Abstract
Background
Multivitamin use is a common health behavior but there is conflicting evidence from prospective studies about whether this behavior increases or decreases prostate cancer (PCa) risk.
Methods
Associations of multivitamin (MVI) use and PCa risk were evaluated using data from the Selenium and Vitamin E Cancer Prevention Trial (SELECT). Cox proportional hazards models estimated associations of MVI use with risk of total, low- and high-grade PCa. Longitudinal data were used to evaluate screening and biopsy patterns. To account for differential biopsy patterns, the probability of PCa was estimated for men with a positive screening value but no biopsy. Incidence Density Ratios were used to approximate hazards ratios, and associations of MVI use with predicted PCa risk were compared to observed.
Results
Analyses of data from observed biopsies suggest a respective 19% (95% CI 10–28%) and 21% (12–31%) higher risk of high-grade PCa for current and long-term MVI use, compared to no use. Current and long-term MVI use was associated with a shorter time to first on-study biopsy, indicating the potential for detection bias. After accounting for differential acceptance of biopsy, associations of MVI use with PCa were attenuated and not statistically significant.
Conclusions
In SELECT, biopsy acceptance patterns differed by MVI use. Estimates of associations of MVI use with PCa risk based on observed biopsy data may be biased by differential acceptance of biopsy.
Impact
Differential biopsy ascertainment may impact associations of risk factors and PCa. Detailed screening and biopsy data can be used to analytically minimize such bias.
Keywords: Prostate Cancer, multivitamin, dietary supplements, prevention
INTRODUCTION
Prostate cancer (PCa) is the most common cancer diagnosis in men today and the second leading cause of cancer-related mortality in men, with an estimated 268,490 new cases diagnosed and 34,500 deaths (the second leading cause of cancer death) in 2022.(1) Other than age, race, genetics and family history, which are non-modifiable, few risk factors for PCa have been established. Diet, in particular nutrient intake, is a modifiable risk factor that has been suggested to play a role in the pathogenesis of PCa.(2) Vitamin and mineral supplements are an important source of micronutrient intakes in the United States.(3,4) In a recent national survey, 52% of adults overall and 45% of men reported use of a dietary supplement in the preceding 30 days.(5) Despite the popularity of dietary supplements, scientific support of their efficacy for disease prevention is limited, and a growing number of clinical trials suggest they may have harmful health effects.(6–9)
Multivitamin and mineral (MVI) supplements, which are comprised of a large number of vitamins and minerals in smaller doses than individual supplements, are the most commonly used dietary supplements in the United States.(5) Several studies have examined the relationship between multivitamin use and risk of PCa. Results from prospective observational studies are inconsistent, with some studies reporting an increased risk of total(10) or aggressive PCa(11) or prostate cancer death(11,12), while others report no association(13). In addition, the only large randomized trial to date found no overall effect of daily multivitamin supplementation on prostate cancer risk or death.(14) Inconsistencies between the randomized trial and observational studies may be due to screening-related biases. Because prostate cancer is highly-prevalent in aging men and is typically asymptomatic, any given risk factor or behavior that increases the likelihood of a biopsy being recommended to or accepted by a man will appear to be associated with the risk of prostate cancer, even if no true association exists.
Here we evaluate the association of multivitamin use with PCa risk using data from the Selenium and Vitamin E Cancer Prevention Trial (SELECT) in a setting with relatively uniform access to screening and health care during the trial. In addition, we use longitudinal PCa screening data to explore potential differential biopsy adoption patterns that could affect the association between MVI use and PCa. We hypothesize that men who use multivitamins have other health-seeking behaviors including higher rates of biopsy or earlier acceptance of biopsy compared to non-users, and that differential biopsy acceptance patterns may impact associations of MVI use and PCa risk.
MATERIALS AND METHODS
Participants/Study Design.
Data are from the Selenium and Vitamin E Cancer Prevention Trial (SELECT), a multicenter randomized, double-blind, placebo-controlled chemoprevention SWOG trial testing whether daily supplementation with selenium, vitamin E, or both would reduce PCa incidence.(15,16) Between July 2001 and May 2004, 35,533 men 55 years or older (>=50 years for African American men) with a baseline prostate-specific antigen (PSA) of ≤4 ng/mL and normal digital rectal examination (DRE) were enrolled at 427 sites in the United States, Canada and Puerto Rico. Men were randomized into 1 of 4 groups: selenium (200 μg/d from L-selenomethionine) with matching vitamin E placebo, vitamin E (400 IU/d of all rac -α-tocopherol acetate) with matching selenium placebo, both agents, or matching placebo. All men provided written informed consent prior to participation, and study procedures were approved by the local institutional review board for each study site. On September 15, 2008, the Data and Safety Monitoring Committee recommended discontinuation of trial supplements due to a prespecified formal futility analysis, and the use of study supplements ended on October 23, 2008.(16)
Data Collection.
At baseline, participants reported information on race/ethnicity, education, history of diabetes, prior biopsy history and family history of PCa, while clinic staff measured height and weight. Body mass index (BMI) was calculated as weight (kg)/ height (m)2. At follow-up study visits scheduled every 6 months after randomization, men reported new medical events and receipt of PCa screening. Participants underwent PSA and DRE screening and received biopsy recommendations according to the local standard of care and participant’s preference.
At baseline, 10-year history of dietary supplement use was captured using a 7-page self-administered questionnaire that assessed total intake of 10 vitamins, 6 minerals, and 20 botanicals/herbals from all types of supplements including multivitamin and mineral supplements.(17) Reliability and validity of self-reported supplement use for this questionnaire have been described previously.(18) Participants who reported regular use of a MVI (at least once a week for a year during the previous 10 years) were also asked to report the number days per week (1–2, 3–4, 5–6, or 7) and number of years taken in the past 10 years (1–3, 4–6, 7–9, or 10). For participants who reported regular use of MVI at baseline but were missing information on the number of years taken in the past 10 years (n=16, 0.05%), these data were imputed using the most common responses given by participants (7 days per week and 2 years taken in past 10 years, respectively). For analyses, MVI use at baseline was parameterized by recency (never vs. former vs. current) and duration (never vs. <5 years vs. 5+ years prior to trial) of use.
PCa status was determined by routine clinical management and confirmed by central pathology review. At each 6-month study visit, PCa status was self-reported, after which medical records were obtained and clinical stage and diagnostic method were abstracted. For cases (n=1,857), Gleason scores provided by local pathologists were used to assign grade. Low- and high-grade tumors were defined by Gleason score of 2 to 6 and 7 to 10, respectively. Grade was undetermined for 203 cases. Prostate cancers diagnosed after the end of study supplementation were censored on October 23, 2008.
Statistical Analysis.
Descriptive statistics were used to evaluate differences in the distributions of demographic and health related characteristics, screening and biopsy patterns between men who were diagnosed with PCa and those who were not, and between men who reported historical MVI use, and those who did not. Cox proportional hazards models were used to estimate hazards ratios (HRs) and 95% confidence intervals (CIs) for the associations between MVI use and PCa risk. Separate models were run for total cancer, and for low-grade and high-grade PCa. In models for low-grade cancer, men with high-grade cancer were censored at the time of diagnosis and vice versa in models of high-grade cancer.
Covariate adjustment was based on a finite list of pre-specified candidates that include well-established PCa risk factors and factors potentially relevant to both MVI use and PCa. All models were adjusted for age at enrollment (continuous), self-reported race (non-Hispanic white, non-Hispanic black, other), family history of PCa at enrollment (yes, no), body mass index at enrollment (<25, 25 to <30, 30+ kg/m2), history of diabetes at enrollment (yes, no/unknown), and study treatment arm assignment (vitamin E, Selenium, both, placebo), as well as the following time-dependent covariates: PSA (continuous), DRE (abnormal, normal) and on-study negative biopsies. An additional model included adjustment for education, smoking status, alcohol. All analyses are based on the time between study entry and either PCa or censor event (last study contact, death or October 23, 2008 (end of trial supplementation)).
To assess potential differential biopsy bias, we first evaluated whether there were differences in screening and biopsy patterns or in the timing of biopsy by MVI use. Descriptive statistics were used to evaluate differences in screening and biopsy patterns. To evaluate potential differences in biopsy timing by MVI use, we used a Cox proportional hazards model, adjusted for age, family history of PCa, and time-dependent longitudinal PSA and DRE, to evaluate time to first on-study biopsy (regardless of the biopsy outcome). For this analysis, men without a biopsy were censored at the date of last study contact, death or October 23, 2008 (end of trial supplementation), whichever came first.
To evaluate the association of MVI use with PCa accounting for differential biopsy patterns, we first used a previously developed approach to mimic uniform biopsy recommendations and acceptance (19) in which Incidence Density Ratios (IDR) were used to approximate a hazards ratio. In step 1 of this approach, for men with screening values suggesting PCa (i.e., PSA>4 ng/ml or suspicious DRE) who did not have a biopsy, we used a man’s estimated probability of prostate cancer to account for lack of biopsy. The Prostate Cancer Prevention Trial Risk Calculators(20,21) were used to estimate probabilities of total PCa and of high-grade PCa at the first ‘for-cause’ screening.Data inputs for the risk calculators included age (high-grade calculator only), PSA and DRE status within the prior year, family history of PCa (any PCa calculator only), prior negative biopsy, and self-identified race (Black vs other, high-grade calculator only). Step 2 involves imputing the resulting estimated probabilities of total PCa and of high-grade PCa as the participant’s predicted ‘outcomes’, with time contributed from study entry until the date of first for-cause screening assessment. In step 3, for men i) diagnosed with PCa on study, ii) with no biopsy and no for-cause screening value, and iii) men with a negative on-study biopsy, an approach analogous to the proportional hazards time to PCa modeling was used. For men diagnosed with PCa, the probability of PCa equals 1, men with no diagnosis and no ‘for-cause’ screening valueare assumed to be disease free (probability=0) for the duration of the study, with time contributed (denominator) from study entry until diagnosis of cancer or last study contact. In the final step, the Incidence Density (ID) for each MVI user group was calculated as the sum of the estimated PCa probabilities as the numerator divided by the sum of total person-time at risk within the group.(22) IDR’s were calculated as the ratio of the ID within each MVI use group, divided by ID for the reference group (no MVI use). As an additional approach to address the potential for indication bias, where MVI use was prompted by an underlying condition that prompted receipt of pre-trial biopsy, analyses were restricted to men with no prior (negative) biopsy. Analyses were restricted to the 34,887 men with available supplement use and covariate data. Statistical analyses were performed using SAS, version 9.4 (SAS Institute Inc., Cary, North Carolina). All statistical tests were 2 sided, and p < .05 was considered statistically significant.
DATA AVAILABIILTY
The data generated in this study are available upon request from the corresponding author.
RESULTS
Baseline characteristics of study participants, stratified by reported multivitamin use at enrollment, are shown in Table 1. Compared to never and former users, men who were current multivitamin users were older, slightly less likely to be obese, have diabetes or be a current smoker, were more likely to be non-Hispanic white and reported higher levels of education. There was substantial overlap between the categories defining recency of (never, former, current) and cumulative years of MVI use (none, <5 years >=5 years); thus, the associations of characteristics with recency and duration of MVI use were similar (Table 1).
Table 1.
Demographic Factors by Prostate Cancer Status and Multivitamin Use at Baseline
| Prostate Cancer | Recency of Multivitamin Use at Baseline | Cumulative years of Multivitamin Use at Baseline | ||||||
|---|---|---|---|---|---|---|---|---|
|
|
|
|
||||||
| No N=33,030 | Yes N=1,857 | Never N=17,719 | Former N=2,821 | Current N=14,347 | None N=17,719 | <5 years N=5,502 | ≥5 years N=11,666 | |
|
| ||||||||
| Age, years | ||||||||
| < 60 | 12104 (36.6%) | 550 (29.6%) | 6678 (37.7%) | 1149 (40.7%) | 4827 (33.6%) | 6678 (37.7%) | 2078 (37.8%) | 3898 (33.4%) |
| 60 – 64 | 8632 (26.1%) | 542 (29.2%) | 4682 (26.4%) | 677 (24.0%) | 3815 (26.6%) | 4682 (26.4%) | 1423 (25.9%) | 3069 (26.3%) |
| 65 – 69 | 6416 (19.4%) | 443 (23.9%) | 3426 (19.3%) | 539 (19.1%) | 2894 (20.2%) | 3426 (19.3%) | 1069 (19.4%) | 2364 (20.3%) |
| >= 70 | 5878 (17.8%) | 322 (17.3%) | 2933 (16.6%) | 456 (16.2%) | 2811 (19.6%) | 2933 (16.6%) | 932 (16.9%) | 2335 (20.0%) |
| Education | ||||||||
| ≤ High school | 7283 (22.0%) | 399 (21.5%) | 4575 (25.8%) | 626 (22.2%) | 2481 (17.3%) | 4575 (25.8%) | 1210 (22.0%) | 1897 (16.3%) |
| Some college | 8895 (26.9%) | 460 (24.8%) | 4684 (26.4%) | 801 (28.4%) | 3870 (27.0%) | 4684 (26.4%) | 1523 (27.7%) | 3148 (27.0%) |
| ≥ College graduate | 16531 (50.0%) | 984 (53.0%) | 8242 (46.5%) | 1369 (48.5%) | 7904 (55.1%) | 8242 (46.5%) | 2722 (49.5%) | 6551 (56.2%) |
| Unknown | 321 (1.0%) | 14 (0.8%) | 218 (1.2%) | 25 (0.9%) | 92 (0.6%) | 218 (1.2%) | 47 (0.9%) | 70 (0.6%) |
| Marital Status | ||||||||
| Current | 26883 (81.4%) | 1547 (83.3%) | 14478 (81.7%) | 2218 (78.6%) | 11734 (81.8%) | 14478 (81.7%) | 4462 (81.1%) | 9490 (81.3%) |
| Former | 4723 (14.3%) | 246 (13.2%) | 2468 (13.9%) | 463 (16.4%) | 2038 (14.2%) | 2468 (13.9%) | 820 (14.9%) | 1681 (14.4%) |
| Never | 1248 (3.8%) | 55 (3.0%) | 633 (3.6%) | 129 (4.6%) | 541 (3.8%) | 633 (3.6%) | 204 (3.7%) | 466 (4.0%) |
| Unknown | 176 (0.5%) | 9 (0.5%) | 140 (0.8%) | 11 (0.4%) | 34 (0.2%) | 140 (0.8%) | 16 (0.3%) | 29 (0.2%) |
| Family history of Prostate cancer | ||||||||
| Yes | 5335 (16.2%) | 529 (28.5%) | 3024 (17.1%) | 471 (16.7%) | 2369 (16.5%) | 3024 (17.1%) | 891 (16.2%) | 1949 (16.7%) |
| Race | ||||||||
| Non-Hispanic White | 26097 (79.0%) | 1474 (79.4%) | 13554 (76.5%) | 2087 (74.0%) | 11930 (83.2%) | 13554 (76.5%) | 4151 (75.4%) | 9866 (84.6%) |
| Non-Hispanic Black | 4018 (12.2%) | 268 (14.4%) | 2433 (13.7%) | 409 (14.5%) | 1444 (10.1%) | 2433 (13.7%) | 796 (14.5%) | 1057 (9.1%) |
| Other | 2915 (8.8%) | 115 (6.2%) | 1732 (9.8%) | 325 (11.5%) | 973 (6.7%) | 1732 (9.8%) | 555 (10.1%) | 743 (6.3%) |
| Cigarette smoking | ||||||||
| Current | 2575 (7.8%) | 107 (5.8%) | 1548 (8.7%) | 215 (7.6%) | 919 (6.4%) | 1548 (8.7%) | 391 (7.1%) | 743 (6.4%) |
| Former | 16289 (49.3%) | 868 (46.7%) | 8498 (48.0%) | 1395 (49.5%) | 7264 (50.6%) | 8498 (48.0%) | 2679 (48.7%) | 5980 (51.3%) |
| Never | 14004 (42.4%) | 877 (47.2%) | 7538 (42.5%) | 1203 (42.6%) | 6140 (42.8%) | 7538 (42.5%) | 2417 (43.9%) | 4926 (42.2%) |
| Unknown | 162 (0.5%) | 5 (0.3%) | 135 (0.8%) | 8 (0.3%) | 24 (0.2%) | 135 (0.8%) | 15 (0.3%) | 17 (0.1%) |
| Alcohol consumption, g/d | ||||||||
| 0 – <0.5 (none to < 1 drink/mo) | 11749 (35.6%) | 644 (34.7%) | 6470 (36.6%) | 1079 (38.2%) | 4844 (33.8%) | 6470 (36.6%) | 2114 (38.5%) | 3809 (32.7%) |
| 0.5 – 9.9 (1 drink/mo to <1 drink/day) | 10482 (31.7%) | 592 (31.8%) | 5506 (31.1%) | 882 (31.3%) | 4652 (32.5%) | 5506 (31.1%) | 1705 (31.0%) | 3829 (32.9%) |
| >10 (≥1 drink per day) | 10799 (32.7%) | 621 (33.5%) | 5723 (32.3%) | 860 (30.5%) | 4837 (33.7%) | 5723 (32.3%) | 1679 (30.5%) | 4018 (34.5%) |
| Unknown/missing | 31 (<1%) | 3 (<1%) | 20 (<1%) | 14 (<1%) | 20 (<1%) | 4 (<1%) | 10 (<1%) | |
| Body Mass Index, kg/m2 | ||||||||
| <25 | 6803 (20.6%) | 363 (19.5%) | 3535 (20.0%) | 593 (21.0%) | 3038 (21.2%) | 3535 (20.0%) | 1127 (20.5%) | 2504 (21.5%) |
| 25.0 to 29.9 | 15804 (47.8%) | 947 (51.0%) | 8416 (47.5%) | 1357 (48.1%) | 6978 (48.6%) | 8416 (47.5%) | 2611 (47.5%) | 5724 (49.1%) |
| ≥30.0 | 10423 (31.6%) | 547 (29.5%) | 5768 (32.6%) | 871 (30.9%) | 4331 (30.2%) | 5768 (32.6%) | 1764 (32.1%) | 3438 (29.5%) |
| Diabetes | ||||||||
| Yes | 3492 (10.6%) | 133 (7.2%) | 1949 (11.0%) | 290 (10.3%) | 1386 (9.7%) | 1949 (11.0%) | 584 (10.6%) | 1092 (9.4%) |
| Study Treatment Arm | ||||||||
| Selenium only | 8298 (25.1%) | 454 (24.4%) | 4465 (25.2%) | 730 (25.9%) | 3557 (24.8%) | 4465 (25.2%) | 1397 (25.4%) | 2890 (24.8%) |
| Selenium + Vitamin E | 8248 (25.0%) | 454 (24.4%) | 4335 (24.5%) | 720 (25.5%) | 3647 (25.4%) | 4335 (24.5%) | 1400 (25.4%) | 2967 (25.4%) |
| Placebo | 8253 (25.0%) | 443 (23.9%) | 4440 (25.1%) | 676 (24.0%) | 3580 (25.0%) | 4440 (25.1%) | 1344 (24.4%) | 2912 (25.0%) |
| Vitamin E only | 8231 (24.9%) | 506 (27.2%) | 4479 (25.3%) | 695 (24.6%) | 3563 (24.8%) | 4479 (25.3%) | 1361 (24.7%) | 2897 (24.8%) |
Table 2 shows associations of baseline MVI use with the risk of total, low- and high-grade PCa. In multivariate models adjusted for age, race, family history of PCa, BMI, diabetes, study treatment arm, PSA and DRE, there were no significant associations of recency of or cumulative years of MVI use with risk of total or low-grade PCa. Compared to no history of MVI use at baseline, former and current MVI use was associated with a 17% (95% CI 1%, 34%) and 21% (95% CI 12%, 31%), respectively, higher risk of high-grade PCa. Similarly, men who reported <5 or ≥5 years of MVI use at baseline had a 14% (95% CI 2%, 27%) and 24% (95% CI 14%, 34%) higher risk of high-grade PCa compared to non-users, respectively. A separate model additionally adjusted for education, smoking status and alcohol consumption yielded similar comparable estimates of risk.
Table 2.
Self-reported History of Multivitamin Use and Prostate Cancer Risk
| Recency of Multivitamin Use at Baseline | Cumulative Years of Multivitamin Use at Baseline | |||||
|---|---|---|---|---|---|---|
|
|
|
|||||
| None n=17,719 | Former n=2,821 | Current n=14,347 | None n=17,719 | 0 – <5 years n=5,502 | ≥5 years n=11,666 | |
|
| ||||||
| Total PCa | ||||||
| # cases (n=1,857) | 943 | 139 | 775 | 943 | 273 | 641 |
| Model 1: HR (95% CI) 1 | Ref | 0.98 (0.91–1.06) | 1.01 (0.97–1.06) | Ref | 0.97 (0.92–1.03) | 1.02 (0.98–1.07) |
| Model 2: HR (95% CI) 2 | Ref | 0.97 (0.90–1.05) | 1.01 (0.97–1.05) | Ref | 0.97 (0.91–1.03) | 1.02 (0.98–1.07) |
| Low-Grade PCa (Gleason 2–6) 3 | ||||||
| # cases (n=1,061) | 585 | 87 | 464 | 585 | 173 | 378 |
| Model 1: HR (95% CI) 1 | Ref | 0.93 (0.84–1.03) | 0.97 (0.92–1.03) | Ref | 0.94 (0.87–1.02) | 0.98 (0.92–1.04) |
| Model 2: HR (95% CI) 2 | Ref | 0.93 (0.84–1.02) | 0.96 (0.91–1.02) | Ref | 0.93 (0.87–1.01) | 0.97 (0.91–1.03) |
| High-Grade Pca (Gleason 7–10) 3 | ||||||
| # cases (n=524) | 275 | 40 | 262 | 275 | 83 | 219 |
| Model 1: HR (95% CI) 1 | Ref | 1.13 (0.99–1.30) | 1.19 (1.10–1.28) | Ref | 1.10 (1.00–1.22) | 1.21 (1.12–1.31) |
| Model 2: HR (95% CI) 2 | Ref | 1.13 (0.99–1.29) | 1.19 (1.10–1.28) | Ref | 1.10 (0.99–1.22) | 1.22 (1.13–1.32) |
Model 1: Hazard ratios are adjusted for age (linear), race (white, black, other), family history of prostate cancer, longitudinal PSA (linear; time-dependent); longitudinal DRE (normal vs abnormal; time-dependent), negative biopsy at baseline or on study (yes vs no; time-dependent), BMI (<25, 25 to <30, 30+), history of diabetes (yes, no/unknown), and treatment arm (vitamin E, selenium, both, placebo).
Model 2: Hazard ratios are adjusted for age (linear), race (white, black, other), family history of prostate cancer, longitudinal PSA (linear; time-dependent); longitudinal DRE (normal vs abnormal; time-dependent), negative biopsy at baseline or on study (yes vs no; time-dependent), BMI (<25, 25 to <30, 30+), history of diabetes (yes, no/unknown), and treatment arm (vitamin E, selenium, both, placebo) AND education, smoking status and alcohol consumption.
For grade-specific models, participants with a different grade of cancer are censored at diagnosis date, and cases with no known Gleason score (N=203) are removed from the analysis.
To investigate whether MVI use impacted screening patterns and biopsy acceptance within SELECT, Table 3 shows the distribution of several PCa screening-related characteristics by reported MVI use at baseline. There were no appreciable differences in the overall length of follow-up and the number of PSA and DRE screens per year of follow-up. Current MVI users had a slightly lower media PSA at first on-study biopsy and slightly higher proportions of participants with an on-study biopsy, or with a for-cause biopsy prompt (PSA>4 or abnormal DRE) who did not receive a biopsy. Differences across cumulative MVI use categories were similar. The proportion of participants who had a negative biopsy prior to entering SELECT was substantially higher among current and long-term supplement users compared to never/former or shorter-term (0 to <5 years) supplement users. In addition, time to first on-study biopsy was significantly shorter for current and long-term supplement users compared to never users while accounting for other known risk factors including on-study PSA and DRE.
Table 3.
Distributions of Clinical and Screening Factors by Multivitamin Use at Baseline
| Recency of Multivitamin Use at Baseline | Cumulative Years of Multivitamin Use at Baseline | |||||
|---|---|---|---|---|---|---|
|
|
|
|||||
| Never n=17,719 | Former n=2,821 | Current n=14,347 | None n=17,719 | <5 years n=5,502 | ≥5 years n=11,666 | |
|
| ||||||
| Length of follow-up time on trial, mean yrs | 5.2 | 5.2 | 5.2 | 5.2 | 5.2 | 5.2 |
| Number of PSA per year of follow-up, mean (SD) | 0.9 (0.3) | 0.9 (0.3) | 1.0 (0.3) | 0.9 (0.3) | 1.0 (0.3) | 1.0 (0.3) |
| Number of DRE per year of follow-up, mean (SD) | 0.8 (0.3) | 0.8 (0.3) | 0.8 (0.3) | 0.8 (0.3) | 0.8 (0.3) | 0.8 (0.3) |
| Median PSA at first on-study biopsy 1 | 3.8 | 3.7 | 3.6 | 3.8 | 3.5 | 3.7 |
| N (%) with on-study biopsy | 2473 (14.0%) | 390 (13.8%) | 2150 (15.0%) | 2473 (14.0%) | 777 (14.1%) | 1763 (15.1%) |
| N with PSA>4 or abnormal DRE at any point in study and no biopsy | 1475 (8.3%) | 246 (8.7%) | 1233 (8.6%) | 1475 (8.3%) | 451 (8.2%) | 1028 (8.8%) |
| N (%) with negative biopsy prior to SELECT | 1563 (8.8%) | 236 (8.4%) | 1608 (11.2%) | 1563 (8.8%) | 482 (8.8%) | 1362 (11.7%) |
| HR (95% CI) | HR (95% CI) | |||||
|
|
||||||
| Time to first biopsy on-study 2 | Ref | 0.99 (0.94–1.05) | 1.10 (1.06–1.13) | Ref | 1.02 (0.98–1.06) | 1.11 (1.07–1.14) |
Most recent recorded PSA at the time of first recorded biopsy, for both cases and non-cases. For cases with no recorded biopsy, diagnosis date is used.
Hazard ratios adjusted for age, family history of prostate cancer, PSA and DRE (longitudinal)
Table 4 compares observed associations of MVI use and PCa risk and detection bias adjusted associations in which PCa status was predicted for men who had a for-cause prompt (PSA>4 or abnormal DRE) but did not undergo biopsy. Compared to observed associations of MVI use with PCa risk, after imputing total and high-grade PCa status for the overall study population, associations of current and long-term MVI use with risk of total or high-grade PCa were somewhat attenuated (Incidence Density Hazard ratios for Current vs None: 1.10 (95% CI 0.96, 1.26) and >5 years vs None 1.20 (95% CI 1.01, 1.44), respectively), although the confidence intervals were slightly wider and still overlap those from Table 2. Analyses limited to men who had no prior (negative) biopsy, found that for observed associations of MVI use, former and < 5 years of use were associated with slight reductions in risk of total PCa; however, after imputation, the associations were essentially null. For high-grade PCa, associations of recency and duration of MVI use among men who had no prior biopsy were slightly weaker compared to the full population, and after imputation, associations for current and long-term MVI use with risk of high-grade PCa were further attenuated and no longer statistically significant (IDR Current use vs None: 1.07 (95% CI 0.93, 1.24) and >5 years use vs None 1.09 (95% CI 0.94–1.27), respectively).
Table 4.
Self-reported History of Multivitamin Use and Prostate Cancer Risk
| Recency of Multivitamin Use at Baseline | Cumulative Years of Multivitamin Use at Baseline | |||||
|---|---|---|---|---|---|---|
|
|
|
|||||
| Never n=17,719 | Former n=2,821 | Current n=14,347 | None n=17,719 | <5 years n=5,502 | ≥5 years n=11,666 | |
|
| ||||||
| Total PCa | ||||||
| HR (95% CI)1 | Ref | 0.98 (0.91–1.06) | 1.01 (0.97–1.06) | Ref | 0.97 (0.92–1.03) | 1.02 (0.98–1.07) |
| IDR (95% CI)2 | Ref | 0.97 (0.85, 1.12) | 1.01 (0.93, 1.09) | Ref | 0.96 (0.86–1.07) | 1.03 (0.95–1.11) |
| High-Grade PCa (Gleason 7–10) 3 | ||||||
| HR (95% CI)1 | Ref | 1.13 (0.99–1.30) | 1.19 (1.10–1.28) | Ref | 1.10 (1.00–1.22) | 1.21 (1.12–1.31) |
| IDR (95% CI)2 | Ref | 0.93 (0.72, 1.21) | 1.10 (0.96, 1.26) | Ref | 0.97 (0.77–1.24) | 1.20 (1.01–1.44) |
|
|
||||||
| Excluding Men who had a biopsy prior to SELECT Trial entry | ||||||
|
|
||||||
| N=16,156 | N=2,585 | N=12,739 | N=16,156 | N=5,020 | N=10,304 | |
| Total PCa | ||||||
| HR (95% CI)1 | Ref | 0.88 (0.81–0.96) | 1.02 (0.97–1.06) | Ref | 0.92 (0.87–0.98) | 1.03 (0.98–1.08) |
| IDR (95% CI)2 | Ref | 0.95 (0.82, 1.10) | 1.01 (0.93, 1.09) | Ref | 0.94 (0.84–1.05) | 1.03 (0.95–1.12) |
| High-Grade PCa (Gleason 7–10) 3 | ||||||
| HR (95% CI)1 | Ref | 0.93 (0.80–1.07) | 1.15 (1.06–1.24) | Ref | 0.99 (0.88–1.10) | 1.17 (1.08–1.27) |
| IDR (95% CI)2 | Ref | 0.86 (0.65, 1.15) | 1.07 (0.93, 1.24) | Ref | 0.94 (0.77–1.16) | 1.09 (0.94–1.27) |
Model 1: Hazard ratios are adjusted for age (linear), race (white, black, other), family history of prostate cancer, longitudinal PSA (linear; time-dependent); longitudinal DRE (normal vs abnormal; time-dependent), negative biopsy at baseline or on study (yes vs no; time-dependent).
Incidence Density Hazard ratios are adjusted for age, PSA, DRE status within the prior year, family history of prostate cancer, prior negative biopsy and race; Imputation was performed separately for any and high-grade PCa.
For the -specific models, participants with a different grade of cancer are censored at diagnosis date, and cases with no known Gleason score (N=203) are removed from the analysis.
DISCUSSION
In the Selenium and Vitamin E Chemoprevention Trial, we utilized detailed PCa screening data collected at each trial contact and identified differences in biopsy timing by MVI use that indicate the potential for PCa detection bias. Analyses of data from reported biopsies suggest a positive association for current and long-term MVI use with time to total and high-grade PCa; however, after accounting for differential biopsy acceptance, MVI use was no longer statistically significantly associated with PC risk. These findings are critical for informing whether MVI use poses a risk for aging men in relation to PCa risk. A naïve approach not accounting for detection bias would have concluded that any use and extended use of MVI could lead to higher risk of high-grade PCa.
Screening related biases are important to consider in epidemiologic studies of PCa. The potential for bias is substantial because the disease is highly prevalent and typically asymptomatic, and in order to receive a diagnosis of PCa, a man must be screened and undergo biopsy. Therefore, any factor that affects the likelihood of having a biopsy or timing of biopsy is likely to appear associated with PCa, even if there is no true association. Screening-related biases can occur at the level of biopsy recommendation by the physician or acceptance by the patient but may also be related to the timing of biopsy. We previously observed that a broad range of demographic and lifestyle factors including age, marital status, family history of PCa, BMI and health conditions, even after adjusting for whether or not a screen was conducted, or the outcome of a screening test, were related to the time to first biopsy in SELECT.(23) In that study, a naïve approach suggested that aspirin use was associated with a 34% reduced risk of PCa; however, after controlling for differential biopsy acceptance, aspirin use was associated with a 20–23% increased risk of PCa. Following this observation, the Aspirin in Reducing Events in the Elderly (ASPREE) trial reported a 31% increased risk of cancer death in participants receiving aspirin.(24) In the present analyses, many of these same factors, such as age, BMI, diabetes, education, smoking and marital status were also associated with MVI use. Although we found no appreciable differences in PSA and DRE screening rates, the threshold for biopsy (PSA at biopsy) or acceptance of biopsy prompt (given a positive screening result) between MVI users and non-users, current and long-term MVI users were more likely to have had a negative biopsy prior to enrollment in SELECT. It is possible that current and long-term MVI users with a prior negative biopsy were taking MVIs out of concern for prostate cancer or the condition that led to screening, or that they were a more highly screened population and perhaps had less undetected PCa at study entry. In addition, after accounting for PSA and DRE results up until biopsy, the timing of the first on-study biopsy was earlier for current and long-term MVI users compared to never users. These data indicate that patterns in biopsy acceptance and timing differed by MVI use. After applying statistical methods that attempt to mimic uniform biopsy recommendations and acceptance and account for differential biopsy patterns, we found attenuated associations of historical MVI use and PCa risk, bringing into question whether there is a true association between MVI use and PCa or just a perception.
Findings from prospective cohort studies have suggested that multivitamin use is associated with an increased risk of total(10) and advanced PCa(11) and PCa death(11,12); however, screening-related bias is a plausible explanation for the results reported in at least two of these studies. In the National Institutes of Health-AARP cohort(10,11), history of PSA or DRE screening in the 3 years before baseline was available on roughly 60% of participants, and men who used supplements were more likely to have been screened than nonusers.(11) In addition, a stratified analysis revealed that the reported association between MVI use and advanced and fatal PCa was apparent only among men with a family history of PCa. We have previously shown that even after controlling for PSA and DRE findings, men with a family history of PCa were more likely to undergo biopsy.(23) Thus, it is likely that analyses within this subset of men may be more susceptible to detection bias. Further, if the outcome is misclassified (incorrect) due to differential biopsy, no amount of covariate adjustment can minimize this bias. In contrast, studies where the potential for differential biopsy bias has been minimized reported no associations of MVI use with PCa. In the Prostate Cancer Prevention Trial (PCPT), which included annual screening (PSA and DRE) and an end of study biopsy, no association was reported between MVI (<1 vs.1–6 or 7+ pills/week) use and risk of low- or high-grade (Gleason grade 8–10) PCa.(13) In addition, in the Physicians Health Study, a large (n=14,641) randomized trial of Centrum Silver® vs. placebo, daily multivitamin supplementation for a median of 11.2 years had no overall effect on PCa risk or death (HR=0.98, 95% CI=0.88 to 1.09; HR=0.91, 95% CI=0.66 to 1.25, respectively).(13)
Since the impact of detection bias is expected to be strongest for low-grade PCa’s, which are primarily detected via screening, the lack of association between MVI use and low-grade or total PCa risk in SELECT deserves comment. The SELECT study population is highly screened, with the majority of participants undergoing annual PSA and DRE testing (85% and 70%, respectively)(16) and participants underwent first on-study biopsy at a somewhat modest PSA (Table 2). It is possible that the lack of variability in screening and adoption of biopsy made it difficult to detect an association between MVI use and low-grade/total PCa. In cohorts with more variability of screening and biopsy recommendation, the associations of MVI use and PCa risk may differ. Although detection bias is primarily attributed to screen-detected or low-grade PCa, it is possible for studies evaluating advanced disease to also be impacted. In this study, after accounting for PSA and DRE values, we found that biopsies occurred earlier among MVI users compared to non-users. Differences in biopsy timing, the point at which a PCa is detected, can impact the calculation of ‘time’ within time to event models, which subsequently affects the hazards ratio. In addition, the absence of screening or biopsy can lead to advanced disease. For example, factors related to comorbid conditions or contraindications for biopsy, such as BMI or diabetes, have been associated with a lower likelihood of having a biopsy(23,25,26) and also with more advanced disease.(27,28)
Strengths of this study include the prospective design, central pathology review and use of a validated questionnaire to collect data on dietary supplement use.(18) A particular strength of this study is the detailed assessment of PCa screening, including PSA, DRE, and biopsy at enrollment and throughout the trial, which enabled evaluation of detection bias. This study is not without limitations. MVI use was self-reported and may not represent actual use. We also cannot rule out the possibility that survival bias may have impacted the representativeness of the study population or the ability to have PCa detected during the trial. In addition, participants in SELECT were participating in a trial, and although screening was not part of the trial protocol, (16) the majority of cancers were screen-detected and early stage. In cohorts with more variability of screening and biopsy recommendation, it is likely that the impact of differential ascertainment of PCa by risk factors could be more pronounced. Lastly, we used the PCPT risk calculator to estimate the predicted probability of any and high-grade PCa among men who had an abnormal screen result but no biopsy.(20,21) There are likely a proportion of men with a negative screen results (PSA<4 and normal DRE) who had indolent prostate cancer; however, for this analysis, we applied rules for imputing a prostate cancer endpoint based on the general screening guidance in place at the time of SELECT. Further, while other risk prediction models exist, the PCPT risk calculator was developed using data from a comparably screened prevention trial, has been validated in a number of cohorts(29), and is widely used within the urologic community(30).
In this and prior analyses in SELECT, we found that a broad range of factors, including multivitamin use, are associated with the probability of having a biopsy and may also be associated with biopsy timing. In the presence of differential patterns of biopsy acceptance, use of observed biopsy data may yield biased estimates of the associations of a potential risk factor like MVI use with PCa risk. Statistical methods using individual-level screening and biopsy data can be used to minimize differential detection bias; although, covariate adjustment alone cannot resolve detection bias resulting in endpoint misclassification. Differential biopsy ascertainment may play a role in evaluating associations of potential risk factors and prostate cancer, and detailed screening and biopsy data can be used to analytically minimize such bias.
Acknowledgments:
These analyses were supported by an Administrative Supplement from the Office of Dietary Supplements
Funding:
U01CA182883, U10CA37429 (legacy), UG1CA189974 and UM1CA182883.
Footnotes
COI: The authors declare no potential conflicts of interest.
REFERENCES
- 1.Siegel RL, Miller KD, Fuchs HE, Jemal A. Cancer statistics, 2022. CA Cancer J Clin 2022;72(1):7–33 doi 10.3322/caac.21708. [DOI] [PubMed] [Google Scholar]
- 2.Chan JM, Gann PH, Giovannucci EL. Role of diet in prostate cancer development and progression. J Clin Oncol 2005;23(32):8152–60 doi 10.1200/JCO.2005.03.1492. [DOI] [PubMed] [Google Scholar]
- 3.Fulgoni VL 3rd, Keast DR, Bailey RL, Dwyer J. Foods, fortificants, and supplements: Where do Americans get their nutrients? J Nutr 2011;141(10):1847–54 doi 10.3945/jn.111.142257. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Newman JC, Malek AM, Hunt KJ, Marriott BP. Nutrients in the US Diet: Naturally Occurring or Enriched/Fortified Food and Beverage Sources, Plus Dietary Supplements: NHANES 2009–2012. J Nutr 2019;149(8):1404–12 doi 10.1093/jn/nxz066. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Kantor ED, Rehm CD, Du M, White E, Giovannucci EL. Trends in Dietary Supplement Use Among US Adults From 1999–2012. JAMA 2016;316(14):1464–74 doi 10.1001/jama.2016.14403. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Albanes D, Heinonen OP, Huttunen JK, Taylor PR, Virtamo J, Edwards BK, et al. Effects of alpha-tocopherol and beta-carotene supplements on cancer incidence in the Alpha-Tocopherol Beta-Carotene Cancer Prevention Study. Am J Clin Nutr 1995;62(6 Suppl):1427S–30S. [DOI] [PubMed] [Google Scholar]
- 7.Kristal AR, Darke AK, Morris JS, Tangen CM, Goodman PJ, Thompson IM, et al. Baseline selenium status and effects of selenium and vitamin e supplementation on prostate cancer risk. J Natl Cancer Inst 2014;106(3):djt456 doi 10.1093/jnci/djt456. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Figueiredo JC, Grau MV, Haile RW, Sandler RS, Summers RW, Bresalier RS, et al. Folic acid and risk of prostate cancer: results from a randomized clinical trial. J Natl Cancer Inst 2009;101(6):432–5 doi 10.1093/jnci/djp019. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Omenn GS, Goodman GE, Thornquist MD, Balmes J, Cullen MR, Glass A, et al. Effects of a combination of beta carotene and vitamin A on lung cancer and cardiovascular disease. N Engl J Med 1996;334(18):1150–5 doi 10.1056/NEJM199605023341802. [DOI] [PubMed] [Google Scholar]
- 10.Lim JE, Weinstein SJ, Liao LM, Sinha R, Huang J, Albanes D. Multivitamin Use and Overall and Site-Specific Cancer Risks in the National Institutes of Health-AARP Diet and Health Study. J Nutr 2022;152(1):211–6 doi 10.1093/jn/nxab322. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Lawson KA, Wright ME, Subar A, Mouw T, Hollenbeck A, Schatzkin A, et al. Multivitamin use and risk of prostate cancer in the National Institutes of Health-AARP Diet and Health Study. J Natl Cancer Inst 2007;99(10):754–64 doi 10.1093/jnci/djk177. [DOI] [PubMed] [Google Scholar]
- 12.Stevens VL, McCullough ML, Diver WR, Rodriguez C, Jacobs EJ, Thun MJ, et al. Use of multivitamins and prostate cancer mortality in a large cohort of US men. Cancer Causes Control 2005;16(6):643–50 doi 10.1007/s10552-005-0384-5. [DOI] [PubMed] [Google Scholar]
- 13.Kristal AR, Arnold KB, Neuhouser ML, Goodman P, Platz EA, Albanes D, et al. Diet, supplement use, and prostate cancer risk: results from the prostate cancer prevention trial. Am J Epidemiol 2010;172(5):566–77 doi 10.1093/aje/kwq148. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Gaziano JM, Sesso HD, Christen WG, Bubes V, Smith JP, MacFadyen J, et al. Multivitamins in the prevention of cancer in men: the Physicians’ Health Study II randomized controlled trial. JAMA 2012;308(18):1871–80 doi 10.1001/jama.2012.14641. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Lippman SM, Goodman PJ, Klein EA, Parnes HL, Thompson IM Jr., Kristal AR, et al. Designing the Selenium and Vitamin E Cancer Prevention Trial (SELECT). J Natl Cancer Inst. Volume 972005. p 94–102. [DOI] [PubMed] [Google Scholar]
- 16.Lippman SM, Klein EA, Goodman PJ, Lucia MS, Thompson IM, Ford LG, et al. Effect of selenium and vitamin E on risk of prostate cancer and other cancers: the Selenium and Vitamin E Cancer Prevention Trial (SELECT). JAMA 2009;301(1):39–51 doi 10.1001/jama.2008.864. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Patterson RE, Kristal AR, Levy L, McLerran D, White E. Validity of methods used to assess vitamin and mineral supplement use. Am J Epidemiol 1998;148(7):643–9 doi 10.1093/aje/148.7.643. [DOI] [PubMed] [Google Scholar]
- 18.Satia-Abouta J, Patterson RE, King IB, Stratton KL, Shattuck AL, Kristal AR, et al. Reliability and validity of self-report of vitamin and mineral supplement use in the vitamins and lifestyle study. Am J Epidemiol 2003;157(10):944–54. [DOI] [PubMed] [Google Scholar]
- 19.Tangen CM, Schenk J, Till C, Goodman PJ, Barrington W, Lucia MS, et al. Variations in prostate biopsy recommendation and acceptance confound evaluation of risk factors for prostate cancer: Examining race and BMI. Cancer Epidemiol 2019;63:101619 doi 10.1016/j.canep.2019.101619. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Thompson IM, Ankerst DP, Chi C, Goodman PJ, Tangen CM, Lucia MS, et al. Assessing prostate cancer risk: results from the Prostate Cancer Prevention Trial. J Natl Cancer Inst 2006;98(8):529–34 doi 10.1093/jnci/djj131. [DOI] [PubMed] [Google Scholar]
- 21.Ankerst DP, Hoefler J, Bock S, Goodman PJ, Vickers A, Hernandez J, et al. Prostate Cancer Prevention Trial risk calculator 2.0 for the prediction of low- vs high-grade prostate cancer. Urology 2014;83(6):1362–7 doi 10.1016/j.urology.2014.02.035. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Kleinbaum DG, Kupper LL, Morgenstern H. Epidemiologic research: principles and quantitative methods. John Wiley & Sons; 1991. [Google Scholar]
- 23.Tangen CM, Goodman PJ, Till C, Schenk JM, Lucia MS, Thompson IM Jr., Biases in Recommendations for and Acceptance of Prostate Biopsy Significantly Affect Assessment of Prostate Cancer Risk Factors: Results From Two Large Randomized Clinical Trials. J Clin Oncol 2016;34(36):4338–44. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.McNeil JJ, Nelson MR, Woods RL, Lockery JE, Wolfe R, Reid CM, et al. Effect of Aspirin on All-Cause Mortality in the Healthy Elderly. N Engl J Med 2018;379(16):1519–28 doi 10.1056/NEJMoa1803955. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Miller EA, Pinsky PF, Pierre-Victor D. The relationship between diabetes, prostate-specific antigen screening tests, and prostate cancer. Cancer Causes Control 2018;29(10):907–14 doi 10.1007/s10552-018-1067-3. [DOI] [PubMed] [Google Scholar]
- 26.Beckmann K, Crawley D, Nordstrom T, Aly M, Olsson H, Lantz A, et al. Association Between Antidiabetic Medications and Prostate-Specific Antigen Levels and Biopsy Results. JAMA Netw Open 2019;2(11):e1914689 doi 10.1001/jamanetworkopen.2019.14689. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Gong Z, Neuhouser ML, Goodman PJ, Albanes D, Chi C, Hsing AW, et al. Obesity, diabetes, and risk of prostate cancer: results from the prostate cancer prevention trial. Cancer Epidemiol Biomarkers Prev 2006;15(10):1977–83 doi 10.1158/1055-9965.EPI-06-0477. [DOI] [PubMed] [Google Scholar]
- 28.Genkinger JM, Wu K, Wang M, Albanes D, Black A, van den Brandt PA, et al. Measures of body fatness and height in early and mid-to-late adulthood and prostate cancer: risk and mortality in The Pooling Project of Prospective Studies of Diet and Cancer. Ann Oncol 2020;31(1):103–14 doi 10.1016/j.annonc.2019.09.007. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Parekh DJ, Ankerst DP, Higgins BA, Hernandez J, Canby-Hagino E, Brand T, et al. External validation of the Prostate Cancer Prevention Trial risk calculator in a screened population. Urology 2006;68(6):1152–5 doi 10.1016/j.urology.2006.10.022. [DOI] [PubMed] [Google Scholar]
- 30.Bandala-Jacques A, Castellanos Esquivel KD, Perez-Hurtado F, Hernandez-Silva C, Reynoso-Noveron N. Prostate Cancer Risk Calculators for Healthy Populations: Systematic Review. JMIR Cancer 2021;7(3):e30430 doi 10.2196/30430. [DOI] [PMC free article] [PubMed] [Google Scholar]