Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
American Journal of Epidemiology Copyright © 2004 by the Johns Hopkins Bloomberg School of Public Health All rights reserved Vol. 159, No. 1 Printed in U.S.A. DOI: 10.1093/aje/kwh010 PRACTICE OF EPIDEMIOLOGY VITamins And Lifestyle Cohort Study: Study Design and Characteristics of Supplement Users Emily White1,2, Ruth E. Patterson1,2, Alan R. Kristal1,2, Mark Thornquist1, Irena King1, Ann L. Shattuck1, Ilonka Evans1, Jessie Satia-Abouta1,3, Alyson J. Littman1,2, and John D. Potter1,2 1 Cancer Prevention Research Program, Fred Hutchinson Cancer Research Center, Seattle, WA. Department of Epidemiology, School of Public Health and Community Medicine, University of Washington, Seattle, WA. 3 Department of Nutrition, School of Public Health, University of North Carolina, Chapel Hill, NC. 2 Received for publication April 22, 2003; accepted for publication July 11, 2003. Vitamin and mineral supplements are among the most commonly used drugs in the United States, despite limited evidence on their benefits or risks. This paper describes the design, implementation, and participant characteristics of the VITamins And Lifestyle (VITAL) Study, a cohort study of the associations of supplement use with cancer risk. A total of 77,738 men and women in western Washington State, aged 50–76 years, entered the study in 2000–2002 by completing a detailed questionnaire on supplement use, diet, and other cancer risk factors, and 70% provided DNA through self-collected buccal cell specimens. Supplement users were targeted in recruitment: 66% used multivitamins, 46% used individual vitamin C, 47% used individual vitamin E, and 46% used calcium, typically for 5–8 of the past 10 years. Analyses to identify confounding factors, the main study limitation, showed that regular nonsteroidal anti-inflammatory drug use, intake of fruits and vegetables, and recreational physical activity were strongly associated with supplement use (p < 0.001). The authors describe a follow-up system in which cancers, deaths, and changes of residence are tracked efficiently, primarily through linkage to public databases. These methods may be useful to other researchers implementing a large cohort study or designing a passive follow-up system. calcium; data collection; dietary supplements; epidemiologic research design; follow-up studies; vitamins Abbreviations: SEER, Surveillance, Epidemiology, and End Results; VITAL, VITamins And Lifestyle. Consumers have little rigorous scientific information to guide them in selecting types and dosages of supplements to use for disease prevention or, specifically, cancer prevention. Many nutrients have shown cancer preventive properties in in vitro studies, animal studies, and human clinical studies (2–4). Hundreds of epidemiologic studies on diet and cancer have been conducted; an extensive review found strong and consistent evidence that diets high in vegetables and fruit reduce the risk of cancers of the mouth and pharynx, esophagus, lung, stomach, colon and rectum, and possibly breast and bladder; however, studies of dietary intake of specific nutrients have been less consistent (5). Vitamin, mineral, and other dietary supplements are among the most commonly used drugs in the United States. According to recent national surveys, 48–55 percent of US adults use vitamin and/or mineral supplements regularly (1; E. Slaughter, Princeton Survey Research Associates, personal communication, 2002). Moreover, some supplements provide much higher levels of nutrients than can be obtained from food. For example, median intake of vitamin E from food is 7–8 mg of α-tocopherol for those aged 51–70 years (2), whereas the common dose in an individual supplement is 20 times this amount (400 IU or 180 mg αtocopherol equivalents). Correspondence to Dr. Emily White, Cancer Prevention Program MP-702, Fred Hutchinson Cancer Research Center, PO Box 19024, Seattle, WA 98109-1024 (e-mail: [email protected]). 83 Am J Epidemiol 2004;159:83–93 84 White et al. TABLE 1. Disposition of questionnaires and response rates in the VITamins And Lifestyle cohort study, Washington State, 2000–2002 Full mailing list No. % 364,418 100 Undeliverable questionnaire 6,900† 1.9 Deceased 7,300† 2.0 Duplicate questionnaire 4,700† Age ineligible Out of area at baseline Total Returned questionnaires Response rate (%) No. % 79,300 100 1.3 73 0.1 4,000† 1.1 844 1.1 2,600† 0.7 402 0.5 3 0.0 Questionnaire completed by a person to whom it was not mailed 163 0.2 Excess missing data‡ 77 0.1 77,738 98.0 21.8* Ineligible Transsexual Failed quality control checks Eligible and passed quality control 338,918† 93.0 22.9§ * Percentage of questionnaires returned. † Estimate based on counts within subsamples or external estimates. ‡ Failed two of three quality control checks: some items on every page of the food frequency questionnaire were completed, at least half of the cancer risk factor questions were completed, at least some items in the supplement section were completed. § Estimated response rate among eligible individuals. Far fewer human studies of cancer risk have explicitly studied nutrients obtained from supplements. We have published two critical reviews of this literature (6, 7). Key findings are the adverse effects of β-carotene on lung cancer (8, 9) and a probable reduction in risk of prostate cancer with use of supplemental vitamin E (10) and selenium (11) (a common component of multivitamins). There is also supportive evidence that supplemental vitamin E (12–14), supplemental calcium (15), and long-term multivitamin use (16, 17) may reduce the risk of colorectal cancer. Randomized trials are limited because they can test only specific agents at specific doses for typically a small number of years, and often in populations selected because of their high risk of certain cancers. In contrast, cohort studies can be an effective way to assess a wide range of supplements used at varying dosages for long durations. We therefore designed and implemented a cohort study with the overall aim of investigating the associations of supplement use with cancer risk. Specific goals focus on the associations of supplemental vitamin C, vitamin E, calcium, and multivitamins with prostate, female breast, lung, and colorectal cancer incidence. The study is called the VITamins And Lifestyle (VITAL) Study. This paper describes the design and implementation of the VITAL Study and presents baseline observations on factors associated with supplement use. We describe the study methods in detail, including recruitment of the cohort, data collection methods, and our passive follow-up system. This information provides a background for future reports from this study and may provide practical information to aid others who are implementing cohort studies. Description of the characteristics of those who use the main supplements of interest addresses the main limitation of this and other observational studies of supplement use—that factors associated with supplement use need to be identified so they can be controlled for as potential confounding factors. MATERIALS AND METHODS VITAL cohort recruitment Men and women were eligible to join the cohort if they were aged 50–76 years and lived in the 13-county area in western Washington State covered by the Surveillance, Epidemiology, and End Results (SEER) program cancer registry. Using names purchased from a commercial mailing list, we mailed 364,418 baseline questionnaires followed by a postcard reminder after 2 weeks. To recruit supplement users into the cohort, the cover letter described the study as one on supplement use and cancer risk, but the study was not restricted to supplement users. Recruitment was conducted from October 2000 to December 2002, during which time 79,300 (21.8 percent) questionnaires were returned. Of these, 77,738 questionnaires passed eligibility and quality control checks. Disposition of the questionnaires is given in table 1. We estimated that 93 percent of those persons on the mailing list would have been eligible for the study (table 1), yielding an estimated response rate of 22.9 percent among those eligible. Data and specimen collection Data were collected at baseline by using a 24-page, selfadministered, sex-specific, optically scanned questionnaire Am J Epidemiol 2004;159:83–93 VITamins And Lifestyle Cohort Study 85 that covered three content areas: supplement use, diet, and health history and risk factors. We plan to update this information at 5 years postbaseline. Measurement of supplement use. Supplement use was the principal exposure in this study; therefore, our supplement questionnaire section was considerably longer (six pages) and more detailed than typical supplement assessment forms used in large epidemiologic studies. This questionnaire was developed based on our earlier methodological work in supplement assessment (18–20). Respondents were asked about their use of supplements during the 10 years prior to baseline. For current multivitamin use, participants either selected one of 16 common brand names or provided dose information on each vitamin and mineral in the brand they used. For those who had used more than one brand over the 10 years or had used multivitamins only in the past, respondents selected from another list of brand names (reflecting past market availability). For analysis, the nutrient content of multivitamins was based on information from the Physicians’ Desk Reference for Nonprescription Drugs and Dietary Supplements (21) and from direct inquiry to manufacturers to determine composition of multivitamins in the past 10 years. Respondents then reported their intake of 10 vitamins and six minerals from all other mixtures not classified as multivitamins (e.g., “stress”/B complex or antioxidant mixtures) and single supplements and of 21 herbal and specialty supplements. We used a closed-ended format to inquire about current versus past use, frequency (times per week) and duration of use over the previous 10 years, and usual dose per day. For herbal and specialty supplements, questions on dose were not included because of a lack of accurate information on their potency. Average daily intake of supplemental vitamin E, C, and calcium over the previous 10 years was estimated by summing intake from individual supplements and intake from multivitamins during the 10-year period; intake from each nutrient was computed as number of years/10 × number of days per week/7 × dose. For multivitamin use, 10-year intake was expressed as daily pills over the 10 years (number of years/10 × number of days per week/7). In a study of 220 participants, the VITAL supplement questionnaire showed excellent reliability/validity when compared with a repeat administration of the questionnaire 3 months after baseline, to a detailed home interview and supplement inventory, and to nutrient biomarkers (22) (table 2). Diet and other covariates. Diet was assessed by using a food frequency questionnaire that captures intakes of 120 food and beverage items and includes adjustment questions on types of foods and preparation techniques. It was adapted from food frequency questionnaires we developed for the Women’s Health Initiative and other studies (23–25), with the addition of highly supplemented foods such as calciumsupplemented orange juice and fortified drinks and bars such as Slim·Fast (Slim·Fast Foods Company, West Palm Beach, Florida) and PowerBar (PowerBar Inc., Berkeley, California). The measurement properties of earlier versions of this questionnaire were published previously (25). The food frequency questionnaire analytic program, based on nutrient Am J Epidemiol 2004;159:83–93 TABLE 2. Validity and reliability of supplement use from the baseline questionnaire (n = 220),* VITamins And Lifestyle cohort study, Washington State, 2000–2002 Questionnaire 10-year dose test-retest ICC† Questionnaire current dose vs. interview/ transcription r‡ Questionnaire current dose vs. biomarker§ r‡,¶ Vitamin C 0.85 0.77 0.29 Vitamin E 0.87 0.81 0.69 Calcium 0.77 0.69 –0.07 Multivitamins 0.81 NA† NA Nutrient * Satia-Abouta J, Patterson RE, King IB, et al. Reliability and validity of self-report of vitamin and mineral supplement use in the Vitamins and Lifestyle study. Am J Epidemiol 2003;157:944–54. † ICC, intraclass correlation; NA, not applicable. ‡ Pearson correlation coefficient. § Plasma vitamin C, serum α-tocopherol, and spot urinary calcium, respectively. ¶ Adjusted for age, sex, race, and current smoking. In addition, plasma vitamin C was adjusted for dietary vitamin C; serum αtocopherol was adjusted for serum total cholesterol, body mass index, and dietary α-tocopherol; and urinary calcium was adjusted for urinary creatinine and dietary calcium. values from the Minnesota Nutrient Data System (26), yields estimated intakes of over 50 nutrients. The remaining parts of the questionnaire covered personal identifiers for tracking purposes, demographic characteristics, health history, drug use with emphasis on nonsteroidal anti-inflammatory drugs, physical activity over the 10 years prior to baseline, cancer screening practices, and other potential confounders of supplement-cancer associations. Health complaints, such as fatigue and joint pain, were also assessed because such nonmedically diagnosed health conditions may be the “indication” that led subjects to select certain supplements. Questionnaire quality control. Questionnaires were reviewed before scanning; correction tape was used on crossed-out answers, and written comments were read and coded. The data entry and scan (ScanTools; Pearson NCS, Bloomington, Minnesota) software we used contain editcheck capabilities (e.g., value ranges). We developed extensive postentry programmed edit checks, including comparing names from the questionnaire and the mailing lists to determine whether the correct person had completed the questionnaire, comparing questionnaire identifiers to spot duplicate questionnaires from the same person, checking that subjects completed the appropriate sexspecific questionnaire (if not, a subject was interviewed by telephone to ascertain the missing information), interpreting duplicate marks, and conducting extensive consistency checks between variables. Collection of buccal cells. Buccal cells were collected as a source of DNA for future gene-nutrient studies. To select the best procedures, we conducted a small study (n = 24) to compare DNA quality, DNA quantity, and costs of cytobrush and mouthwash methods (27). The cytobrush method 86 White et al. TABLE 3. Supplement use over 10 years prior to baseline by participants in the VITamins And Lifestyle cohort study, Washington State, 2000–2002 Supplement (units) Individual use of supplement (%) Median dose per day Median no. of years of use High-dose supplement use* over 10 years (%) Multivitamin (no. of pills/day) 65.5 1 8 53.9 Vitamin C (mg) 45.8 500 8 34.1 Vitamin E (mg of dl α-tocopherol) 47.2 400 5 30.6 Calcium (mg) 45.2 500 5 22.6 * High-dose use was defined as an average intake over 10 years of ≥5 multivitamin pills per week, ≥180 mg of vitamin C per day, ≥150 mg of vitamin E per day, or ≥350 mg of calcium per day. yielded sufficient DNA quality and quantity (12.0 µg, sufficient for an estimated 150–225 polymerase chain reactions requiring short- and intermediate-length DNA fragments) at a cost acceptable for this large-scale study ($8.50 per participant). Three months after the completed baseline questionnaire was received, a DNA kit was mailed containing a cover letter; three sterile cytology brushes, each packaged in a separate bar-coded tube; detailed instructions with pictures; and a consent form. A second mailing was sent to nonresponders approximately 1 year later. When returned, buccal brushes were logged into the specimen tracking system and were linked to the participant’s identity, and the name on the consent form was entered to ensure that the correct person provided specimens. The three brushes were stored in separate –80°C freezers. The response rate was 70 percent. Follow-up methods for cancer and censoring After discounting rates to account for a possible “healthy volunteer effect,” we expect approximately 1,000 cancers to be diagnosed in the cohort in each of the next 5 years, including an estimated 220 prostate, 155 breast, 130 lung, and 90 colorectal cancers. The censoring date (the last date for follow-up) for each participant will be the earliest of the date of death, date on which the participant moved out of the SEER catchment area, date on which he or she asked to be removed from the study, or most recent date that endpoints were ascertained. Information on cancers, deaths, and moves out of the area are obtained primarily from passive follow-up by using linkages to public databases. Cancer incidence. Incident cancer cases are ascertained by linking the study cohort to the western Washington SEER program cancer registry, maintained by the Fred Hutchinson Cancer Research Center under contract to the National Cancer Institute. Data are collected on all newly diagnosed cancers (except nonmelanoma skin cancers) occurring in residents of the 13 counties of the Seattle-Puget Sound area. Cases are ascertained through all hospitals in the area; offices of pathologists, oncologists, and radiotherapists; and state death certificates. Extensive quality control procedures ensure that registry data are accurate and complete. Linkage methods. We link the VITAL cohort to the SEER file annually. We have designed and implemented a comprehensive linkage system, which is largely automated, by using data items common to both sets of data. First, potential matches are identified based on linking the two files by nine increasingly broad (i.e., less specific) sets of matching criteria. Examples follow: 1) full social security number (provided by 33 percent of our cohort); 2) last four digits of the social security number (provided by 36 percent of our cohort), first five characters of the last name, and date of birth; and 3) “sounds like” (based on phonetic sound) last name and date of birth. Second, each potential match is ranked electronically to determine whether it is “good” (enough data items in common to be considered a match), “bad” (too few data items in common to be considered a match), or “needs visual inspection” (some data items in common). The ranking criteria are based on the type and number of other data items that match (name, sex, date of birth, street address, zip code, telephone number, marital status, and birthplace), with the criteria more stringent for the less specific linkage criteria. Third, matches requiring visual inspection are reviewed by staff using screens that display all relevant information from VITAL and SEER. This inspection allows the use of human judgment for matches not made electronically because of misspellings of names, nicknames, transposition of numbers, and so forth. Finally, for records for which the visual match is still uncertain, the participant is telephoned directly. Mortality. Deaths occurring in the cohort in Washington State are ascertained by linkage to the state death file and by using other methods. Our linkage system to the death file is almost identical to the system described above for linkage to the SEER file. In addition, unconfirmed deaths will be ascertained from annual linkage to the Social Security Death Index, from tracking those lost to follow-up, from the US Postal Service as a result of the 5-year mailings, and from family members who receive the 5-year mailings. We will link the unconfirmed deaths and all 5-year questionnaire nonresponders to the National Death Index. Confirmed or unconfirmed date of death will be the censoring date. Moves out of area. Subjects who move out of the 13county area are tracked by electronic linkage to the National Am J Epidemiol 2004;159:83–93 VITamins And Lifestyle Cohort Study 87 TABLE 4. Herbal and specialty supplement use in the VITamins And Lifestyle cohort study, Washington State, 2000–2002 Prevalence of use* (%) Type of supplement Men Women Glucosamine 14.0 19.0 Fiber 10.6 14.1 Chondroitin 8.7 13.0 Saw palmetto† 8.0 Ginkgo biloba 6.5 Fish oil/EPA‡/omega-3/cod liver oil 6.4 7.4 Garlic pills 7.3 6.8 7.9 * Use at least once a week for a year. † Asked of men only. ‡ EPA, eicosapentaenoic acid. Change of Address system. All change of address data submitted by people who relocate across the entire country are entered into this system. The resulting file is provided to licensed private companies and is updated every 2 weeks. We link the VITAL cohort to the National Change of Address file annually by using a vendor who contracts with the US Postal Service. If a name and address match exactly (last name for family moves and first and last names for individual moves), the new address and move date are provided. If a new address is not known, a move date is provided along with information about the move (e.g., moved, no forwarding address). If a record does not match exactly but enough data items are common to suggest the possibility of a move, codes are returned to indicate a “possible” move, but no new address is given. This system is estimated to identify about 65 percent of moves. Tracking those lost to follow-up. Participants considered “lost to follow-up” are determined primarily from 1) linkages to the National Change of Address system returned as a match for a move without a forwarding address or as a “possible move” but not an exact match, and 2) mailed data collection instruments returned by the US Postal Service as “undeliverable” or “moved, no forwarding address.” Intensive tracking procedures are conducted for these cohort members by using 1) public databases and directories owned by the Fred Hutchinson Cancer Research Center (e.g., Wash- TABLE 5. Odds ratios† for 10-year high-dose supplement use‡ by participants in the VITamins And Lifestyle cohort study, by participant characteristics (n = 76,072),§ Washington State, 2000–2002 Characteristic % Multivitamins Vitamin C Vitamin E Calcium Age (years) 50–59 46.0 Reference Reference Reference Reference 60–69 34.7 1.29 1.28 1.64 1.65 ≥70 19.3 1.39 1.34 1.81 1.94 0.001 0.001 0.001 0.001 p for trend Sex Male 48.3 Reference Reference Reference Reference Female 51.7 1.44*** 1.30*** 1.42*** 4.79*** Education High school or less 20.2 Reference Reference Reference Reference Some college 38.3 1.30 1.29 1.17 1.25 College degree 24.5 1.39 1.40 1.26 1.47 Advanced degree 17.0 1.45 1.51 1.36 1.50 0.001 0.001 0.001 0.001 p for trend Race/ethnicity White 93.2 Reference Reference Reference Reference Hispanic 0.9 0.64*** 0.71** 0.85 0.78* Black 1.3 0.64*** 0.65*** 0.54*** 0.44*** American Indian/Alaska Native 1.5 0.92 1.05 0.95 0.98 Asian/Pacific Islander 2.5 0.67*** 0.63*** 0.63*** 0.65*** * p < 0.05; ** p < 0.01; *** p < 0.001 for categorical variables. For ordinal variables, p for trend is given. † Based on logistic regression models adjusted for other characteristics in the table. ‡ High-dose use was defined as an average intake over 10 years of ≥5 multivitamin pills per week, ≥180 mg of vitamin C per day, ≥150 mg of vitamin E per day, or ≥350 mg of calcium per day. § Omitted were 114 late responders and 1,552 subjects for whom data on adjustment factors (education or race) were missing. Am J Epidemiol 2004;159:83–93 88 White et al. TABLE 6. Odds ratios† for 10-year high-dose supplement use‡ by participants in the VITamins And Lifestyle cohort study, by factors putatively associated with breast or prostate cancer, Washington State, 2000–2002 Risk factor % Multivitamins Vitamin C Vitamin E Calcium Risk factors for breast cancer in women (n = 39,301)§ Family history of breast cancer No 83.9 Reference Reference Reference Reference Yes 16.1 1.02 1.02 1.08* 1.03 Mammogram in the past 2 years No 9.0 Reference Reference Reference Reference Yes 91.0 1.52* 1.26* 1.43* 1.60* ≤24 58.6 Reference Reference Reference Reference 25–29 19.6 0.95 0.92 0.88 1.01 ≥30 8.9 0.82 0.80 0.71 0.90 Age at first birth (years) No births 12.9 p for trend 1.05 1.17 1.10 1.11 0.61 0.03 0.62 0.08 Hormone replacement therapy Never/former use 52.9 Reference Reference Reference Reference Current use 47.1 1.34* 1.25* 1.29* 1.41* Normal (<25) 41.1 Reference Reference Reference Reference Overweight (25–<30) 33.5 0.92 0.86 0.90 0.79 Body mass index (kg/m2) Obese (30–<35) 15.5 0.88 0.76 0.79 0.69 Very obese (≥35) 10.0 0.80 0.70 0.73 0.58 0.001 0.001 0.001 0.001 p for trend Alcohol intake (no. of drinks/ week) None 50.8 Reference Reference Reference Reference 1–2 18.4 0.97 0.97 0.99 1.00 3–6 14.2 0.97 1.10 1.13 1.09 ≥7 16.6 0.91 1.02 1.13 1.06 0.003 0.10 0.001 0.01 p for trend Table continues ington State Department of Licensing), 2) public records offices (e.g., county tax assessors’ offices), and 3) Internet sources and online database services (e.g., Autotrack; ChoicePoint Asset Company, Boca Raton, Florida). RESULTS Supplement use in the cohort We successfully recruited a cohort in which the level of supplement use was high and of long duration. Of multivitamins and the 16 individual vitamin and mineral supplements we inquired about, 21 percent of the cohort used none, 26 percent used one or two, 21 percent used three or four, and 32 percent used five or more. Of the main exposures of interest, 66 percent of participants used multivitamins, 46 percent used individual vitamin C, 47 percent used individual vitamin E, and 46 percent used calcium—each typically for 5–8 of the previous 10 years (table 3). The most common daily doses were 500 mg of vitamin C, 400 mg of vitamin E, and 500 mg of calcium. The last column of table 3 gives the percentage of the cohort whose 10-year average level of use of each supplement was high, defined as higher than could be obtained by 10-year daily use of common formulations of multivitamins. By identifying such users, we will be able to consider separately the effects of specific supplemental nutrients and the effects of multivitamins. According to this definition, 34 percent of our subjects used high doses of vitamin C (≥180 mg, or about twice the Recommended Dietary Allowance of 90 mg for men and 75 mg for women (3)), 31 percent used high doses of vitamin E (≥150 mg of dl α-tocopherol (150 IU), or five times the Recommended Dietary Allowance of 30 IU (4)), and 23 percent used high doses of calcium (≥350 mg, or about one third the adequate calcium intake of 1,200 mg (28)). Multivitamins generally contain 100 percent of the Recommended Dietary Allowance for those micronutrients for which there Am J Epidemiol 2004;159:83–93 VITamins And Lifestyle Cohort Study 89 TABLE 6. Continued Risk factor % Multivitamins Vitamin C Vitamin E Calcium Risk factors for prostate cancer in men (n = 36,771)¶ Family history of prostate cancer No 86.3 Reference Reference Reference Reference Yes 13.7 1.02 1.04 1.00 0.96 No 26.6 Reference Reference Reference Reference Yes 73.4 1.43* 1.37* 1.50* 1.32* Normal (<25) 27.5 Reference Reference Reference Reference Overweight (25–<30) 48.8 0.97 0.91 0.98 0.90 Obese (30–<35) 17.7 0.89 0.83 0.88 0.88 Very obese (≥35) 6.0 0.81 0.68 0.80 0.82 0.001 0.001 0.001 0.002 Prostate-specific antigen screening in the past 2 years Body mass index (kg/m2) p for trend Diet Energy from fat (%) >40 23.8 Reference Reference Reference Reference 36–40 25.4 1.13 1.14 1.18 1.00 31–35 24.4 1.35 1.31 1.47 1.18 ≤30 26.4 1.63 1.65 1.87 1.26 0.001 0.001 0.001 0.001 p for trend Fruit/vegetables (no. of servings/day) 0–2 28.1 Reference Reference Reference Reference >2–3 24.7 1.27 1.29 1.26 1.28 >3–5 25.2 1.31 1.48 1.50 1.47 >5 22.0 1.60 1.86 1.89 1.93 0.001 0.001 0.001 0.001 p for trend * p < 0.001 for categorical variables. For ordinal variables, p for trend is given. † Based on logistic regression models adjusted for age, education, and race/ethnicity. ‡ High-dose use was defined as an average intake over 10 years of ≥5 multivitamin pills per week, ≥180 mg of vitamin C per day, ≥150 mg of vitamin E per day, or ≥350 mg of calcium per day. § Omitted were 38 late responders and 1,006 subjects for whom data on adjustment factors (education or race) were missing. Those subjects for whom data for the risk factor of interest were missing (range, 0.6–6.1%) were also omitted. ¶ Omitted were 76 late responders and 546 subjects for whom data on adjustment factors (education or race) were missing. Those subjects for whom data for the risk factor of interest were missing (range, 1.4–7.4%) were also omitted. are recommendations, except for calcium and certain other minerals. On average, men and women consume about the Recommended Dietary Allowance of most nutrients from food (29), so the 54 percent who met our definition of use of high doses of multivitamins (≥5 pills per week for 10 years) had about twice the intake of many nutrients compared with nonusers of supplements. The VITAL participants also commonly used other supplements. Of the 13 other individual vitamins and minerals of interest, use was highest for folate and other B vitamins (14–17 percent), zinc (14 percent), and selenium (14 percent). Overall, about one third of participants used at least one herbal or specialty supplement. The most frequently used other supplements are given in table 4. Am J Epidemiol 2004;159:83–93 Participant characteristics and their association with supplement use The cohort is 52 percent female, the median age is 61 years, 42 percent of the participants are college graduates, and 93 percent are White (table 5). To identify factors that may confound supplement use-cancer relations in this observational study, we examined associations of demographic factors and cancer risk factors with supplement use. Table 5 shows associations of demographic factors with 10-year high-dose use (as defined above) of four specific supplements of initial major interest (multivitamins, vitamin C, vitamin E, and calcium). High-dose use of each supplement increased with age and education, and women were more 90 White et al. likely to be high-dose users of each supplement, particularly calcium. Whites showed higher levels of use of each supplement than other racial/ethnic groups except for Native Americans, who were similar to Whites in this regard. Tables 6 and 7 present associations of selected cancer risk factors for the four most common cancers with 10-year highdose use of the four main supplements of interest. Of the selected risk factors for breast cancer (table 6), prostate cancer (table 6), colorectal cancer (table 7), and lung cancer (table 7), all were statistically significantly associated with all four supplements except for family history of cancer and a woman’s age at first birth. Associations were particularly strong (odds ratios close to 2) for nonsteroidal antiinflammatory drug use (which includes low-dose aspirin commonly used for heart disease prevention), physical activity, and fruit and vegetable intake DISCUSSION A large and increasing number of Americans are using dietary supplements, despite limited scientific information to guide them. Therefore, we designed and implemented a cohort study to test the associations of supplement use with cancer risk. One advantage of this study is that participants’ use of supplements is considerably greater (more types, higher doses, and longer durations) than for those subjects in earlier cohorts (30, 31) because the present cohort was recruited recently and targeted supplement users. We also collected more detailed information on supplement use than has been collected for most other cohorts, and our validation study demonstrated good measurement properties of the supplement questionnaire (22). We will be able to examine cumulative dose over a long period of time (the 10-year period ending at baseline), which is critical considering the likelihood that 5–15 years of supplement use is needed to affect cancer risk (16, 17). The estimated response rate in the VITAL study was 23 percent. This low rate was probably due to the length of the questionnaire (24 pages), use of only a single mailing, and use of a recruitment letter that targeted supplement users to participate. Our cross-sectional analyses on correlates of supplement use (to identify confounding factors in this cohort) would not be generalizable because of this low response rate, although our findings are consistent with those from past studies (see below). However, selection bias is not a major concern in prospective studies, such as our future studies of supplement-disease associations, because participation generally cannot be jointly related to supplement use and to the participants’ unknown, future disease status. This study is limited, as are all observational studies, by possible confounding. Specifically, supplement users may be more health conscious and therefore more likely to practice other disease prevention behaviors (32). We therefore evaluated whether cancer risk or prevention factors were associated with long-term, high-dose use of each of the four main supplements of interest (multivitamins, vitamin C, vitamin E, and calcium) in our population. Most associations were in the direction that those participants with a possibly lower risk of cancer (those with a lower body mass index, lower dietary fat intake, higher fruit and vegetable intake, regular use of nonsteroidal anti-inflammatory drugs, greater recreational physical activity, and not a current smoker) were more likely to use each supplement. High-dose supplement use was also associated with cancer screening, which increases diagnoses of some cancers (e.g., prostate) and decreases diagnoses of others (e.g., colorectal) and with current use of hormone replacement therapy, which increases the risk of breast cancer but reduces the risk of colorectal cancer (33). To our knowledge, our finding that use of nonsteroidal anti-inflammatory drugs is associated with supplement use has not been reported previously. The remainder of our findings are consistent with those from earlier studies, even though our population is highly selfselected (34–42).. Conversely, supplement users may be less healthy; certain diseases or minor health complaints may have prompted them to use supplements. Failure to control for these factors may lead to “confounding by indication” (43). Gray et al. (40) found that supplement users were more likely to have multiple health care visits in the past year. We previously reported associations of supplement use with 21 health conditions assessed at baseline (44). However, in our study, only one cancer-related condition was found to be associated with supplement use: reflux disease with calcium supplements (44). To account for these health behaviors and medical conditions that may be confounding factors in future analyses from this study, we will statistically control for them in the analyses. In addition, we will assess the dose-response gradient of supplement use on the cancer outcome among users of only that supplemental nutrient. For example, users of supplemental vitamin C will be compared only with other users, by dose of vitamin C, reducing bias caused by comparing nonusers with users (45). This study uses the cost-effective approach of ascertaining endpoints and censoring dates by linkages to available cancer registry, death, and change-of-address files rather than by active contact with participants. Systems such as ours, with computerized linkage to identify close matches (rather than exact matches) and final decisions arrived at by human judgment, are considered the most efficient and reliable (46, 47). Other cohort studies also use a primarily passive follow-up system (48, 49). Two concerns about our passive follow-up system are possible selection biases from 1) censoring participants when they move out of the catchment area of the western Washington State SEER program cancer registry and 2) erroneously including some out-of-area subjects as in-area because of our inability to identify all moves. Censoring those who move out of the area is significantly cheaper than tracking the outcomes of those who have moved. The effect of outmigration is expected to be a 4 percent loss of personyears (50). Failure to capture all out-of-area moves with our tracking efforts would lead to a small error (an estimated 1 percent of person-years) in that we will include some subjects who moved out of the area as contributing personyears of observation. Censoring subjects when they move out of the area or erroneously including some out-of-area subjects as in-area would bias the relative risk to only a small degree, because these events are not likely jointly dependent Am J Epidemiol 2004;159:83–93 VITamins And Lifestyle Cohort Study 91 TABLE 7. Odds ratios† for 10-year high-dose supplement use‡ by participants in the VITamins And Lifestyle cohort study, by factors putatively associated with colorectal or lung cancer (n = 76,072),§ Washington State, 2000–2002 Risk factor % Multivitamins Vitamin C Vitamin E Calcium Risk factors for colorectal cancer Family history of colorectal cancer No 88.5 Reference Reference Reference Reference Yes 11.5 1.03 1.00 1.00 1.01 No 42.9 Reference Reference Reference Reference Yes 57.1 1.29* 1.23* 1.32* 1.39* No 73.5 Reference Reference Reference Reference Yes 26.5 1.67* 1.81*** 2.14* 1.68* Normal (<25) 34.4 Reference Reference Reference Reference Overweight (25–<30) 41.0 0.94 0.88 0.93 0.82 Obese (30–<35) 16.6 0.88 0.79 0.82 0.74 Very obese (≥35) 8.0 0.81 0.70 0.76 0.62 0.001 0.001 0.001 0.001 Reference Reference Reference Reference Sigmoidoscopy in the past 10 years Regular use of NSAIDs¶ in the past 10 years# Body mass index (kg/m2) p for trend Physical activity (MET¶-hours/week) None 15.0 1st tertile (0–<4.4) 28.3 1.36 1.32 1.36 1.27 2nd tertile (4.4–<13) 28.3 1.61 1.62 1.69 1.63 3rd tertile (≥13) 28.4 1.77 1.99 2.10 1.94 0.001 0.001 0.001 0.001 p for trend Fruit/vegetables (no. of servings/day) 0–2 22.8 Reference Reference Reference Reference >2–3 22.1 1.24 1.30 1.29 1.31 >3–5 25.1 1.34 1.51 1.45 1.47 >5 30.1 1.63 1.89 1.84 1.96 0.001 0.001 0.001 0.001 p for trend Risk factors for lung cancer Smoking Never 47.4 Reference Reference Reference Reference Former 44.3 1.08* 1.10* 1.12* 1.11* Current 8.3 0.75* 0.79* 0.76* 0.67* * p < 0.001 for categorical variables. For ordinal variables, p for trend is given. † Based on logistic regression models adjusted for age, sex, education, and race/ethnicity. ‡ High-dose use was defined as an average intake over 10 years of ≥5 multivitamin pills per week, ≥180 mg of vitamin C per day, ≥150 mg of vitamin E per day, or ≥350 mg of calcium per day. § Omitted were114 late responders and 1,552 subjects for whom data on adjustment factors (education or race) were missing. Those subjects for whom data for the risk factor of interest were missing (range, 0.5–8.9%) were also omitted. ¶ NSAIDs, nonsteroidal anti-inflammatory drugs (aspirin, ibuprofen, or naproxen); MET, metabolic equivalent. # Regular use was defined as ≥4 times per week for ≥4 years. on exposure and outcome in a cohort study. Furthermore, the error in terms of person-years is sufficiently small so that even a nonindependent association would bias the relative risk by a small degree (51). Note that we retain the ability to identify cancers and deaths of subjects who move within the Am J Epidemiol 2004;159:83–93 catchment area whether or not we have tracked their current address correctly. Future results of this study on supplement-cancer associations may have scientific and public health implications. It has been suggested that cohort studies may be the most 92 White et al. rational way to choose specific agents to test in future clinical trails (52). The results of this study may also help guide persons who choose to use supplements toward those that have benefits. In addition, if some supplements are found to have no effect or to be harmful, this information will be important to the approximately 100 million US adults who use them. 13. 14. 15. ACKNOWLEDGMENTS This study was supported by a grant (R01CA74846) from the National Cancer Institute. The authors thank Anne Oswald and Melissa Mouton for their contributions to data collection for the study and Kayla Stratton for the statistical analyses for this report. 16. 17. 18. 19. REFERENCES 1. Blendon RJ, DesRoches CM, Benson JM, et al. Americans’ views on the use and regulation of dietary supplements. Arch Intern Med 2001;161:805–10. 2. Institute of Medicine. Vitamin E. Dietary reference intakes for vitamin C, vitamin E, selenium, and carotenoids. Washington, DC: National Academy Press, 2000:186, 190–2, 196, 416. 3. Institute of Medicine. Vitamin C. Dietary reference intakes for vitamin C, vitamin E, selenium, and carotenoids. Washington, DC: National Academy Press, 2000:154, 416. 4. Institute of Medicine. Vitamin A. Dietary reference intakes for vitamin A, vitamin K, arsenic, boron, chromium, copper, iodine, iron, manganese, molybdenum, nickel, silicon, vanadium, and zinc. Washington, DC: National Academy Press, 2002:92–3. 5. World Cancer Research Fund. Food, nutrition and the prevention of cancer: a global perspective. Washington, DC: American Institute for Cancer Research, 1997. 6. Patterson RE, White E, Kristal AR, et al. Vitamin supplements and cancer risk: the epidemiologic evidence. Cancer Causes Control 1997;8:786–802. 7. Patterson RE, Kristal AR, Neuhouser ML. Vitamin supplements and cancer risk: epidemiologic research and recommendations. In: Bendich A, Deckelbaum RJ, eds. Primary and secondary preventive nutrition. Totowa, NJ: Humana Press, Inc, 2001. 8. The effect of vitamin E and beta carotene on the incidence of lung cancer and other cancers in male smokers. The AlphaTocopherol, Beta Carotene Cancer Prevention Study Group. N Engl J Med 1994;330:1029–35. 9. Omenn GS, Goodman GE, Thornquist MD, et al. Effects of a combination of beta carotene and vitamin A on lung cancer and cardiovascular disease. N Engl J Med 1996;334:1150–5. 10. Heinonen OP, Albanes D, Virtamo J, et al. Prostate cancer and supplementation with alpha-tocopherol and beta-carotene: incidence and mortality in a controlled trial. J Natl Cancer Inst 1998;90:440–6. 11. Clark LC, Combs GF Jr, Turnbull BW, et al. Effects of selenium supplementation for cancer prevention in patients with carcinoma of the skin. A randomized controlled trial. Nutritional Prevention of Cancer Study Group. JAMA 1996;276: 1957–63. 12. Albanes D, Malila N, Taylor PR, et al. Effects of supplemental alpha-tocopherol and beta-carotene on colorectal cancer: 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. results from a controlled trial (Finland). Cancer Causes Control 2000;11:197–205. Bostick RM, Potter JD, McKenzie DR, et al. Reduced risk of colon cancer with high intake of vitamin E: the Iowa Women’s Health Study. Cancer Res 1993;53:4230–7. White E, Shannon JS, Patterson RE. Relationship between vitamin and calcium supplement use and colon cancer. Cancer Epidemiol Biomarkers Prev 1997;6:769–74. Baron JA, Beach M, Mandel JS, et al. Calcium supplements for the prevention of colorectal adenomas. Calcium Polyp Prevention Study Group. N Engl J Med 1999;340:101–7. Giovannucci E, Stampfer MJ, Colditz GA, et al. Multivitamin use, folate, and colon cancer in women in the Nurses’ Health Study. Ann Intern Med 1998;129:517–24. Jacobs EJ, Connell CJ, Patel AV, et al. Multivitamin use and colon cancer mortality in the Cancer Prevention Study II cohort (United States). Cancer Causes Control 2001;12:927–34. Patterson RE, Neuhouser ML, White E, et al. Measurement error from assessing use of vitamin supplements at one point in time. Epidemiology 1998;9:567–9. Patterson RE, Kristal AR, Levy L, et al. Validity of methods used to assess vitamin and mineral supplement use. Am J Epidemiol 1998;148:643–9. Patterson RE, Levy L, Tinker LF, et al. Evaluation of a simplified vitamin supplement inventory developed for the Women’s Health Initiative. Public Health Nutr 1999;2:273–6. Physicians’ desk reference for nonprescription drugs and dietary supplements. Montvale, NJ: Medical Economics Company, 2002. Satia-Abouta J, Patterson RE, King IB, 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: 944–54. Kristal AR, Feng Z, Coates RJ, et al. Associations of race/ ethnicity, education, and dietary intervention with the validity and reliability of a food frequency questionnaire: the Women’s Health Trial Feasibility Study in Minority Populations. Am J Epidemiol 1997;146:856–69. (Comments published in Am J Epidemiol 1998;148:819–20). Kristal AR, Patterson RE, Neuhouser ML, et al. Olestra Postmarketing Surveillance Study: design and baseline results from the sentinel site. J Am Diet Assoc 1998;98:1290–6. Patterson RE, Kristal AR, Tinker LF, et al. Measurement characteristics of the Women’s Health Initiative food frequency questionnaire. Ann Epidemiol 1999;9:178–87. Schakel SF, Buzzard IM, Gebhardt SE. Procedures for estimating nutrient values for food composition databases. J Food Comp Anal 1997;10:102–14. King IB, Satia-Abouta J, Thornquist MD, et al. Short communication: buccal cell DNA yield, quality, and collection costs. Comparison of methods for large-scale studies. Cancer Epidemiol Biomarkers Prev 2002;11:1130–3. Institute of Medicine. Calcium. Dietary reference intakes for calcium, phosphorus, magnesium, vitamin D, and fluoride. Washington, DC: National Academy Press, 1999:71, 389. Murphy SP, Rose D, Hudes M, et al. Demographic and economic factors associated with dietary quality for adults in the 1987–88 Nationwide Food Consumption Survey. J Am Diet Assoc 1992;92:1352–7. Kushi LH, Fee RM, Sellers TA, et al. Intake of vitamins A, C, and E and postmenopausal breast cancer. The Iowa Women’s Health Study. Am J Epidemiol 1996;144:165–74. Wu K, Willett WC, Chan JM, et al. A prospective study on supplemental vitamin E intake and risk of colon cancer in women and men. Cancer Epidemiol Biomarkers Prev 2002;11:1298– 304. Am J Epidemiol 2004;159:83–93 VITamins And Lifestyle Cohort Study 93 32. Block G, Cox C, Madans J, et al. Vitamin supplement use, by demographic characteristics. Am J Epidemiol 1988;127:297– 309. 33. Rossouw JE, Anderson GL, Prentice RL, et al. Risks and benefits of estrogen plus progestin in healthy postmenopausal women: principal results from the Women’s Health Initiative randomized controlled trial. JAMA 2002;288:321–33. (Comment/summary for patients published in CMAJ 2002 Aug 20; 167:377–8). 34. Patterson RE, Neuhouser ML, White E, et al. Cancer-related behavior of vitamin supplement users. Cancer Epidemiol Biomarkers Prev 1998;7:79–81. 35. Slesinski MJ, Subar AF, Kahle LL. Dietary intake of fat, fiber and other nutrients is related to the use of vitamin and mineral supplements in the United States: the 1992 National Health Interview Survey. J Nutr 1996;126:3001–8. 36. Ervin RB, Wright JD, Kennedy-Stephenson J. Use of dietary supplements in the United States, 1988–94. Vital Health Stat 11 1999;Jun:i-iii, 1–14. 37. Lyle BJ, Mares-Perlman JA, Klein BE, et al. Supplement users differ from nonusers in demographic, lifestyle, dietary and health characteristics. J Nutr 1998;128:2355–62. 38. Kirk SF, Cade JE, Barrett JH, et al. Diet and lifestyle characteristics associated with dietary supplement use in women. Public Health Nutr 1999;2:69–73. 39. Hoggatt KJ, Bernstein L, Reynolds P, et al. Correlates of vitamin supplement use in the United States: data from the California Teachers Study cohort. Cancer Causes Control. 2002;13: 735–40. 40. Gray SL, Hanlon JT, Fillenbaum GG, et al. Predictors of nutritional supplement use by the elderly. Pharmacotherapy 1996; 16:715–20. 41. Neuhouser ML, Kristal AR, Patterson RE, et al. Dietary supplement use in the Prostate Cancer Prevention Trial: implications for prevention trials. Nutr Cancer. 2001;39:12–18. 42. Foote JA, Murphy SP, Wilkens LR, et al. Factors associated Am J Epidemiol 2004;159:83–93 43. 44. 45. 46. 47. 48. 49. 50. 51. 52. with dietary supplement use among healthy adults of five ethnicities: The Multiethnic Cohort Study. Am J Epidemiol 2003; 157:888–97. Psaty BM, Koepsell TD, Lin D, et al. Assessment and control for confounding by indication in observational studies. J Am Geriatr Soc 1999;47:749–54. Satia-Abouta J, Kristal AR, Patterson RE, et al. Dietary supplement use and medical conditions. Am J Prev Med 2003;24:43– 51. Breslow NE, Day NE, eds. Statistical methods in cancer research. Vol 2. The design and analysis of cohort studies. Lyon, France: International Agency for Research on Cancer, 1987. (IARC scientific publication no. 82). Roos LL, Wajda A. Record linkage strategies. Part I: estimating information and evaluating approaches. Methods Inf Med 1991;30:117–23. Shannon HS, Jamieson E, Walsh C, et al. Comparison of individual follow-up and computerized record linkage using the Canadian Mortality Data Base. Can J Public Health 1989;80: 54–7. Zheng W, Anderson KE, Kushi LH, et al. A prospective cohort study of intake of calcium, vitamin D, and other micronutrients in relation to incidence of rectal cancer among postmenopausal women. Cancer Epidemiol Biomarkers Prev 1998;7:221–5. Parikh-Patel A, Allen M, Wright WE. Validation of selfreported cancers in the California Teachers Study. Am J Epidemiol 2003;157:539–45. 1992 population trends for Washington State. Olympia, WA: Washington State Office of Financial Management, 1992. Armstrong BK, White E, Saracci R. Principles of exposure measurement in epidemiology. Oxford, United Kingdom: Oxford University Press, 1994. Taylor PR, Albanes D, Tangrea JA. To supplement or not to supplement, that is the question. Cancer Causes Control 1997; 8:685–7.