relative to a one-step process (consented all participants prior to randomization) (OR: 2.2, 95% CI: 2.0–2.5). Non-compliance percentages increased with further distance from screening centers, but ORs were not significantly different from 1.
Conclusions
Many factors modestly influenced compliance. Consent process was the strongest predictor of compliance.
Keywords: Mass screening, adherence, compliance, cancer, randomized controlled trial as subject
BACKGROUND
The success of randomized controlled trials (RCTs) depends upon many accomplishments, including meeting or exceeding pre-specified levels of compliance with interventions. Failure to meet compliance goals can reduce statistical power, which may necessitate recruitment of more participants than originally planned or extension of follow-up. These changes can be deleterious, particularly if funds are not available for unanticipated activities or the pool of potential participants has become limited. Therefore, it is critical to identify characteristics that affect compliance.
Patient-related predictors of compliance with therapeutic drug regimens for cancer, both in experimental and community-based settings, have been studied extensively, demonstrating the clinical community’s recognition of the importance of compliance when patients or subjects are ill. Compliance with chemopreventive regimens has been reported for RCTs of persons at above-average risk of cancer of the breast [1,2] and lung [3], as well as RCTs like the Women’s Health Initiative, which enrolled average risk women and had breast and colon cancer as primary endpoints [4]. Predictors of screening regimen compliance in RCTs of persons at average risk who reside in the developed world have been published for only one trial (the UK flexible sigmoidoscopy trial) [5]1, but only race and attitudes concerning colorectal cancer and screening for that cancer were examined. Also, this trial randomized persons prior to consenting them for screening exams, and thus did so without their knowledge.
To explore multiple predictors of compliance in mass screening RCTs conducted in the US, we analyzed data from the Prostate, Lung, Colorectal and Ovarian Cancer Screening Trial (PLCO), an RCT of cancer screening efficacy in men and women ages 55–74 years [7]. Roughly half the 154,897 participants were randomly assigned to an intervention arm and offered specific screening exams multiple times during their first six years of enrollment. These exams required a clinic visit that included a blood draw and, at certain study visits, other invasive and non-invasive clinical procedures.
METHODS
The PLCO Trial
PLCO, a multiphasic RCT, began in 1992, enrolled participants through mid-2001, screened through 2006, and followed each participant until withdrawal, death, 13 years of follow-up, or December 31, 2009 (whichever occurred first) [7]. Final primary results were published in 2011 and 2012 [8–11]. A total of 154,897 participants, aged 55–74 years at entry, were enrolled at one of ten screening centers nation-wide and were randomized to either an intervention or control arm [12]. At baseline, control arm participants were advised to receive their usual medical care, and intervention arm participants were offered a blood-based PSA exam and digital rectal exam (prostate, men only), a single view chest x-ray (lung), a flexible sigmoidoscopy (colorectal), and a blood-based CA-125 exam and transvaginal ultrasound (ovarian, women only). For prostate and ovarian cancer, blood-based exams were offered annually for five more years, and invasive exams were offered annually for three more years. For lung cancer, ever smokers were offered chest x-ray annually for three more years; never smokers were offered that test annually for two more years. For colorectal cancer, one additional flexible sigmoidoscopy was offered at either year 3 or 5, with year of exam dependent on date of enrollment due to a mid-study protocol change. In almost all instances, all exams for a given study year were performed at a single clinic visit that lasted no more than 2 hours. All screening exams were offered at no cost to the participant; some screening centers provided gas cards or taxi vouchers to offset the cost of transportation to the screening center. Either at or prior to the first screening visit, participants were asked to complete the Baseline Questionnaire Form (BQF), which was gender-specific and queried participants about numerous factors, including demographics, prior and current health history, and family history of cancer.
Informed consent
All screening centers received Institutional Review Board approval to conduct trial activities. Two methods of consent were used: single and dual. Seven centers used a single consent process, which consented for enrollment and randomization at the same time. The remaining three (Henry Ford Health System, Washington University School of Medicine, and Pacific Health Research and Education Institute) initially used a dual consent process. This process began with consent for administration of the BQF and follow-up for cancer incidence and vital status. Consented participants were then randomized, without their knowledge, to the intervention or control arm, and only participants randomized to the intervention arm were told of their assignment; they also received a second consent, which consented for administration of screening exams. The single consent method was expected to increase compliance in the intervention arm but lead to contamination in the control arm; the dual consent method, while expected to decrease participation in the intervention arm, was expected to decrease contamination in the control arm. Contamination refers to the receipt of the screening exams under study by control arm participants. Because the rate of refusal to participate among dual-consent participants randomized to the intervention arm was unacceptably high, the three screening centers switched to the single consent process, with two switching in 1995 and one in 1997.
Compliance
We examined compliance with the first screening round, because all intervention arm participants were offered all screening exams. We chose not to create a compliance index that reflected compliance across study years for a number of reasons: participant relocation, changes in eligibility for exams due to protocol changes and cancer diagnoses, and expected-to-be important compliance predictors, such as declining health status, for which we had no or incomplete data. Participants were classified as non-compliant if no screening exams were completed within 11 months of randomization, and compliant otherwise.
Analysis
All intervention arm participants who did not die or withdraw between randomization and the first screening visit were included in our analyses, although we excluded those who did not complete the BQF or omitted an answer for at least one of the questions under consideration. We examined the relationship between non-compliance and age at randomization, gender, race, educational attainment, body mass index (BMI), presence or history of a co-morbidity at baseline (bronchitis, cirrhosis, diabetes, emphysema, heart attack, hepatitis, stroke, or personal history of cancer), smoking status, marital status, occupation at baseline, family history of a PLCO cancer, screening center, consent type, and year of randomization. The three screening centers (University of Colorado Anschutz Medical Campus, Henry Ford Health System, University of Utah Health Sciences Center) with an interest in the relationship of travel distance and non-compliance used Mapquest.com to calculate, for 1500 randomly selected participants who were eligible for the first screening visit (500 at each center), the distance from the participants’ baseline home addresses to the screening center. Distance was categorized using screening center-specific tertile values due to variation in population density in the catchment areas of the three centers. We used logistic regression models to calculate both unadjusted and adjusted odds ratios. SAS statistical software (Version 9) was used for all statistical analyses.
RESULTS
Entire cohort
Of the 77,445 participants randomized to the intervention arm, 77,436 were eligible for the first screening visit. Our analyses included the 73,036 participants with complete covariate information, or 94% of those eligible for the first screening visit. The average rate of non-compliance in this group was 11%. Rates of non-compliance varied by screening center: they were very low at the University of Alabama at Birmingham (2%) and the University of Minnesota School of Public Health (2%) screening centers, but high at the Henry Ford Health System screening center (26%). For factors other than screening center, the lowest non-compliance rates were observed for males (8%), participants with college degrees or post-graduate training (8%), and persons enrolled using the single consent method (9%); the highest non-compliance rates were observed for participants enrolled using the dual method (29%), participants with a BMI of 18.5 or less (21%) and participants who classified their occupational status as disabled/extended sick leave (20%). (Table 1)
Table 1.
All | Sampled for distance questions | |||||||||
---|---|---|---|---|---|---|---|---|---|---|
| ||||||||||
Eligible* for any T0 screen | Compliant | Not compliant | Eligible* for any T0 screen | Compliant | Not compliant | |||||
| ||||||||||
N | N | % | N | % | N | N | % | N | % | |
| ||||||||||
All intervention arm participants | 77,436 | 67,466 | 87.1 | 9970 | 12.9 | 1500 | 1223 | 81.5 | 277 | 18.5 |
| ||||||||||
Complete covariate information | 73,036 | 65,243 | 89.3 | 7793 | 10.7 | 1425 | 1176 | 82.5 | 249 | 17.5 |
| ||||||||||
Characteristics | ||||||||||
| ||||||||||
Age at randomization (years) | ||||||||||
Younger than 59 | 24,450 | 21,932 | 89.7 | 2518 | 10.3 | 425 | 349 | 82.1 | 76 | 17.9 |
60–64 | 22,470 | 20,201 | 89.9 | 2269 | 10.1 | 465 | 393 | 84.5 | 72 | 15.5 |
65–69 | 16,413 | 14,631 | 89.1 | 1782 | 10.9 | 315 | 263 | 83.5 | 52 | 16.5 |
70 or older | 9703 | 8479 | 87.4 | 1224 | 12.6 | 220 | 171 | 77.7 | 49 | 22.3 |
| ||||||||||
Gender | ||||||||||
Female | 36,998 | 32,213 | 87.1 | 4785 | 12.9 | 755 | 606 | 80.3 | 149 | 19.7 |
Male | 36,038 | 33,030 | 91.7 | 3008 | 8.3 | 670 | 570 | 85.1 | 100 | 14.9 |
| ||||||||||
Race | ||||||||||
White, non-Hispanic | 64,756 | 58,160 | 89.8 | 6596 | 10.2 | 1230 | 1018 | 82.8 | 212 | 17.2 |
Black, non-Hispanic | 3623 | 3059 | 84.4 | 564 | 15.6 | 109 | 84 | 77.1 | 25 | 22.9 |
Hispanic | 1358 | 1154 | 85.0 | 204 | 15.0 | 70 | 61 | 87.1 | 9 | 12.9 |
Asian | 2719 | 2399 | 88.2 | 320 | 11.8 | 8 | 8 | 100 | 0 | . |
Pacific Islander | 384 | 310 | 80.7 | 74 | 19.3 | 0 | . | . | 0 | . |
American Indian | 196 | 161 | 82.1 | 35 | 17.9 | 8 | 5 | 62.5 | 3 | 37.5 |
| ||||||||||
Education | ||||||||||
Less than high school | 5305 | 4490 | 84.6 | 815 | 15.4 | 119 | 84 | 70.6 | 35 | 29.4 |
High school grad | 16,692 | 14,732 | 88.3 | 1960 | 11.7 | 304 | 248 | 81.6 | 56 | 18.4 |
Post HS/some college | 25,109 | 22,248 | 88.6 | 2861 | 11.4 | 516 | 423 | 82.0 | 93 | 18.0 |
College grad/postgrad | 25,930 | 23,773 | 91.7 | 2157 | 8.3 | 486 | 421 | 86.6 | 65 | 13.4 |
| ||||||||||
BMI (kg/m^2) | ||||||||||
0 – 18.5 | 552 | 439 | 79.5 | 113 | 20.5 | 10 | 8 | 80.0 | 2 | 20.0 |
18.6 – 25.0 | 23,950 | 21,273 | 88.8 | 2677 | 11.2 | 467 | 377 | 80.7 | 90 | 19.3 |
25.1 – 30.0 | 30,792 | 27,771 | 90.2 | 3021 | 9.8 | 613 | 527 | 86.0 | 86 | 14.0 |
> 30 | 17,742 | 15,760 | 88.8 | 1982 | 11.2 | 335 | 264 | 78.8 | 71 | 21.2 |
| ||||||||||
Comorbidity score | ||||||||||
No | 52,994 | 47,879 | 90.3 | 5115 | 9.7 | 1003 | 847 | 84.4 | 156 | 15.6 |
Yes | 20,042 | 17,364 | 86.6 | 2678 | 13.4 | 422 | 329 | 78.0 | 93 | 22.0 |
| ||||||||||
Cigarette smoking status | ||||||||||
Never smoker | 34,207 | 30,799 | 90.0 | 3408 | 10.0 | 699 | 592 | 84.7 | 107 | 15.3 |
Current smoker | 7792 | 6622 | 85.0 | 1170 | 15.0 | 153 | 115 | 75.2 | 38 | 24.8 |
Former smoker | 31,037 | 27,822 | 89.6 | 3215 | 10.4 | 573 | 469 | 81.8 | 104 | 18.2 |
| ||||||||||
Marital status | ||||||||||
Married/living as married | 55,450 | 50,205 | 90.5 | 5245 | 9.5 | 1082 | 895 | 82.7 | 187 | 17.3 |
Formerly married | 15,111 | 12,897 | 85.3 | 2214 | 14.7 | 303 | 251 | 82.8 | 52 | 17.2 |
Never married | 2475 | 2141 | 86.5 | 334 | 13.5 | 40 | 30 | 75.0 | 10 | 25.0 |
| ||||||||||
Current occupation | ||||||||||
Homemaker | 8303 | 7252 | 87.3 | 1051 | 12.7 | 177 | 133 | 75.1 | 44 | 24.9 |
Working | 29,381 | 26,438 | 90.0 | 2943 | 10.0 | 527 | 442 | 83.9 | 85 | 16.1 |
Unemployed | 757 | 643 | 84.9 | 114 | 15.1 | 11 | 9 | 81.8 | 2 | 18.2 |
Retired | 31,326 | 28,136 | 89.8 | 3190 | 10.2 | 641 | 542 | 84.6 | 99 | 15.4 |
Disabled /extended sick leave | 1704 | 1360 | 79.8 | 344 | 20.2 | 32 | 21 | 65.6 | 11 | 34.4 |
Other | 1565 | 1414 | 90.4 | 151 | 9.6 | 37 | 29 | 78.4 | 8 | 21.6 |
| ||||||||||
Family history of PLCO cancer | ||||||||||
Yes | 20,470 | 18,530 | 90.5 | 1940 | 9.5 | 393 | 326 | 83.0 | 67 | 17.0 |
No | 50,808 | 45,177 | 88.9 | 5631 | 11.1 | 991 | 819 | 82.6 | 172 | 17.4 |
Possibly | 1758 | 1536 | 87.4 | 222 | 12.6 | 41 | 31 | 75.6 | 10 | 24.4 |
| ||||||||||
Screening center** | ||||||||||
Colorado | 6194 | 5351 | 86.4 | 843 | 13.6 | 480 | 411 | 85.6 | 69 | 14.4 |
Georgetown | 3470 | 3335 | 96.1 | 135 | 3.9 | 0 | 0 | . | 0 | . |
Pacific Health | 4992 | 4271 | 85.6 | 721 | 14.4 | 0 | 0 | . | 0 | . |
Henry Ford | 11,585 | 8573 | 74.0 | 3012 | 26.0 | 448 | 311 | 69.4 | 137 | 30.6 |
Minnesota | 12,786 | 12,512 | 97.9 | 274 | 2.1 | 0 | 0 | . | 0 | . |
Washington University | 7333 | 6371 | 86.9 | 962 | 13.1 | 0 | 0 | . | 0 | . |
Pittsburgh | 8405 | 7855 | 93.5 | 550 | 6.5 | 0 | 0 | . | 0 | . |
Utah | 7144 | 6526 | 91.3 | 618 | 8.7 | 497 | 454 | 91.3 | 43 | 8.7 |
Marshfield | 8074 | 7454 | 92.3 | 620 | 7.7 | 0 | 0 | . | 0 | . |
Alabama | 3053 | 2995 | 98.1 | 58 | 1.9 | 0 | 0 | . | 0 | . |
| ||||||||||
Consent type | ||||||||||
Single | 65,971 | 60,210 | 91.3 | 5761 | 8.7 | 1075 | 954 | 88.7 | 121 | 11.3 |
Dual | 7065 | 5033 | 71.2 | 2032 | 28.8 | 350 | 222 | 63.4 | 128 | 36.6 |
| ||||||||||
Randomization year | ||||||||||
1993–1994 | 7950 | 7244 | 91.1 | 706 | 8.9 | 221 | 190 | 86.0 | 31 | 14.0 |
1995–1996 | 24,563 | 21,572 | 87.8 | 2991 | 12.2 | 541 | 408 | 75.4 | 133 | 24.6 |
1997–1998 | 22,330 | 19,865 | 89.0 | 2465 | 11.0 | 397 | 346 | 87.2 | 51 | 12.8 |
1999–2000 | 16,855 | 15,382 | 91.3 | 1473 | 8.7 | 248 | 216 | 87.1 | 32 | 12.9 |
2001 | 1338 | 1180 | 88.2 | 158 | 11.8 | 18 | 16 | 88.9 | 2 | 11.1 |
| ||||||||||
Distance from screening center | ||||||||||
Tertile 1 | N/A | N/A | N/A | N/A | N/A | 478 | 410 | 85.8 | 68 | 14.2 |
Tertile 2 | 476 | 391 | 82.1 | 85 | 17.9 | |||||
Tertile 3 | 471 | 375 | 79.6 | 96 | 20.4 |
Excludes participants who withdrew or died between randomization and the first screening visit.
Complete names and locations of screening centers can be found in the Funding section of this manuscript.
Unadjusted odds ratios for consent type, race, smoking status, and occupation were attenuated after adjustment for all variables in multivariate logistic regression models. In some instances, adjustment resulted in a change from a statistically significant association to one that was not. An example is the odds of non-compliance for Black, non-Hispanic participants: the unadjusted odds ratio was 1.6 (95% CI: 1.5–1.8) and the adjusted odds ratio was 0.9 (95% CI: 0.8–1.0). (Table 2)
Table 2.
% Non-compliant | Unadjusted* OR (95% CI) | Adjusted*,** OR (95% CI) | |
---|---|---|---|
| |||
Age at randomization (years) | |||
Younger than 50 | 10.3 | Reference | Reference |
60–64 | 10.1 | 1.0 (0.9–1.0) | 1.0 (0.9–1.1) |
65–59 | 10.9 | 1.1 (1.0–1.1) | 1.1 (1.0–1.2) |
70 or older | 12.6 | 1.3 (1.2–1.4) | 1.3 (1.2–1.4) |
| |||
Gender | |||
Female | 12.9 | Reference | Reference |
Male | 8.3 | 0.6 (0.6–0.6) | 0.7 (0.7–0.8) |
| |||
Race | |||
White, non-Hispanic | 10.2 | Reference | Reference |
Black, non-Hispanic | 15.6 | 1.6 (1.5,1.8) | 0.9 (0.8–1.0) |
Hispanic | 15.0 | 1.6 (1.3–1.8) | 1.1 (0.9–1.3) |
Asian | 11.8 | 1.2 (1.0–1.3) | 0.8 (0.6–0.9) |
Pacific Islander | 19.3 | 2.1 (1.6–2.7) | 1.2 (0.9–1.6) |
American Indian | 17.9 | 1.9 (1.3–2.8) | 1.2 (0.8–1.8) |
| |||
Education | |||
Less than high school | 15.4 | 1.4 (1.2–1.5) | 1.2 (1.1–1.3) |
High school grad | 11.7 | Reference | Reference |
Post HS/some college | 11.4 | 1.0 (0.9–1.0) | 1.0 (0.9–1.0) |
College grad/postgrad | 8.3 | 0.7 (0.6–0.7) | 0.8 (0.7–0.8) |
| |||
BMI (kg/mˆ2) | |||
0 – 18.5 | 20.5 | 2.0 (1.7–2.5) | 1.6 (1.3–2.0) |
18.6 – 25.0 | 11.2 | Reference | Reference |
25.1 – 30.0 | 9.8 | 0.9 (0.8–0.9) | 1.0 (0.9–1.0) |
> 30 | 11.2 | 1.0 (0.9–1.1) | 1.0 (0.9–1.0) |
| |||
Comorbidity score | |||
No | 9.7 | Reference | Reference |
Yes | 13.4 | 1.4 (1.4–1.5) | 1.3 (1.2–1.4) |
| |||
Cigarette smoking status | |||
Never smoker | 10.0 | Reference | Reference |
Current smoker | 15.0 | 1.6 (1.5–1.7) | 1.3 (1.2–1.4) |
Former smoker | 10.4 | 1.0 (1.0–1.1) | 1.1 (1.0–1.1) |
| |||
Marital status | |||
Married/living as married | 9.5 | Reference | Reference |
Formerly married | 14.7 | 1.6 (1.6–1.7) | 1.3 (1.2–1.3) |
Never married | 13.5 | 1.5 (1.3–1.7) | 1.3 (1.2–1.5) |
| |||
Current occupation | |||
Working | 10.0 | Reference | Reference |
Homemaker | 12.7 | 1.3 (1.2–1.4) | 1.0 (0.9–1.1) |
Unemployed | 15.1 | 1.6 (1.3–2.0) | 1.3 (1.1–1.7) |
Retired | 10.2 | 1.0 (1.0–1.1) | 0.9 (0.9–1.0) |
Disabled/extended sick leave | 20.2 | 2.3 (2.0–2.6) | 1.5 (1.3–1.8) |
Other | 9.6 | 1.0 (0.8–1.1) | 0.9 (0.8–1.1) |
| |||
Family history of a PLCO cancer | |||
No | 11.1 | Reference | Reference |
Yes | 9.5 | 0.8 (0.8–0.9) | 0.8 (0.8–0.9) |
Possibly | 12.6 | 1.2 (1.0–1.3) | 1.1 (1.0–1.3) |
| |||
Screening center*** | |||
Colorado | 13.6 | Reference | Reference |
Georgetown | 3.9 | 0.3 (0.2–0.3) | 0.3 (0.2–0.3) |
Hawaii | 14.4 | 1.1 (1.0–1.2) | 1.0 (0.9–1.2) |
Henry Ford | 26.0 | 2.2 (2.1–2.4) | 1.4 (1.3–1.5) |
Minnesota | 2.1 | 0.1 (0.1–0.2) | 0.1 (0.1–0.2) |
Washington | 13.1 | 1.0 (0.9–1.1) | 0.8 (0.7–0.9) |
Pittsburgh | 6.5 | 0.4 (0.4–0.5) | 0.4 (0.4–0.5) |
Utah | 8.7 | 0.6 (0.5–0.7) | 0.6 (0.5–0.6) |
Marshfield | 7.7 | 0.5 (0.5–0.6) | 0.5 (0.4–0.5) |
Alabama | 1.9 | 0.1 (0.1–0.2) | 0.1 (0.1–0.1) |
| |||
Consent type | |||
Single | 8.7 | Reference | Reference |
Dual | 28.8 | 4.2 (4.0–4.5) | 2.2 (2.0–2.5) |
| |||
Randomization year | |||
1993–1994 | 8.9 | 0.7 (0.6–0.8) | 0.4 (0.4–0.5) |
1995–1996 | 12.2 | Reference | Reference |
1997–1998 | 11.0 | 0.9 (0.8–0.9) | 0.9 (0.9–1.0) |
1999–2000 | 8.7 | 0.7 (0.6–0.7) | 0.8 (0.8–0.9) |
2001 | 11.8 | 1.0 (0.8–1.1) | 0.9 (0.7–1.0) |
Adjusted for all covariates listed in table
Complete names and locations of screening centers can be found in the Funding section of this manuscript.
A number of adjusted odds ratios were significant but indicated modest increases or decreases in risk of non-compliance. A notable exception was use of the dual consent method (OR: 2.2, 95% CI: 2.0–2.5), relative to the single consent method. Significant odds ratios of 1.5 or greater also were observed for those with a BMI of 18.5 or less, relative to those who had a BMI of 18.6–25.0 (OR: 1.6, 95% CI: 1.3–2.0), and those who were disabled or on extended sick leave, relative to those who were working (OR: 1.5, 95% CI: 1.3–1.7). Significant odds ratios less than 1 were observed for Asian participants relative to White, non-Hispanic participants (OR: 0.8, 95% CI: 0.6–0.9), persons with college degrees or post-graduate training relative to those whose highest attained education was high school (OR: 0.8, 95% CI: 0.7–0.8), males (OR: 0.7, 95% CI: 0.7–0.8), and participants randomized in 1993 or 1994 (OR: 0.4, 95% CI: 0.4–0.5) and 1999 or 2000 (OR: 0.8, 95% CI: 0.8–0.9), relative to participants randomized in 1995 or 1996. Many screening centers had significant odds ratios less than 1, as the University of Colorado Anschutz Medical Campus screening center was used as the reference category and had one of the higher rates of non-compliance. (Table 2)
Distance subset
Of the 1500 randomly selected participants for distance analyses, 1425 (95%) had complete covariate information and were included. Non-compliance in the distance subset was 18%, which was the same as non-compliance among the totality of participants from those three centers. Patterns of non-compliance were similar in the subset to those in the entire cohort, although rate of non-compliance for the dual consent method was 37% (OR: 7.4, 95% CI: 3.0–18.3). The non-compliance percentages for the tertiles of distance from the screening center were 14% (nearest), 18% (midrange), and 20% (furthest), respectively. The odds ratios for the midrange and far categories, relative to the bottom tertile, were 1.2 and 1.4, respectively, but neither was significantly different from 1.0 (95% CIs: 0.8–1.8 and 1.0–2.1, respectively).
DISCUSSION
Compliance with PLCO’s first screening visit was nearly 90%. Few factors impacted non-compliance in a more-than-modest manner. Our strongest finding, that a dual-step consent process increased the odds of non-compliance more than two-fold, suggests that willingness to participate in observational research does not guarantee willingness to participate in interventional research, and that processes that consent participants for activities after they are randomized should be carefully considered before use.
The UK flexible sigmoidoscopy trial, which examined the association of compliance and race, observed that relative to whites, blacks were more likely to be compliant and Asians less likely to be compliant, but odds ratios were not significantly different from 1 [5]. These patterns are opposite those in the PLCO cohort. Given the minimal data available for compliance in screening RCTs, we compared our findings to RCTs of cancer chemoprevention. In the CARET trial, a trial of lung cancer conducted in above-average risk males and females, patterns of non-compliance by age and gender were comparable to those seen in PLCO [3]. BCPT, a trial of women at above-average risk of breast cancer, observed patterns similar to those of PLCO for age, smoking status, and education [2]. In the WHI supplemental calcium and vitamin D trial, a study of healthy women that examined multiple endpoints including breast and colorectal cancer, patterns of non-compliance for marital and smoking status were similar to PLCO in terms of direction, but patterns of non-compliance for race (in particular, for African-Americans) and age were not [4]. In the aforementioned chemoprevention studies, magnitude of associations was typically modest, as in our data. The fact that in our data certain unadjusted odds ratios for variables typically thought to be associated with non-compliance, like smoking and race, were attenuated after adjustment indicates that it may be misleading to examine and draw conclusions about these variables in other studies without consideration of factors that correlate with them.
The expected benefit of the dual-consent method was to keep contamination low, but low rates of participation among those randomized to the intervention arm were higher than expected. Therefore, the centers that began with the dual-consent method ultimately changed to the single-consent method. There are two lessons to be learned: that participation in an observational study should not be taken to mean willingness to participate in interventional research, and that it is best for to consent participants for interventional activities, especially those that are invasive, prior to randomization so that non-participation, including non-compliance, is minimized and study power can be maintained. If a situation arises in which a “randomize-before-consent” approach must be used, researchers and study coordinators must go to extra efforts to ensure that trial integrity is not compromised due to non-compliance. The fact that the strength of the association of the dual-consent method and non-compliance was meaningfully attenuated after adjustment suggests that targeted approaches to decrease non-compliance in such situations might be effective.
Our restriction to compliance with the first screen limits the generalizability of our results. Risk of non-compliance is likely to change over time as life events, such as aging, disease development, and change in employment, occur. Our findings, therefore, only are relevant to non-compliance soon after trial enrollment. In addition, the BQF was not designed to capture reasons for non-compliance, so we have no data on factors that are certain to impact non-compliance, such as mobility difficulties and other impediments to traveling to a screening center. We would like to note, however, that in our experience the most important factor is the relationship between participants and study staff, something that is multi-dimensional and thus difficult, if not impossible, to measure.
Strengths of our study include a large sample size, one that allowed us to examine odds of non-compliance for some covariate levels that typically have low prevalence, such as low levels of education. We also were able to include many covariates in multivariate models to account for potential confounding.
CONCLUSIONS
In PLCO, a large multi-phasic cancer screening RCT, many factors significantly influenced non-compliance with the first round of exams, including BMI, employment status, and race, but did so only modestly. A process that consented intervention arm participants for screening exams after randomization increased odds of non-compliance two-fold, suggesting that use of this method should be carefully considered before its implementation.
Table 3.
% Non-compliant | Unadjusted* OR (CI) | Adjusted*,** OR (CI) | |
---|---|---|---|
| |||
Age at randomization (years) | |||
Younger than 50 | 17.9 | Reference | Reference |
60–64 | 15.5 | 0.8 (0.6–1.2) | 1.0 (0.7–1.5) |
65–59 | 16.5 | 0.9 (0.6–1.3) | 1.4 (0.8–2.2) |
70 or older | 22.3 | 1.3 (0.9–2.0) | 1.7 (1.0–2.9) |
| |||
Gender | |||
Female | 19.7 | Reference | Reference |
Male | 14.9 | 0.7 (0.5–0.9) | 0.8 (0.6–1.2) |
| |||
Race | |||
White- non-Hispanic | 17.2 | Reference | Reference |
Black, non-Hispanic | 22.9 | 1.4 (0.9–2.3) | 0.7 (0.4–1.2) |
Hispanic | 12.9 | 0.7 (0.3–1.4) | 0.6 (0.3–1.3) |
Asian | 0.0 | – | – |
Pacific Islander | 0.0 | – | – |
American Indian | 37.5 | 2.9 (0.7–12.1) | 1.8 (0.4–8.8) |
| |||
Education | |||
Less than high school | 29.4 | 1.8 (1.1–3.0) | 1.4 (0.8–2.4) |
High school grad | 18.4 | Reference | Reference |
Post HS/some college | 18.0 | 1.0 (0.7–1.4) | 1.1 (0.8–1.7) |
College grad/postgrad | 13.4 | 0.7 (0.5–1.0) | 0.9 (0.6–1.4) |
| |||
BMI (kg/mˆ2) | |||
0 – 18.5 | 20.0 | 1.0 (0.2–5.0) | 0.9 (0.2–5.7) |
18.6 – 25.0 | 19.3 | Reference | Reference |
25.1 – 30.0 | 14.0 | 0.7 (0.5–0.9) | 0.7 (0.5–1.0) |
> 30.0 | 21.2 | 1.1 (0.8–1.6) | 1.0 (0.7–1.4) |
| |||
Comorbidity score | |||
No | 15.6 | Reference | Reference |
Yes | 22.0 | 1.5 (1.2–2.0) | 1.3 (0.9–1.8) |
| |||
Cigarette smoking status | |||
Never smoker | 15.3 | Reference | Reference |
Current smoker | 24.8 | 1.8 (1.2–2.8) | 1.4 (0.9–2.4) |
Former smoker | 18.2 | 1.2 (0.9–1.6) | 1.0 (0.7–1.4) |
Marital status | |||
Married/living as married | 17.3 | Reference | Reference |
Formerly married | 17.2 | 1.0 (0.7–1.4) | 0.9 (0.6–1.3) |
Never married | 25.0 | 1.6 (0.8–3.3) | 1.6 (0.7–3.6) |
Current occupation | |||
Working | 16.1 | Reference | Reference |
Homemaker | 24.9 | 1.7 (1.1–2.6) | 1.3 (0.7–1.9) |
Unemployed | 18.2 | 1.2 (0.2–5.4) | 0.9 (0.2–4.7) |
Retired | 15.4 | 1.0 (0.7–1.3) | 0.7 (0.5–1.0) |
Disabled/extended sick leave | 34.4 | 2.7 (1.3–5.9) | 1.7 (0.7–3.9) |
Other | 21.6 | 1.4 (0.6–3.2) | 1.0 (0.4–2.6) |
| |||
Family history of a PLCO cancer | |||
No | 17.4 | Reference | Reference |
Yes | 17.0 | 1.0 (0.7–1.3) | 1.1 (0.8–1.5) |
Possibly | 24.4 | 1.5 (0.7–3.2) | 1.8 (0.8–3.9) |
| |||
Screening center*** | |||
Colorado | 14.4 | Reference | Reference |
Henry Ford | 30.6 | 2.6 (1.9–3.6) | 0.5 (0.2–1.2) |
Utah | 8.7 | 0.6 (0.4–0.8) | 0.5 (0.3–0.8) |
| |||
Consent type | |||
Single | 11.3 | Reference | Reference |
Dual | 36.6 | 4.5 (3.4–6.1) | 7.4 (3.0–18.3) |
| |||
Randomization year | |||
1993–1994 | 14.0 | 0.5 (0.3–0.8) | 0.5 (0.3–0.8) |
1995–1996 | 24.6 | Reference | Reference |
1997–1998 | 12.8 | 0.5 (0.3–0.6) | 1.0 (0.7–1.7) |
1999–2000 | 12.9 | 0.5 (0.3–0.7) | 1.2 (0.7–2.1) |
2001 | 11.1 | 0.4 (0.1–1.7) | 1.4 (0.2–7.9) |
| |||
Distance | |||
Tertile 1 | 14.2 | Reference | Reference |
Tertile 2 | 17.9 | 1.3 (0.9–1.9) | 1.2 (0.8–1.8) |
Tertile 3 | 20.4 | 1.5 (1.1–2.2) | 1.4 (1.0–2.1) |
Bolding indicates statistically significant ORs.
Adjusted for all covariates listed in table.
Complete names and locations of screening centers can be found in the Funding section of this manuscript.
HIGHLIGHTS.
Many factors modestly affected non-compliance with the first screening exam visit
The strongest predictor was method of consent
Consent after randomization significantly increased non-compliance two-fold.
These results are relevant for RCTs of healthy persons ages 55–74
Acknowledgments
The authors thank the PLCO participants and the hundreds of individuals who worked on the trial throughout its course. The authors also wish to acknowledge our colleague and friend, Eduard Gamito (deceased). Ed was the Recruitment/Retention Coordinator at the University of Colorado PLCO site. He conceived the idea for this project, directed the initial work and collaborated with the authors. His inspiration and contributions made this paper possible. We dedicate it to his memory.
FUNDING: Supported by contracts from the Division of Cancer Prevention, National Cancer Institute, to the coordinating center (N01-CN-25476 to Westat, Inc.), the ten screening centers (N01-CN-25514 to the University of Colorado Anschutz Medical Campus (Denver Metro area); N01-CN-25522 to Georgetown University Medical Center (Washington, DC metro area); N01-CN-25515 to Pacific Health Research and Education Institute (Honolulu, HI); N01-CN-25512 to Henry Ford Health System (Detroit, MI); N01-CN-25513 to University of Minnesota School of Public Health (Minneapolis/St. Paul, MN); N01-CN-25516 to Washington University School of Medicine (St. Louis, MO); N01-CN-25511 to University of Pittsburg Medical Center (Pittsburgh, PA) ; N01-CN-25524 to the University of Utah Health Sciences Center (Salt Lake City, UT); N01-CN-25518 to Marshfield Clinic Research Foundation (Marshfield, WI); N01-CN-75022 to University of Alabama at Birmingham (Birmingham, AL); and N02-CN-55203-76 and N02-CN-35001-45 to Information Management Services, Inc. (Rockville, MD).
ABBREVIATIONS
- PLCO
Prostate, Lung, Colorectal and Ovarian Cancer Screening Trial
- BQF
Baseline Questionnaire Form
Footnotes
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Results have been published from the French component of the European Randomized Study of Prostate Cancer Screening [6], but are not available in English, and thus are not considered.
CONFLICT OF INTEREST: The authors have no conflicting interests, financial or otherwise.
AUTHOR CONTRIBUTIONS: SLO, LHG, JCC, SMP, LEL were leaders in data collection at PLCO screening centers; along with KW, they identified this research topic and led discussions concerning issues that were important in data analysis. JM, BT, and TR managed data and conducted statistical analysis. PMM provided guidance on research question development and statistical analysis. PMM and SLO led the drafting of this manuscript. HMR created the tables for the manuscript. All authors read and approved the final manuscript.
Contributor Information
Pamela M. Marcus, Email: marcusp@mail.nih.gov.
Sheryl L. Ogden, Email: Sheryl.ogden@ucdenver.edu.
Lisa H. Gren, Email: Lisa.gren@hsc.utah.edu.
Jeffery C. Childs, Email: jeff.childs@hsc.utah.edu.
Shannon M Pretzel, Email: Shannon.pretzel@ucdenver.edu.
Lois E. Lamerato, Email: llamera1@hfhs.org.
Kayo Walsh, Email: walsh@hcp.med.harvard.edu.
Heather M. Rozjabek, Email: heather.rozjabek@gmail.com.
Jerome Mabie, Email: mabiej@imsweb.com.
Brett Thomas, Email: thomasb@imsweb.com.
Tom Riley, Email: rileyt@imsweb.com.
References
- 1.Land SR, Cronin WM, Wickerham DL, et al. Cigarette smoking, obesity, physical activity, and alcohol use as predictors of chemoprevention adherence in the National Surgical Adjuvant Breast and Bowel Project P-1 Breast Cancer Prevention Trial. Cancer Prev Res. 2011;4:1393–400. doi: 10.1158/1940-6207.CAPR-11-0172. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Cronin WM, Cecchini RS, Wickerham D, Vogel VG, Costantino JP, Wolmark N. Factors associated with participant adherence in the NSABP study of tamoxifen and raloxifine (STAR). [abstract]. J Clin Oncol; American Society of Clinical Oncology (ASCO) Annual Meeting; 2008; 2008. May 20, abstr 6518. [Google Scholar]
- 3.Thornquist MD, Patrick DL, Omenn GS. Participation and adherence among older men and women recruited to the Beta-Carotene and Retinol Efficacy Trial. Gerentologist. 1991;5:593–597. doi: 10.1093/geront/31.5.593. [DOI] [PubMed] [Google Scholar]
- 4.Brunner R, Dunbar-Jacob J, Leboff MS, et al. Predictors of adherence in the Women’s Health Initiative calcium and vitamin D trial. Behav Med. 2009;34:145–55. doi: 10.3200/BMED.34.4.145-155. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Robb KA, Power E, Atkin W, Wardle J. Ethnic differences in participation in flexible sigmoidoscopy screening in the UK. J Med Screen. 2008;15:130–6. doi: 10.1258/jms.2008.007112. [DOI] [PubMed] [Google Scholar]
- 6.Jegu J, Tretarre B, Grosclaude P, et al. Results and participation factors to the European Randomized study of Screening for Prostate Cancer (ERSPC) with Prostate Specific Antigen: French departments of Tarn and Hérault. Prog Urol. 2009;17:487–8. doi: 10.1016/j.purol.2009.03.001. [French] [DOI] [PubMed] [Google Scholar]
- 7.Prorok PC, Andriole GL, Bresalier RS, et al. Design of the Prostate, Lung, Colorectal and Ovarian (PLCO) Cancer Screening Trial. Control Clin Trials. 2000;21:273S–309S. doi: 10.1016/s0197-2456(00)00098-2. [DOI] [PubMed] [Google Scholar]
- 8.Buys SS, Partridge E, Black A, Johnson CC, Lamerato L, Isaacs C, et al. Effect of screening on ovarian cancer mortality: the Prostate, Lung, Colorectal and Ovarian (PLCO) Cancer Screening Randomized Controlled Trial. JAMA. 2011;305:2295–303. doi: 10.1001/jama.2011.766. [DOI] [PubMed] [Google Scholar]
- 9.Oken MM, Hocking WG, Kvale PA, Andriole GL, Buys SS, Church TR, et al. Screening by chest radiograph and lung cancer mortality: the Prostate, Lung, Colorectal, and Ovarian (PLCO) randomized trial. JAMA. 2011;306:1865–73. doi: 10.1001/jama.2011.1591. [DOI] [PubMed] [Google Scholar]
- 10.Andriole GL, Crawford ED, Grubb RL, 3rd, Buys SS, Chia D, Church TR, et al. Prostate cancer screening in the randomized Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial: mortality results after 13 years of follow-up. J Natl Cancer Inst. 2012;104:125–32. doi: 10.1093/jnci/djr500. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Schoen RE, Pinsky PF, Weissfeld JL, Yokochi LA, Church T, Laiyemo AO, et al. Colorectal-cancer incidence and mortality with screening flexible sigmoidoscopy. N Engl J Med. 2012;366:2345–57. doi: 10.1056/NEJMoa1114635. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Gren L, Broski K, Childs J, et al. Recruitment methods employed in the Prostate, Lung, Colorectal and Ovarian Cancer Screening Trial. Clin Trials. 2009;6:52–9. doi: 10.1177/1740774508100974. [DOI] [PMC free article] [PubMed] [Google Scholar]