Yu Chen
Department of Epidemiology, Mailman School of Public Health, Columbia University, New York, USA
Habibul Ahsan*
Herbert Irving Comprehensive Cancer Center, Columbia University, New York, USA
Faruque Parvez
Department of Environmental Health Sciences, Mailman School of Public Health, Columbia University, New York, USA
Geoffrey R. Howe
Department of Epidemiology, Mailman School of Public Health, Columbia University, New York, USA
Validation study: Food-frequency questionnaires: Food diaries: Nutritional epidemiology
Socio-economic and Policy
Validation
Parent study: The parent HEALS is a large ongoing prospective cohort study to investigate the health effects of As exposure from drinking water in Bangladesh. Briefly, water samples and household and geographic data were collected and analysed for 5966 contiguous wells serving nearly 66 000 individuals in a well-defined 25 km2 geographic area in Araihazar, Bangladesh to create a sampling frame for subject recruitment in the cohort study (van Geen et al. 2002). From 22 October 2000 to 19 May 2002, 11 746 men and women were recruited at baseline from the users of 5966 contiguous wells. The extensive baseline interview included inquiries on water-drinking patterns and history, demographics and lifestyle characteristics. Information on diet was assessed using a newly developed FFQ. In addition, trained physicians completed comprehensive physical examinations with special emphasis on As-related signs and symptoms such as As-induced skin lesions and skin cancers. Biological specimens, including whole blood and spot urine samples, were collected from 97 % of study participants. Study design for the validation study Stratified sampling was conducted for the FFQ validation study. Parallel with the baseline interview of the cohort study, 300 wells were randomly selected from the 4236 wells that enumerated the overall cohort study participants. From cohort study participants who were users of these 300 wells, 200 subjects were randomly chosen and invited to the FFQ validation study. The food-frequency questionnaire The semi-quantitative thirty-nine-item FFQ was designed to assess the long-term daily diet of subjects in the cohort study. Since the average education level of residents in the present study area was low (44 % without any formal education), trained interviewers completed the FFQ through in-person interviews. HEALS investigators, with help from local nutrition experts, first identified all the food items available at the village market in the study area. Then, a preliminary version of the FFQ was developed following extensive discussions with ten focus groups, each consisting of six to ten participants. The FFQ was finalised after pilot testing among 120 local individuals who were not part of the cohort study. Only common food items were included in the FFQ (Appendix 1), and food items with intake frequencies less than once per month during the past 1 year were deemed to be insignificant. About 10 % of the food items were removed from the original food list because of infrequent consumption. Although a close-ended format may theoretically help subjects describe their dietary patterns, its fixed categorisations are not appropriate for foods that have large differences in seasonal availability and its numerous categories may be cumbersome in an interview setting. Therefore, to simplify the FFQ, openended questions for the amount per meal, frequency per d, month and year were used. For instance, for mango, subjects were asked about the number of months per year they consumed mango, the number of days per week they ate it during those months, the number of times in a typical day they ate it and, last, the amount they consumed on average each time. Different locally used plates and utensils were shown to the subjects to define the portion sizes during the interviews. Seven-day food diaries The two 7 d FD were completed in two major separate seasons in order to capture the seasonal variation of food availability. The first FD took place between April 2001 and July 2001 and the second between January 2002 and March 2002. Because of the large amount of work involved, twenty village interviewers were recruited. They received a 3 d training that included discussion sessions on the methods of the FD, detailed explanations of the study purpose, demonstrations of interviews and interview practice through role-playing. During data collection, each of the twenty trained interviewers visited six to ten subjects at home three times per d either during or immediately after the meal (no later than 2 h) to directly measure and record consumptions of the thirty-nine food items listed in the FFQ. Two study coordinators and two nutrition experts supervised the interviewers, checked the completeness of the FD and held discussion sessions on a daily basis for quality control. Among the 200 subjects invited, 198 completed the first 7 d FD, of which 193 completed the second. Among these 193 subjects with both 7 d FD completed, four were excluded because of incomplete FFQ. Therefore, the final sample for the validation study included 189 subjects. Baseline interviews and FFQ were conducted for these subjects between October 2000 and June 2001. FD were conducted after the FFQ for all study subjects. On average, the duration from the FFQ to the first 7 d FD was 76 d, and the duration from the first to the second FD was 190 d. Statistical analysis Average daily intakes of food items and nutrients were calculated separately based on the FFQ and FD. Since nutrient values specifically for foods in Bangladesh were not available, we used both the United States Department of Agriculture (USDA) Nutrient Database for Standard Reference (abbreviated version) (United States Department of Agriculture, 2002) and an Indian food nutrient database (Gopalan et al. 1989). Given the geographical and socio-cultural similarities between India and Bangladesh, we were able to identify exact food items listed in the FFQ from the Indian food composition table. While food items listed in the FFQ were closely matched to those in the Indian food composition tables, the USDA data may have more accurate assessments of nutrients. Several studies have documented As contents of foodstuffs collected in As-affected areas (Roychowdhury et al. 2002; Misbahuddin, 2003). The content of As in foodstuffs depends on the As concentration in the soil, in the water used for washing and cooking, in the water used for irrigation purpose and in the pesticides (Roychowdhury et al. 2002). Therefore, the evaluation of the As content in the diet is complex and will be the subject of a separate future analysis.
British Journal of Nutrition (2004), 92, 851–859
Journal