Validity of Methods Used to Assess Vitamin and Mineral Supplement Use

Transcription

1 American Journal of Epidemiology Copyright O 1998 by The Johns Hopkins University School of Hygiene and Public Health All rights reserved Vol. 148, No. 7 Printed In U.SA. Validity of Methods Used to Assess Vitamin and Mineral Supplement Use Ruth E. Patterson, 12 Alan R. Kristal, 12 Usa Levy, 1 Dale McLerran, 1 and Emily White 12 Assessing vitamin and mineral supplement use is important because supplement use per se is an exposure of interest for the risk of several chronic diseases and because supplements contribute a large proportion of total (diet plus supplement) micronutrient intake, another important exposure in epidemioiogic research. Unfortunately, little is known about methods for obtaining valid information about supplement use. The authors conducted a validation study in 1996 comparing supplement data collected in a telephone interview and from a self-administered questionnaire with data derived from a detailed, in-person interview and of the labels of supplement bottles (i.e., a gold standard) among adult supplement users in Washington State (n = 104). Spearman correlation coefficients comparing average daily supplemental vitamin and mineral intake from the interview or questionnaire with the gold standard ranged from 0.76 (95% confidence interval ) for vitamin C to 0.08 (95% confidence interval to 0.29) for iron, with a mean of about 0.5. The principal sources of error were inaccurate assumptions about the micronutrient composition of multiple vitamins and respondent confusion regarding the distinction between multiple vitamins and single supplements. These results suggest that commonly used epidemioiogic methods of assessing supplement use may incorporate significant amounts of error in estimates of some nutrients. Am J Epidemiol 1998;148: epidemioiogic methods; nutrition assessment; questionnaires; vitamins Epidemioiogic studies play an important role in understanding the relation between dietary factors and chronic disease (1). In particular, many observational studies have found inverse associations of micronutrient intake with disease, including antioxidants with cardiovascular disease (2, 3) and cancer (3, 4), folate with cardiovascular disease (5), and calcium with colon cancer (6) and osteoporosis (7). One of the biggest impediments to epidemioiogic research on diet and health has been the exceptionally difficult challenge of accurately assessing dietary intake (8). For this reason, the reliability and validity of dietary assessment instruments have been intensively studied. However, to obtain a measure of total micronutrient intake, it is also necessary to collect data on supplement use so that micronutrients from supplements can be summed with those from foods (9). Assessing micronutrient intake from supplements is Received for publication October 7, 1997, and accepted for publication March 4, Abbreviations: Cl, confidence interval; FFQ, food frequency questionnaire. 1 Cancer Prevention Research Program, Fred Hutchinson Cancer Research Center, Seattle, WA. 2 Department of Epidemiology, School of Public Health and Community Medicine, University of Washington, Seattle, WA. Reprint requests to Dr. Ruth E. Patterson, Cancer Prevention Research Program, Fred Hutchinson Cancer Research Center, 1100 Fairview Avenue North MP 702, P. O. Box 19024, Seattle, WA important because of the high prevalence of their use (10) and the large amount of some nutrients in supplements compared with that which can be obtained from food (9). There is also interest in assessing the effect of vitamin supplements per se on cancer risk (11, 12), cardiovascular disease (13), and other diseases (14). Unfortunately, little is known about methods of obtaining valid information about supplement use. Epidemioiogic studies typically use personal interviews or self-administered questionnaires to obtain information on 3-5 general classes of multiple vitamins and on single supplements, the dose of single supplements, and, sometimes, frequency and/or duration of use. These short and simple assessment methods were developed because collecting detailed data on supplement use presents many difficulties: More than 3,400 vitamin and mineral preparations are available to consumers (15), and these formulations change over time. Even if respondents were able to recall or record the exact name, brand, and dosage of their supplements, we know of no readily accessible and well-maintained database for over-the-counter supplements like that available for foods (16) and prescription drugs (17). Therefore, for large-scale studies, it is not practical to perform a supplement inventory as is recommended for prescription drugs (17). Block et al. (9) asserted that collection of vitamin 643

2 644 Patterson et al. supplement data need not be extensive to be useful. However, there has been no research on vitamin and mineral supplement assessment methodology, and therefore, little is known about the sources, types, and magnitude of errors that can occur when assessing micronutrient intake from supplements. The objective of this study is to investigate the accuracy of epidemiologic methods used to assess supplement use. Specifically, we compare supplement data collected in a telephone interview and from a self-administered questionnaire with data derived from transcribing the labels of supplement bottles (i.e., a gold standard). We address whether respondents can accurately report the type of supplement taken and the bias and precision of micronutrient estimates obtained from supplement use questionnaires. We also examine the impact of misclassification of micronutrient intake from supplements on misclassification of total micronutrient intake (from diet and supplements) because total intake is often the exposure of interest in studies of diet and disease. MATERIALS AND METHODS Subjects and recruitment We recruited our study sample from participants in the Washington State Cancer Risk Behavior Survey, a random-digit-dial survey of adults (aged 18 years or older) to monitor cancer risk behavior. The conservatively estimated effectiveness rate (completed interviews divided by known plus estimated eligible participants) was 63.5 percent (18). Details of sample selection and methods have been published previously (19). Approximately 1-2 months after the completion of the Cancer Risk Behavior Survey, we contacted survey participants who indicated that they took vitamin or mineral supplements at least three times a week. Because the study protocol required a clinic visit, we restricted our sample to adults living in the Seattle metropolitan area (King County, Washington). We excluded anyone taking only herbal supplements and women taking prenatal vitamins. Of the 192 eligible supplement users, we recruited 104 (54.2 percent), we were unable to contact 24 (12.5 percent), and 64 (33.3 percent) refused to participate or did not complete the study. Participants were paid $25.00 as a reimbursement for their time and travel expenses. Data collection Study procedures. After we contacted participants and ascertained eligibility and willingness to participate, we administered a vitamin supplement questionnaire over the telephone and scheduled a clinic appointment. Immediately after the interview, we mailed participants a self-administered supplement questionnaire and a food frequency questionnaire (FFQ). At the clinic visit, we conducted an in-person interview in which we collected data on frequency and duration of current supplement use and photocopied the participants' supplement bottles. Supplement use assessments. We assessed supplement use with an interview and a self-administered questionnaire because these are the most commonly used methods of obtaining self-reported data from participants (20). The interview and the questionnaire collected the same data, but differed in administration (interview or self-administered) and format (open or close ended). For the telephone interview, we asked questions on whether, over the previous year, they had taken multiple vitamins, such as One-a-Day (Miles, Inc., Elkhart, Indiana); stress multivitamin, B-complex, or antioxidant mixtures, such as Protegra (Lederle, Wayne, New Jersey); single vitamin supplements, such as vitamin A, C, E, or niacin; and single mineral supplements, such as calcium, iron, selenium, or zinc. When participants answered yes to a question, we ascertained the type(s) of multiple or single supplements) and the frequency (times per week) and duration (months or years) of use. We asked for the dose of single supplements, which was recorded exactly. Participants were instructed to respond from memory. The self-administered supplement questionnaire was similar to that used in the National Cancer Institute/Block Health Habits Questionnaire (9). The questionnaire asked about multiple vitamins, such as One-a-Day (Miles, Inc.); stress multivitamin, B- complex, or antioxidant mixtures, such as Protegra (Lederle); vitamin C, vitamin E, calcium, and zinc (capturing dose information); and vitamin A, vitamin D, folic acid, niacin, iron, and selenium (no dose information). For multiple vitamins and the commonly consumed single supplements (vitamin C, vitamin E, calcium, and zinc), we assessed frequency of use with a closed-ended response format: 1) one per week, 2) 2-6 per week, 3) one per day, and 4) two or more per day; and duration: less than 6 months or 6 months or more. We also used close-ended questions to assess the dose of vitamin C (250, 500, or 1,000 mg), vitamin E (250, 400, or 1,000 IU), calcium (150, 500, or 1,000 mg), and zinc (15, 50, or 100 mg). For the infrequently used, single supplements (vitamin A, vitamin D, folic acid, niacin, iron, and selenium), we asked participants only if they took them three times per week or more. Participants were instructed to look at their supplement bottles when completing this form. Gold standard. During the in-person interview, we

3 Vitamin Supplement Assessment 645 photocopied the supplement bottle labels and used a scripted interview to assess frequency (times per week) and duration of supplement use over the previous year with probes for changes in routine (e.g., vacation and illness). After the interview, we classified the type of multiple vitamin taken and transcribed the nutrient data from the photocopies of the supplement bottles. Micronutrient intake from supplements. For each of the three methods of assessing supplement use, we calculated average daily dose of micronutrients from supplements over the previous year by summing across multiple vitamins and single supplements after adjustment for frequency and duration of use over the previous year. We calculated these variables for the most frequently used single supplements (vitamin C, vitamin E, and calcium) and for a set of micronutrients of scientific interest (folic acid, iron, and selenium). For the telephone interview and the questionnaire, we made the following assumptions about the formulations of the multiple vitamins: once-a-day type multivitamins without minerals contained 60 mg vitamin C, 30 IU vitamin E, and 400 pig folic acid; once-a-day type with minerals contained vitamin C, vitamin E, and folic acid as above plus 162 mg calcium, 18 mg iron, and 20 /i-g selenium; stress supplements contained 500 mg vitamin C, 30 IU vitamin E, and 400 /xg folic acid; B-complex contained 400 /ig folic acid; and antioxidant mixtures contained 250 mg vitamin C, 200 IU vitamin E, and 15 mg selenium. When the singlesupplement dose was unknown or missing, we used the following defaults: 500 mg vitamin C, 400 IU vitamin E, 400 yug folic acid, 500 mg calcium, 50 mg iron, and 50 /xg selenium. These assumptions were based on the composition of leading brands (21), characteristics of supplement products in the United States (15), and informal surveys of vitamin supplements available at drug stores in King County, Washington. For the closed-ended questions on the selfadministered questionnaire, we made the following assumptions about frequency and duration of use. We coded frequency of use as "4 pills per week" for participants who indicated that they took supplements 2-6 times per week. For duration, less than 6 months was coded as 3 months and more than 6 months was coded as 1 year. Finally, for the infrequently used supplements for which we ascertained only that participants took them three or more times per week (folic acid, iron, and selenium), we assumed that participants had taken them daily for the previous year. FFQ. We assessed diet using an FFQ developed at the Fred Hutchinson Cancer Research Center (22). In format, this FFQ is similar to the National Cancer Institute/Block FFQ (23) with 15 adjustment questions on types of foods and preparation techniques, 99 food items with questions on frequency of use and portion size, and two summary questions on the usual intake of fruits and vegetables. The nutrient database used to convert food frequency information into nutrients is from the University of Minnesota's Nutrition Coding Center database (16). Data analysis We used the kappa statistic to assess agreement between the types of supplements taken as assessed by interview and self-administered questionnaire compared with the label (gold standard). Data on micronutrient intake from supplements were highly skewed and nonnormal regardless of data transformation. Therefore, to examine the bias between the assessment methods, we present the percentage of the sample taking supplemental vitamins and minerals and the median intake among users by each method. We calculated Spearman correlation coefficients and their 95 percent confidence intervals (24) to assess the precision of micronutrient intakes from the interview and the self-administered questionnaire compared with the. We calculated the correlation coefficients for both supplemental and total (diet plus supplements) vitamin and mineral intakes. RESULTS The mean age of our participants was 44 years (standard deviation, 13), 57 percent were female, and 91 percent were Caucasian (table 1). Our study sample was similar to the random sample of supplement users in King County that comprised our sampling frame and to Washington State residents (supplement users and nonusers), except for education distribution. The study sample was well educated, with over 50 percent of adults having 16 or more years of education compared with 45 percent of the King County supplement users and 30 percent of the Washington State residents. Seventy-seven percent of participants took a oncea-day type multivitamin with or without minerals (the most common multiple vitamin). The most commonly used single supplements were vitamin C (52 percent), vitamin E (34 percent), and calcium (10 percent) (table 2). The agreement (kappa) between the interview and the label for type of multiple vitamin ranged from 0.14 for other supplement mixtures to 0.92 for once-a-day multivitamins; single supplements ranged from 0.43 for calcium to 0.91 for vitamin E. Agreement was slightly lower for the selfadministered questionnaires and ranged from 0.46 for other supplement mixtures to 0.68 for once-a-day mul-

4 646 Patterson et al. TABLE 1. Characteristics of study sample (adult supplement users in King County Washington) compared with random samples of supplement user* in King County and Washington State residents, 1996 Characteristic Age (years) Females Race/ethnicity Caucasian Asian/Pacific Islander Other Education (years) Study sample (n»104) * Adjusted to intracensal state census. Supplement Washington users tn State King County residents (n - 264) (n = 2,450)' (%) (*) tivitamins and from 0.36 for calcium in single supplements to 0.70 for vitamin E. There was no evidence of systematic bias in estimates of average supplemental vitamin or mineral intakes from either the interview or the selfadministered questionnaire (table 3). Median micronutrient estimates from the interview and questionnaire were within ±10 percent of the label for folic acid, iron, and selenium and were generally within 25 percent for vitamin C, vitamin E, and calcium. Correlation coefficients comparing vitamin and mineral intake from supplements assessed by the interview with label ranged from 0.27 for iron to 0.75 for vitamin C, with a mean of 0.50 (table 4). Correlation coefficients comparing the questionnaire and label ranged from 0.08 for iron to 0.76 for vitamin C, with a mean of Correlation coefficients for total micronutrient intake (from supplements and diet) were appreciably higher, with a mean of 0.77 for the interview and 0.71 for the questionnaire. Age, gender, and education had no statistically significant effect on these between method correlation coefficients (data not shown). DISCUSSION The accuracy of epidemiologic methods in assessment of micronutrient intakes varied from good (approximately 0.75 for vitamin C and E) to poor (about zero for iron). We know of no research on the validity of vitamin supplement assessment with which to compare our findings. In FFQ validation studies, betweenmethod correlation coefficients for these micronutrients from food have generally ranged from 0.4 to 0.6 and, including supplemental intakes, generally increased the correlation coefficients (23, 25-27). The higher correlation coefficients may result from greater accuracy in estimates of supplement use than from food intake or may reflect increased variability in nutrients intake, which would increase the magnitude of the correlation coefficients. One major limitation in assessment instruments is inaccuracy in assumptions regarding the micronutrient composition of multiple vitamins. For example, approximately half of the once-a-day multivitamins without minerals and half of the stress supplements contained iron, even though iron is not usually considered a constituent of these multiple vitamins. In addition, the amount of iron within the multiplevitamin classes varied considerably. Block et al. (9) recommended that questionnaires distinguish between once-a-day vitamins and therapeutic types because the former contain approximately 60 mg vitamin C, while the latter approximately 200 mg (9). However, we did not find uniform formulations for high-dose multiple vitamins. About one third of the once-a-day multivitamins with minerals had nonstandard formulations with large or small amounts of vitamins and minerals. For example, vitamin C ranged from 30 to 1,000 mg, vitamin E from 30 to 400 IU, calcium from zero to 1,000 mg, iron from zero to 50 mg, and selenium from zero to 200 /xg, and these micronutrients did not covary in a consistent fashion. In addition, the labels on many once-a-day type multiple-vitamins (e.g., Centrum (Lederle)) indicate that they are "high-potency" or "therapeutic," while many of the high-dose formulations labeled themselves "daily" or "complete." It is unlikely that respondents could classify their multivitamins as once-a-day or therapeutic, and even if they could, there appears to be no accurate way to define the micronutrient composition of so-called therapeutic multiple vitamins. The distinction between multiple vitamins and single supplements is critical for accurate assessment (9). However, this distinction was particularly problematic for calcium supplements. In this study, only 10 percent of the sample took single calcium supplements (or Turns; SmithKline Beecham, Philadelphia, Pennsylvania), while over 20 percent took calcium mixed with vitamin D, magnesium, and/or zinc. Therefore, a question about use of calcium in single supplements is ambiguous about whether calcium in mixtures should be recorded. This ambiguity is reflected in the low agreement between reported use of single calcium supplements versus actual use and in the low agree-

5 Vitamin Supplement Assessment 647 TABLE 2. Agreement between type of vitamin/mineral supplement assessed by telephone interview and self-administered questionnaire compared with label (i.e., gold standard) (n» 104), 1996 Supplement Multivitamint Stress/B complex Other multivitamins Single supplements Vitamin C Vitamin E Calcium Supplement users (%)* Interview vs. Kappa %Clt 0.83 to to to to to to 0.67 Questionnaire vs. Kappa 95% a to to to to to to 0.53 * Assessed by label. t Cl, confidence interval. i Once-a-day type multivitamin with or without minerals. All other mixtures of vitamins and/or minerals, such as antioxtdants or calciumanagnesium/zinc. TABLE 3. Nutrient intakes from vitamin supplements (summed across all supplements) assessed by label (I.e., gold standard), telephone interview, and self-administered questionnaire (n o 104), 1996 Nutrient Vitamin C (mg) Vitamin E (III) Folic acid ( ig) Calcium (mg) Iron (mg) Selenium ( ig) Users (%) Transcription Median* Users (%) Interview Median* Median intakes among participants taking supplements) containing the nutrient Users (%) Questionnaire Median* TABLE 4. Spearman correlation coefficients and 95% confidence intervals between micronutrient intakes from vitamin supple ments (summed across all supplements) assessed by telephone Interview and self-administered questionnaire compared with label, 1996 Vitamin C Vitamin E Folic acid Calcium Iron Selenium No.t Interview vs. r 95% at Supplement nutrient kitake 0.65 to to to to to to 0.50 No.t Questionnaire vs. r 95% a to to to to to to 0.53 r Interview vs. 95% Cl 0.76 to to to to to to 0.91 Total nutrient intake' Questionnaire vs. r 95% at to to to to to to 0.86 * From supplements plus diet (diet assessed by food frequency questionnaire) (n = 104). t Participants taking supplement(s) containing the nutrient assessed by either instrument, t Cl, confidence interval. ment between method correlation coefficients for calcium compared with vitamin C or E. Another potential source of error in assessing supplemental vitamin and mineral intakes derives from mistakes by respondents. However, we found that participants were fairly accurate in classifying their supplements as once-a-day multivitamins, stress/bcomplex mixtures, and single supplements of vitamins C and E. There were instances in which participants incorrectly reported that they took single supplements (often reporting use of all of the single supplements), and we hypothesize that they were reporting on the micronutrients contained in their multiple vitamins. It is interesting to note that correlation coefficients for the telephone interview, where participants answered from memory, were not appreciably different than for the questionnaire completed in participants' homes, where they were instructed to review their supplement bottles. Therefore, we conclude that participants are able to recall from memory the type and dose of

6 648 Patterson et al. supplements that they take. A limitation of this study is that results can be generalized only to people who are motivated to participate in a research study requiring completion of questionnaires and a clinic visit. Nevertheless, this was a population-based sample and may be quite representative of vitamin supplement users. A strength of the study is the quality of the reference data. Since we were able to observe directly the supplements taken by our participants, our gold standard was probably much closer to the truth than the reference measures typically used in dietary assessment validation studies (e.g., food records or dietary recalls). The analyses presented here addressed how accurately epidemiologic methods of assessing supplement use rank supplement users along the distribution of micronutrient intake. The low correlation coefficients for calcium, iron, and selenium could seriously attenuate measures of associations in studies investigating the effects of supplement use on disease risk. Inclusion of participants who do not take supplements will increase between method correlation coefficients because the coefficients include the ability of the assessment instrument to distinguish between respondents who do and those who do not takes supplements, which generally would be high. Therefore, in study populations in which supplement use is low, the validity would be significantly greater than that presented here, which would primarily reflect the ability to classify users versus nonusers. We also found higher between-method correlation coefficients for total micronutrient intake (diet plus supplement) than for supplement intake alone. This was to be expected because the same measure of diet was used for each of the three supplement assessments. Nonetheless, these results suggest that in studies of total nutrient intake in which the majority of a micronutrient comes from foods as opposed to supplements (e.g., calcium), the assessment error will be reduced. Therefore, the degree to which exposure measurement error will attenuate measures of association depends on the assessment instrument, the nutrient being assessed, the research question (supplement use vs. total micronutrient intake), and the proportion of the study population that use supplements. Supplement assessment instruments need to be more carefully designed to eliminate ambiguity and to help participants answer questions on multiple vitamins versus single supplements. However, it is difficult to design supplement assessment instruments without population-level data on supplement use. Information on the most commonly used brands of multiple vitamins would enable researchers to decide which multiple vitamins to ask about and to choose the most representative vitamin and mineral formulation (i.e., nutrient database) for supplements. In addition, population-level data on the percentage of micronutrients obtained from different types of supplements (e.g., percentage of total supplemental calcium obtained from calcium/magnesium mixtures) would allow investigators to design assessment instruments that capture accurate information on specific supplemental vitamins or minerals. ACKNOWLEDGMENTS Supported by National Cancer Institute grants P01 CA and R01 CA REFERENCES 1. National Research Council, Committee on Diet and Health, Food and Nutrition Board, Commission on Life Sciences. Diet and health: implications for reducing chronic disease risk. Washington, DC: National Academy of Sciences, Gey KF. Cardiovascular disease and vitamins. Concurrent correction of "suboptimal" plasma antioxidant levels may, as important part of "optimal" nutrition, help to prevent early stages of cardiovascular disease and cancer, respectively. Bibl NutrDieta 1995;52: Diplock AT. Antioxidant nutrients and disease. Nutr Health 1993;9: Giacosa A, Filiberti R, Hill MJ, et al. Vitamins and cancer prevention. Eur J Cancer Prev 1997;6 (Suppl.):S47-S Boushay CJ, Beresford SAAB, Omenn GS, et al. A quantitative assessment of plasma homocysteine as a risk factor for vascular disease. JAMA 1995;274: Slattery ML, Sorenson AW, Ford MH. Dietary calcium intake as a mitigating factor in colon cancer. Am J Epidemiol 1988; 128: Murray TM. Prevention and management of osteoporosis: consensus statements from the Scientific Advisory Board of the Osteoporosis Society of Canada. 4. Calcium nutrition and osteoporosis. Can Med Assoc J 1996; 155: Willett W. Nutritional epidemiology. Oxford, England: Oxford University Press, 1990: Block G, Sinha R, Gridley G. Collection of dietary supplement data and implications for analysis. Am J Clin Nutr 1994;59 (Suppl.):232S-9S. 10. Bender MM, Levy AS, Schucker RE, et al. Trends in prevalence and magnitude of vitamin and mineral supplement usage and correlation with health status. J Am Diet Assoc 1992;92: Patterson RE, White E, Kristal AR, et al. Vitamin supplements and cancer risk: a review of the epidemiologic evidence. Cancer Causes Control 1997,8: White E, Shannon J, Patterson RE. Vitamin supplement use in relation to colon cancer. Cancer Epidemiol Biomarkers Prev 1997;6: Losonczy KG, Harris TB, Havlik RJ. Vitamin E and vitamin C supplement use and risk of all-cause and coronary heart disease mortality in older persons: the Established Populations for Epidemiologic Studies of the Elderly. Am J Clin Nutr 1996;64: Dawson-Hughes B, Harris SS, Krall EA, et al. Effect of calcium and vitamin D supplementation on bone density in men and women 65 years of age or older. N Engl J Med 1997;337:670-6.

7 Vitamin Supplement Assessment Park YK, Kim I, Yetley EA. Characteristics of vitamin and mineral supplement products in the United States. Am J Clin Nutr 1991;54: Schakel SF, Sievert YA, Buzzard IM. Sources of data for developing and maintaining a nutrient database. J Am Diet Assoc 1988;88: Psaty BM, Lee M, Savage PJ, et al. Assessing the use of medications in the elderly: methods and initial experience in the Cardiovascular Health Study. J Clin Epidemiol 1992;45: Kristal AR, Levy L, Patterson RE, et al. Trends in food label use associated with new nutrition labeling regulations. Am J Public Health 1998;88: Kristal AR, White E, Davis JR, et al. The effects of enhanced calling efforts in random digit dial surveys in response rates and population and population level estimates of health behavior and costs. Public Health Rep 1993; 108: Armstrong BK, White E, Saracci R. Principles of exposure measurement in epidemiology. Monographs in epidemiology and biostatistics. Vol. 21. Oxford, England: Oxford University Press, 1994: Physicians' desk reference for nonprescription drugs. 17th ed. Montvale, NJ: Medical Economics Data Publication Co., Thompson FE, Byers T. Dietary assessment resource manual. J Nutr 1994;124 (Suppl.):2279S. 23. Block G, Woods M, Potosky A, et al. Validation of a selfadministered diet history questionnaire using multiple diet records. J Clin Epidemiol 1990;43: Zar JH. Biostatistical analysis. Second ed. Englewood Cliffs, NJ: Prentice-Hall, Inc., Willett WC, Sampson L, Stampfer MJ, et al. Reproducibility and validity of a semiquantitative food frequency questionnaire. Am J Epidemiol 1985;122: Rimm EB, Giovannucci EL, Stampfer MJ, et al. Reproducibility and validity of an expanded self-administered semiquantitative food frequency questionnaire among male health professionals. Am J Epidemiol 1992;135: Munger RG, Folsom AR, Kushi LH, et al. Dietary assessment of older Iowa women with a food frequency questionnaire: nutrient intake, reproducibility, and comparison with 24-hour dietary recall interviews. Am J Epidemiol 1992,136: