American Journal of Epidemiology Copyright O 1992 by The Johns Hopkns University School of Hygiene and Pubfc Health Al rights reserved

Vol. 136, No. 2 Printed in U.S.A

Dietary Assessment of Older Iowa Women with a Food Frequency Questionnaire: Nutrient Intake, Reproducibility, and Comparison with 24-Hour Dietary Recall Interviews Ronald G. Munger,1 Aaron R. Folsom,2 Lawrence H. Kushi,3 Susan A. Kaye,2 and Thomas A. Sellers2

The authors report the results of a dietary survey of 38,121 Iowa women, 55-69 years of age in 1986, based on a semiquantitative food frequency questionnaire previously tested among Boston-area women aged 34-59 years. The Iowa women, compared with the younger Boston-area women, consumed a similar amount of calories (1,767 vs. 1,844 kcal) and a similar amount of total calories from fat (35 vs. 37%) but had markedly greater intake of the following mtaronutrients after including supplement use: iron (+18%), calcium (+33%), vitamin A (+43%), riboflavin (+46%), thiamine (+50%), and pyridoxine (+122%). The reproducibility of the questionnaire was examined in two more administrations to 44 of the Iowa women in January and June of 1988. Reproducibirrty was highest for alcohol (Pearson's r = 0.99), caffeine (r = 0.95), and vitamin E (r = 0.90) and lowest for sucrose (r = 0.53), polyunsaturated fat {r = 0.56), and iron (r = 0.59). Micronutrient intakes were generally more reproducible than macronutrient intakes. The agreement between the June 1988 questionnaire and the average of five 24-hour dietary recalls was also assessed in the 44 subjects. The median correlations of energy-adjusted intake were as follows: for macronutrients, r = 0.45; for micronutrients without supplements, r = 0.33; and for micronutrients with supplements, r = 0.64. This food frequency questionnaire appears to be reasonably reproducible and accurate, so that Its use may be extended to epkJemiologic studies of older women with a broad range of socioeconomic backgrounds. Am J Epidemiol 1992:136:192-200. diet; food; nutrition; nutrition surveys; questionnaires

The semiquantitative food frequency questionnaire has emerged as a useful tool to assess diet in large-scale epidemiologic

studies (1). Willett et al. (2) demonstrated reasonable levels of reproducibility and validity for their food frequency questionnaire in a sample of nurses aged 34-59 years from the Boston, Massachusetts, area. The nurses in the Boston study were educated, urban dwelling, and employed. The usefulness of this questionnaire in other populations has not been well documented but is of interest, given the importance of prospective health studies of the elderly, minorities, the poor, and less educated individuals in both rural and urban areas. The Willett questionnaire was adapted for use in a prospective study of the health of older Iowa women. We report here the nutrient intake of this large cohort, the reproducibility of the questionnaire, and the comparison of the results of the questionnaire with a separate dietary

Received for publication April 9, 1991, and in final form January 8, 1992. ' Department of Preventive Medicine and Environmental Health, University of Iowa Coflege of Medicine, Iowa City, IA. 'Division of Epidemiology, University of Minnesota School of Public Health, Minneapolis, MN. 3 Division of Human Development and Nutrition, University of Minnesota School of Public Health, Minneapolis, MN. Reprint requests to Dr. Ronald G. Munger, Department of Preventive Medicine and Environmental Health, 2830 Stemdler Building, University of Iowa College of Medicine, Iowa City, IA 52242. Supported by grant R01-CA 39742 from the National Cancer Institute. The authors gratefully acknowledge the assistance of Ray Hsieh, Cyndy Long, Dr. Laura Sampson, Dr. Mary Fran Sowers, Dr. Robert Wallace, Joyce Wenz, and Dr Walter Wlett.

192

Dietary Assessment of Older Iowa Women

assessment using multiple 24-hour dietary recalls. MATERIALS AND METHODS

The Iowa Women's Health Study is a longitudinal study of cancer incidence in Iowa women aged 55-69 years at baseline. Questionnaires were mailed during January 1986 to 99,826 Iowa women whose names were selected from the driver's license data files of the Iowa Department of Transportation. A total of 41,837 women (42 percent) returned the questionnaire. Of the women who returned the 1986 baseline questionnaire, 99 percent were white, and their mean age was 61.7 ± 4.2 years (standard deviation). This sample included women from rural and urban areas with both high and low levels of education. Thirty-four percent resided in cities with a population of > 10,000, and 27 percent lived in rural areas. Thirty-nine percent had greater than a high school education, and 19 percent had less than a high school education. Other details of the Iowa Women's Health Study have been published elsewhere (3). A semiquantitative food frequency questionnaire, adapted from the 126-item questionnaire used in the Nurses' Health Study (2), was included in the 1986 baseline questionnaire of the Iowa Women's Health Study. Three minor modifications of the Nurses' Health Study questionnaire were made: 1) green peppers were added to the list of foods; 2) vitamin D doses were added to the supplement section; and 3) iron was added to the listed supplements. All other questions remained the same. The questionnaire was designed for optical scanning, and nutrients were scored using the computer program and nutrient data base developed for the Nurses' Health Study. Two exclusions were applied. Participants with >29 missing items were excluded because our preliminary analyses revealed that estimates of total energy intake dropped rapidly when more than this number of food items were left blank. A total energy intake of 3,5OO kcal was judged to be implausible; participants with these val-

193

ues were also excluded. After exclusions, 38,121 subjects (91 percent) remained. A follow-up survey was conducted in the fall of 1987, and participants were asked if they would take part in additional diet assessments. A total of 244 participants were randomly selected, of whom 71 had declined participation in further dietary studies. Recruitment letters were sent to the remaining 173 participants, and consent was received from 77 of them within 5 weeks. Of these 77 women, 56 were recruited, because the 21 others indicated that they would be traveling away from home during the period of study. Two additional food frequency questionnaires, identical to the one used at baseline, were completed by participants: one from January to March of 1988 and one during June 1988. Each questionnaire was used to assess nutrient intake during the previous 1-year period. Five 24-hour dietary recall interviews were completed with each participant in the substudy during February and March of 1988. Five to six recalls per subject have been shown to reduce the attenuation of correlation coefficients due to unreliability to 10 percent (4). The dietary recall interviews were conducted by telephone by one trained dietitian with extensive experience in nutrition assessment and coding. Each interview lasted approximately 20 minutes. Dietary recalls included 1-2 weekend days and 3-4 weekdays selected at random. A two-dimensional visual guide to portion sizes (Nutrition Consulting Enterprises, Newton, Massachusetts) was mailed to each participant before the first interview, and participants were instructed to keep the guide near the telephone for use in the interviews but were not told when they would be called. Questions were also asked during each recall about participants' intake of nutritional supplements over the same 24-hour period. The sequence of data collection is depicted in table 1. Reported intake from the 24-hour dietary recalls was coded by food type, portion size, and daily frequency. Nutrients, including supplements, were scored with the same nutrient data base used to score the food fre-

194 Mungeretal. TABLE 1. Sequence of dietary assessments of the Iowa Women's Health Study, 1986-1988 Dietary assessment

Baseline food frequency questionnaire (FFQ 1) Second food frequency questionnaire (FFQ 2) Five 24-hour recalls Third food frequency questionnaire (FFQ 3)

Dates administered

January to April, 1986

(p < 0.05). The purpose of our analysis was to describe classification of nutrient intake and its relevance to epidemiologic studies rather than hypothesis testing; therefore, p values are not presented.

January to March, 1988 February and March, 1988 June 1988

quency questionnaires. The mean nutrient intake of the 5 recalled days, weighted to a 7-day week, was used in the data analyses. Forty-four of the 56 participants had complete data and plausible calorie intakes and were included in the analyses. The following statistical procedures were used. The mean ± standard deviation was computed for total intake of each nutrient (food and supplement sources combined) and for nutrient intake from food alone (supplements excluded). The supplemental intake of micronutrients is highly skewed, so median values for food and supplemental sources are presented in table 3 to provide a more accurate summary of their relative importance. In addition, nutrient variables used in correlation and regression analyses were loge transformed to improve normality and adjusted for total energy intake using the regression technique described by Willett et al. (1). Separate regression models were constructed for each nutrient with total energy intake as the independent variable and nutrient intake as the dependent variable; energy-adjusted nutrient scores for each participant were calculated by adding the residual to the expected nutrient score for the mean energy intake of all subjects in the substudy. Separate regression models were used for each of the three administrations of the food frequency questionnaire and for the mean of the 24-hour dietary recalls. Pearson's correlation coefficients were calculated for the comparisons of crude and energyadjusted nutrient intake. With a sample size of 44, correlation coefficients of >0.25 are statistically significant

RESULTS

Values for the nutrient intake (mean ± standard deviation) at baseline in 1986 for the 38,121 cohort members with adequately completed questionnaires are presented in table 2. The Iowa women on average consumed 1,767 calories per day, 293 mg of cholesterol per day, and 80 g of protein per day; obtained 35 percent of their calories from fat; and had a polyunsaturated/saturated fat intake ratio of 0.51. The means of nutrient intake calculated from the 1986 baseline questionnaire of the subsample of 44 women in the reliability study were comparable to those for the entire cohort at baseline. The means of nutrient intake of the subsample were also stable with repeated administrations of the food frequency questionnaire (data not presented). The mean intake estimated by the 24-hour dietary recalls was greater than that estimated by the first food frequency questionnaire for total energy (+13 percent), total fat (+19 percent), sodium (+90 percent), alcohol (+223 percent), and sucrose (+238 percent) and lower for vitamin B l2 (-33 percent) and vitamin A (-26 percent). Medians of micronutrient intake of the entire cohort, with and without supplements, and the percentage of participants taking supplements of each nutrient are presented in table 3. Sixty-three percent of women reported using one or more nutritional supplements. The amount of supplements used relative to dietary intake varied considerably. The median supplemental retinol was equivalent to 80 percent of the median retinol consumed in food alone; the other leading supplements included vitamin D (equivalent to 51 percent of intake from food sources) and calcium (41 percent). Relative to dietary sources, the supplements used less included folate (18 percent of dietary intake) and iron (11 percent). The

Dietary Assessment of Older Iowa Women

195

TABLE 2. Nutrient Intake (mean ± standard deviation) from food of 38,121 Iowa women aged 55-69 years in 1986 and of 44 women in a subsamp4e for the study of reprodudbility and comparison with 24-hour dietary recall Interviews Nutrient

Cohort at baseline Of FFQ1*, t (n = 38,121)

Subsampteof FFCMt (n-44)

Subsampte, mean of five 24-hour dietary recalls (n-44)

Total energy (kcal)* Protein (g) Total fat (g) Saturated fat (g) Monounsaturated fat (g) Polyunsaturated fat (g) Cholesterol (mg) Total carbohydrate (g) Sucrose (g) Crude fiber (g) Phosphorus (mg) Sodium (mg) Potassium (mg) Alcohol (g) Caffeine (mg) Vitamin A (IU)* Retinol (IU) Vitamin D (IU) Vitamin E (mg) Thlamlne (mg) Riboflavin (mg) Pyridoxlne (mg) Fcrfate (Mg) Vitamin B12 (M9.) Vitamin C (mg) Calcium (mg) Iron(mg)

1,767 ± 545 80 ±28 67 ±25 24 ±10 25 ± 10 12±5 293±135 213 ±73 45 ±25 5.6 ± 2.5 1,280 ± 447 2,005 ± 728 3,240 ±1,029 3.9 ± 9.2 289 ± 249 12,209 ±8,657 2,910 ±3,315 252 ±151 7.7 ± 4.3 1.3 ± 0.5 2.0 ± 0.8 2.1 ± 0.8 282 ±121 10.8 ±10.3 150 ±78 824 ± 390 13.1 ±5.6

1,793 ± 443 84 ±25 69 ±21 25 ± 8 26 ±9 11 ±4 312 ± 115 214 ±60 48 ±20 5.7 ± 2.0 1,318 ±375 1,962 ±638 3,364 ± 1,075 3.0 ± 9.5 303 ±291 12,669 ±6,294 2,381 ±1,667 233 ±125 6.8 ± 2.3 1.3 ±0.4 2.0 ± 0.7 2.1 ± 0.9 281 ±120 9.1 ± 5.0 148 ±82 848 ±358 13.0 ±3.9

2,020 ± 530 79 ±23 82 ±30 28 ± 9 28 ±9 16 ± 7 306±129 259 ± 79 114±190 5.9 ± 2.7 1,339 ±414 3,732 ± 2,250 3,285 ± 961 6.6 ±21.0 309 ±310 9,379 ±6,017 1,930 ± 1,293 260 ±129 9.6 ± 5.1 1.4 ± 0.5 2.4 ± 1.9 5.1 ± 12.4 289 ±188 6.1 ± 3.7 130 ±61 912 ±346 25.2 ± 60.6

• FFQ 1, food frequency questionnaire 1; kcal, kBocalorie(s); IU, IntemationaJ unlt(s). t Nutrient Intake assessed with semiquantttative food frequency questionnaire at baseline in 1986.

TABLE 3. Median micronutrient intake of 38,121 Iowa women aged 55-69 years in 1986 from dietary and supplemental sources and supplement intake as percentage of dietary intake* Nutrient

Vitamin A (IU)t Retinol (IU) Vitamin D (IU) Vitamin E (mg) Thiamine (mg) Riboflavin (mg) Pyridoxine (mg) Folate (Mg) Vitamin B,2 (Mg) Vitamin C (mg) Calcium (mg) Iron (mg)

10,376 2,139 222 7.0 1.2 1.9 2.0 265 8.4 138 736 12.2

Median total Intake, Inducing supplements

Supplement intake as % of Intake from food

%of women taking supplement

12,436 3,847 335 9.2 1.5 2.4 2.5 312 11.5 188 1,039 13.6

20 80 51 32 25 26 25 18 37 36

34 34 36 37 36 36 36 28 35 45 48 20

• Sixty-three percent of participants reported using one or more supplements, t IU, International unit.

41 11

196

Mungeretal.

percentage of cohort members taking each supplement is also listed in table 3. The supplements most often consumed included calcium (48 percent of participants), vitamin C (45 percent), and retinol (34 percent); supplements used less often included folate (28 percent) and iron (20 percent). Reproducibility of nutrient intake

The correlations of nutrient intakes estimated from the food frequency questionnaires between the intervals of January 1986 to January 1988, January 1986 to June 1988, and January 1988 to June 1988 are summarized in table 4. Pearson's correlation coefficients for nutrient intakes were greatest for the comparison of the last two food frequency questionnaires with a 5-month interval (median r = 0.76; range = 0.530.99) compared with the first and second administrations with a 2-year interval (median r = 0.61; range = 0.33-0.99) and the first and third administrations with a 2.5year interval (median r = 0.62; range = 0.30-0.98). Energy adjustment did not appreciably alter these results (data not presented). Micronutrient intakes including supplements were slightly more reproducible than were the macronutrient intakes. In the comparison of the last two food frequency questionnaires given 5 months apart, the highest correlations were for alcohol (r = 0.99), caffeine (r = 0.95), and vitamin E (r = 0.90), and the lowest correlations were for sucrose (r = 0.53), polyunsaturated fat (r = 0.56), and iron (r = 0.59). Comparison of dietary assessment methods

The Pearson correlation coefficients for the comparison of the June 1988 food frequency questionnaire with the mean of the five 24-hour dietary recalls appear in table 5. Adjustment for total energy intake generally improved the correlations for macronutrients (median unadjusted r = 0.29, range = —0.04 to 0.82; median adjusted r = 0.45, range = -0.07 to 0.82) but did not improve the correlations for micronutrients

TABLE 4. Reproducibility of food frequency questionnaire among 44 Iowa women aged 55-69 years in 1986: Pearson's correlation coefficients of nutrient intake, unadjusted for total energy intake Questionnaires compared* Nutrient

Total energy (kcal)t Protein (g) Total fat (g) Saturated fat (g) Monounsaturated fat (g) Polyunsaturated fat (g) Cholesterol (mg) Total carbohydrate (g) Sucrose (g) Crude fiber (g) Phosphorus (mg) Potassium (mg) Alcohol (g) Caffeine (mg) Vitamin A (IU)t Total Without supplements Retinol (III) Total Without supplements vitamin D (III) Total Without supplements Vitamin E (mg) Total Without supplements Thiamine (mg) Total Without supplements Riboflavin (mg) Total Without supplements Pyrtdoxine (mg) Total Without supplements Folate (^g) Total Without supplements Vitamin B12 (nQ) Total Without supplements Vitamin C (mg) Total Without supplements Calcium (mg) Total Without supplements Iron (mg) Total Without supplements

FFQ1t

FFQ1

vs.

vs.

FFQ2 V3.

FFQ2f

FFQ3f

FFQ3

0.51 0.59 0.51 0.45 0.57 0.50 0.53 0.47 0.49 0.57 0.54 0.61 0.99 0.88

0.61 0.70 0.60 0.62 0.64 0.44 0.59 0.54 0.40 0.60 0.70 0.69 0.98 0.85

0.67 0.74 0.71 0.70 0.73 0.56 0.64 0.66 0.53 0.82 0.76 0.80 0.99 0.95

0.64 0.58

0.67 0.63

0.75 0.69

0.66 0.70

0.62 0.51

0.75 0.67

0.63 0.60

0.63 0.64

0.73 0.82

0.82 0.58

0.70 0.30

0.90 0.84

0.75 0.46

0.68 0.55

0.84 0.78

0.73 0.71

0.66 0.78

0.83 0.84

0.69 0.42

0.58 0.48

0.86 0.87

0.33 0.62

0.41 0.59

0.76 0.87

0.53 0.67

0.37 0.55

0.73 0.69

0.81 0.69

0.84 0.57

0.83 0.81

0.63 0.63

0.57 0.66

0.82 0.76

0.85 0.47

0.40 0.58

0.59 0.81

• Semtquantitative food frequency questionnaires; dates listed In table 1. tFFQ 1, food frequency questionnaire 1; FFQ 2, food frequency questionnaire 2; FFQ 3, food frequency questionnaire 3; kcal, kbocalorie(s); IU, International unit(s).

Dietary Assessment of Older Iowa Women

TABLE 5. Comparison of nutrient intake from third food frequency questionnaire with five 24-hour dietary recalls, both administered In 1988 to 44 Iowa women, aged 55-69 years in 1986: Pearson's correlation coefficients for unadjusted and energyadjusted nutrient intake Food frequency questionnaire 3 vs. 24hour dietary recalls Nutrient Unadjusted

Total energy (kcal)* Protein (g) Total fat (g) Saturated fat (g) Monounsaturated fat (g) Polyunsaturated fat (g) Cholesterol (g) Total carbohydrate (g) Sucrose (g) Crude fiber (g) Phosphorus (mg) Potassium (mg) Alcohol (g) Caffeine (mg) Vitamin A (IU)* Total Without supplements Retinol (IU) Total Without supplements Vitamin D (IU) Total Without supplements Vitamin E (mg) Total Without supplements Thlamine (mg) Total Without supplements Riboflavin (mg) Total Without supplements Pyridoxine (mg) Total Without supplements Folate (Mg) Total Without supplements Vitamin B12 (^g) Total Without supplements Vitamin C (mg) Total Without supplements Calcium (mg) Total Without supplements Iron (mg) Total Without supplements

Adjusted for total energy Intake

0.30 0.27 0.27 0.27 0.28 0.34 0.21 0.39 -0.04 0.35 0.46 0.52 0.32 0.82

0.16 0.62 0.59 0.62 0.43 0.21 0.45 -0.07 0.24 0.52 0.65 0.32 0.82

0.60 0.37

0.56 0.14

0.70 0.50

0.68 0.38

0.51 0.59

0.51 0.61

0.62 0.66

0.55 0.79

0.95 0.60

0.95 0.68

0.92 0.36

0.93 0.32

0.70 -0.09

0.69 -0.16

0.43 0.35

0.43 0.26

0.58 0.38

0.76 0.47

0.82 0.55

0.76 0.53

0.67 0.54

0.64 0.49

0.00 0.07

-0.01 -0.09

* kcal, kllocakxie(s); IU, International unrt(s).

197

without supplements (median unadjusted r = 0.38, range = -0.09 to 0.66; median adjusted r = 0.38, range = -0.16 to 0.79) or for micronutrients with supplements (median unadjusted r = 0.62, range = 0.00 to 0.95; median adjusted r = 0.64, range = -0.01 to 0.95). Micronutrient intakes that included supplements had generally greater correlations between methods than did micronutrients from food sources alone and macronutrient intakes. The highest correlations were found for total thiamine (unadjusted and adjusted r = 0.95), total riboflavin (unadjusted r = 0.92; adjusted r = 0.93), caffeine (unadjusted and adjusted r = 0.82), and total vitamin C (unadjusted r = 0.82; adjusted r = 0.76). The lowest correlations were for total iron (adjusted r = -0.01), sucrose (adjusted r = -0.07), and pyridoxine from food sources alone (adjusted r = -0.16).

DISCUSSION

The semiquantitative food frequency questionnaire developed by Willett et al. was shown to be reproducible and valid for use in a group of Boston-area nurses and has been used in a large-scale epidemiologic study of women nurses (1, 2). We administered a nearly identical questionnaire to a group of older Iowa women and, despite notable differences in patterns of nutrient intake between the Iowa and Boston women, found similar levels of reproducibility and agreement relative to the mean of five 24hour dietary recalls. Although the primary purpose of a food frequency questionnaire is to rank individuals according to nutrient intake, use of the same method in two or more populations may allow comparisons of dietary patterns. The older Iowa women, compared with the Boston-area women, had slightly lower total energy intake (4 percent less), a similar percentage of total calories from fat (35 vs. 37 percent), and a similar polyunsaturated/ saturated fat intake ratio (0.51 vs. 0.52). The Iowa women consumed more protein (46 vs. 41 g/1,000 kcal), 18 percent more crude fiber, and 37 percent more vitamin A from

198

Munger et al.

food sources, and they had a markedly greater intake of the following micronutrients after including supplement use: iron (+18 percent), calcium (+33 percent), vitamin A (+43 percent), riboflavin (+46 percent), thiamine (+50 percent), and pyridoxine (+122 percent). Use of vitamin C supplements was similar and considerable in both groups of women. The difference between the Iowa and Bostonarea women in food and supplement use is probably due to differences in age, racial composition, geographic area, occupations, and education. The use of nutritional supplements is greater among women compared with men, varies by region of the United States, and increases with greater income and level of education (5-8). The use of supplements by US women in 1986 increased with age until 50 years, decreased slightly between 50 and 74 years, and then dropped more sharply beyond 75 years (9). Supplement use has increased markedly in the United States since the early 1970s. The National Health Interview Survey during 1971-1974 found that 26 percent of women reported any supplement use (7). Among women aged 51 -74 years in the second National Health Interview Survey (1976— 1980), 44 percent reported any supplement use (6). Sixty-three percent of the Iowa women consumed one or more supplements, and median supplement intake exceeded the median food intake of retinol, vitamin D, calcium, vitamin B,2, and vitamin C by 33 percent or more. The Boston-area women were surveyed 2 years before the Iowa women, probably not enough time to result in important secular changes. Despite the differences in age and demographic characteristics between these two groups, the food frequency questionnaire performed similarly with regard to reproducibility and in comparison with other methods of dietary assessment. The methods used in the Iowa and Boston-area studies were, however, not exactly the same. The reproducibility studies completed by Willett et al. (1) were based on an earlier version of the 61 -item food frequency questionnaire administered 1 year apart and re-

sulted in Pearson's correlations with a median of 0.60 and a range of 0.52-0.71. The Iowa study used a 127-item food frequency questionnaire administered over intervals of 2.5 years, 2.0 years, and 4-5 months. As expected, the correlations for repeated administrations in the Iowa study were greatest over the shortest interval of 5 months, and the correlations were, in general, slightly greater than those reported by Willett et al. (1). We report here the results for a greater number and proportion of micron utrients than in the paper by Willett et al. from 1985 (1), and these micronutrients are generally more reproducible than macronutrients. The larger correlations between the second and third administrations of the food frequency questionnaire may reflect the overlap in time periods in which diet was assessed and a possible "training effect" of repeatedly filling out the questionnaires. One measure of the ability of participants to fill out the questionnaire is the number of missing items. The mean number of missing items did not decrease with subsequent administrations of the questionnaire, but this observation cannot rule out entirely the possibility that subjects selected food frequencies more regularly with successive administrations. Estimates of alcohol and caffeine intake were the most reproducible of all of the dietary variables examined. Nutrient intake derived from only a few foods may be less variable as subjects may recall the use of these few items (e.g., cups of coffee or alcoholic drinks) or abstentions with more regularity. The mean caffeine intake was similar between the two methods of dietary assessment, but the mean alcohol intake was much higher according to the average of five 24-hour dietary recalls. The Iowa study compared the results of the 127-item food frequency questionnaire with nutrient intake based on five 24-hour dietary recalls administered by telephone with the subjects referring to a booklet with pictures of portion sizes. The collection of 24-hour dietary recalls by telephone from older women was shown to be comparable to 24-hour dietary recalls collected by per-

Dietary Assessment of Oder Iowa Women 199 sonal interview in the 1971-1974 Health and Nutrition Examination Survey (10). The 24-hour dietary recalls provided higher estimates of intake for many, but not all, nutrients compared with the estimates from the food frequency questionnaire. Mean nutrient intakes from the mean of five 24-hour dietary recalls were markedly greater than those from the food frequency questionnaire for energy, total fat, polyunsaturated fat, iron, alcohol, and sucrose and were lower for vitamin B12 and vitamin A. Differences in reported intake described by these methods may be due to real variations in diet over the varying time periods used, variations in the ability to recall dietary habits over these time periods, the effects of interviewer probing during the 24-hour dietary recall interview, imprecision in translating 24-hour recalls into the food frequency format for scoring, or the alteration of dietary patterns by the methods themselves. Correlations between nutrient intakes derived from the food frequency questionnaire and the dietary recalls were greater for micronutrients with supplements than for micronutrients without supplements. Supplements use was reported by 63 percent of participants and accounted for a large proportion of total micronutrient intake. Addition of supplements extends the range of intake among the participants and may, therefore, enhance the correlation coefficients. Supplement intake may also be more consistently reported compared with the large number of foods that contribute to the intake of a particular nutrient and result in higher correlations when included. In their 1980-1981 validation study, Willett et al. (1) compared estimates of energyadjusted nutrient intakes from their 61-item questionnaire with those based on four 7day records of weighed food and found Pearson's correlations with a median of 0.55 and a range of 0.39-0.67. The comparison of Willett's 126-item questionnaire administered in 1984 with their 1980 food record data produced consistent correlations with a median of 0.54 and a range of 0.28-0.58). These correlations are in general greater than the Iowa correlations for energy-adjusted

macronutrients and energy-adjusted micronutrients without supplements, but they are most often less than the Iowa correlations for energy-adjusted micronutrients with supplements. Generalization of these results should be tempered with some caution. The 44 women who participated in the dietary substudies and who had complete data may not be representative of the entire cohort in every respect. Of the initial random sample of 244 cohort members at baseline, 173 indicated that they would be later willing to participate in dietary studies (70.9 percent); of these, 77 responded to our mailed inquiry within 5 weeks (44.5 percent). The number of available participants was further reduced to 56 after 21 indicated that they were going to be away from home during the period of study. Of the 56 who enrolled in the additional dietary studies, 44 had complete data with the plausible total energy intakes discussed previously. The 44 compliant women who completed the dietary substudy may have been more precise and attentive to detail than others, but they still had mean nutrient intakes that were very similar to those of the entire cohort. We conclude that the food frequency questionnaire developed and tested by Willett et al. among younger Boston-area women should also be useful in studies of diet and disease among older Iowa women with a broader range of socioeconomic backgrounds and different dietary patterns compared with those of the Boston-area women. The results of the Iowa study were reasonably reproducible and accurate for use in classifying individuals by nutrient intake in large epidemiologic studies.

REFERENCES 1. Willett WC, Sampson U Stampfer MJ, et al. Reproducibility and validity of a semiquantitative food frequency questionnaire. Am J Epidemiol 1985;122:51-65. 2. Willett WC, Sampson L, Browne ML, et al. The use of a self-administered questionnaire to assess diet four years in the past Am J Epidemiol 1988; 127:188-99. 3. Folsom AR, Prineas RJ, Kaye SA, et al. Body fat

200

4. 5.

6.

7.

Munger et al.

distribution and self-reported prevalence of hypertension, heart attack, and other heart disease in older women. Int J Epidemiol 1989; 18:361-7. Rush D, Kristal A. Methodologic studies during pregnancy: the reliability of the 24-hour dietary recall. Am J Clin Nutr 1982;35:1259-68. Stewart ML, McDonald JT, Schucker RE, et al. Vitamin/mineral supplement use: a telephone survey of adults in the United States. J Am Diet Assoc 1985;85:1585-9O. Koplan JP, Annest JL, Layde PM, et al. Nutrient intake and supplementation in the United States (NHANES II). Am J Public Health 1986; 76:287-9. Block G, Cox C, Madans J, et al. Vitamin supple-

ment use, by demographic characteristics. Am J Epidemiol 1988; 127:297-309. 8. Kushi LH, Folsom AR, Jacobs DR, et al. Educational attainment and nutrient consumption patterns: the Minnesota Heart Survey. J Am Diet Assoc 1988;88:1230-6. 9. Moss AJ, Levy AS, Kim I, et al. Use of vitamin and mineral supplements in the United States: current users, types of products, and nutrients. In: Advance data. No. 174. Rockville, MD: National Center for Health Statistics, 1989:1 -19. 10. Posner BM, Borman CL, Morgan JL, et al. The validity of a telephone-administered 24-hour dietary recall methodology. Am J Clin Nutr 1982; 36:546-53.

Dietary assessment of older Iowa women with a food frequency questionnaire: nutrient intake, reproducibility, and comparison with 24-hour dietary recall interviews.

The authors report the results of a dietary survey of 38,121 Iowa women, 55-69 years of age in 1986, based on a semiquantitative food frequency questi...
598KB Sizes 0 Downloads 0 Views