All three FFQs performed reasonably well, although not perfectly, to estimate dietary calcium intake obtained from 7-day food records. Mean calcium intakes from food records and all three FFQs were similar to mean calcium intake from food (892 mg/day) reported for a national sample (NHANES 2001–2002), based on two 24-hour recalls (38
). Although the DHQ showed some systematic bias resulting in underestimation of calcium intake, it was relatively small. The DHQ also had a greater magnitude bias than the Calcium Questionnaire or the Short Calcium Questionnaire, suggesting the DHQ may further underestimate calcium intake for those individuals with the highest calcium intakes. One explanation for the magnitude bias exhibited by the DHQ may be that some respondents may be consuming larger food portions of high-calcium foods than those expected by the DHQ database.
Differences among the nutrient databases used for the three FFQs and for the food record analyses may have accounted, in part, for the apparent underestimation of calcium intake by the DHQ. For example, based on package label information, Total brand cereals (General Mills, Inc, Minneapolis, MN) contain 1,000 mg of calcium in a ¾ to 1 cup (30 to 55 g) serving; yet in the DHQ database, a serving of <1 cup of Total and other fortified cereals is assigned a calcium value of only 104 mg for women and 140 mg for men (http://riskfactor.cancer.gov/DHQ/database/current.html
). The Calcium Questionnaire and Short Calcium Questionnaire databases, as well as the Nutrition Data System for Research database (used to analyze food records), used the food manufacturer’s figures. The absence of some calcium-fortified food items on the DHQ might also have led to underestimations of calcium intake. For example, calcium-fortified orange juice, which contains >300 mg calcium per cup is not differentiated from nonfortified orange juice on the DHQ. Similarly, no option for calcium-fortified soy milk is available in the DHQ database.
It is understandable then that the DHQ, as a total diet FFQ, may not be as sensitive for measuring calcium intake as a trade-off for measuring a larger number of nutrients in the diet. When intake data for several nutrients is needed in the research setting, use of a total diet FFQ is a logical choice, and estimations of calcium intake from the DHQ may be considered acceptable in this context. The DHQ was developed more than 8 years ago; periodic updating of the DHQ database will be necessary to maintain the integrity of this instrument.
The Calcium Questionnaire demonstrated the largest systematic bias, yet it showed no significant magnitude bias. As FFQs are frequently used to rank order subjects by nutrient intake, the Calcium Questionnaire would appear to perform well in this capacity. The tendency of the Calcium Questionnaire to overestimate calcium intake may be due, in part, to the design of this FFQ; the large number of calcium-containing foods listed and the open-ended format for frequency responses may lead respondents to overestimate their consumption. The absence of magnitude bias for the Calcium Questionnaire suggests that the portion-size categories listed are appropriate. In addition, the ability to enter specific calcium content information for nutritional supplements and meal replacements lessens this potential source of magnitude bias. Because the Calcium Questionnaire is not meant to be a total diet FFQ, it is not surprising that estimated mean daily energy intake was only 77% of that estimated by food records. Researchers and practitioners can use the energy intake estimates generated by the Calcium Questionnaire analysis to gauge validity using cutpoints similar to those used in this study; this is not possible with the Short Calcium Questionnaire.
The Short Calcium Questionnaire showed an insignificant systematic bias, suggesting that use of an adjustment factor of +200 mg/day to reflect the contribution of calcium from foods eaten but not listed on the questionnaire was appropriate. The magnitude bias evidenced by the Short Calcium Questionnaire may be a result of the fact that portion-size selection is not an option on this FFQ; the default portions assigned to food items may underestimate portions consumed for those respondents consuming larger portions. The Short Calcium Questionnaire would appear to be suitable for use as a screening tool to identify those at risk for low calcium intakes in order to make treatment recommendations.
When selecting an appropriate tool for dietary assessment, practical issues related to the administration of FFQs should also be examined. When respondent burden is a concern, the length of time needed for an individual to complete each FFQ should be considered: about 1 hour for the DHQ (10
), 15 to 30 minutes for the Calcium Questionnaire, and <10 minutes for the Short Calcium Questionnaire. Analysis times and required resources also vary among the three FFQs. The DHQ requires scanning and/or on-line analysis, the Calcium Questionnaire can be coded in 10 to 15 minutes, and the Short Calcium Questionnaire can be analyzed in <5 minutes. When assessment of only calcium intake is indicated, the shorter Calcium Questionnaire or Short Calcium Questionnaire would thus appear to be better choices.
Possible limitations of using food records as the criterion method should be noted. The act of prospectively recording food intake may influence subjects’ normal consumption resulting in distortion of usual intake, as well as lead to reporting bias; these effects may be more pronounced with obese subjects (39
). Of note, our study enrolled proportionally more overweight subjects (BMI >24.9; 86.5%) compared to a national average of 66.3% overweight adults (40
). Yet, the food records in this study appeared to be of high quality, as judged by the research RDs, and as evidenced by energy and calcium intake estimates that were similar to national data (38
Strengths of this validation study include a large, relatively diverse (although geographically limited) sample of adults, with data collection spanning all seasons. Our findings might not be applicable to elderly adults, to those who are not proficient in the English language, or to those with low literacy levels. Weaknesses of this study are primarily related to design and logistical issues. Although it would have been preferable for subjects to have kept food records periodically throughout the year preceding administration of the FFQs (rather than postadministration of the FFQs, as was done in this study), this was not possible because of the design of the primary study. Similarly, logistics of the primary study dictated that the Short Calcium Questionnaire and the Calcium Questionnaire be completed in the same order on the same day; completing these questionnaires 1 or 2 weeks apart in random order would have been preferable. Verbal directions were provided for the Short Calcium Questionnaire and Calcium Questionnaire, which were completed in the clinic; verbal instructions were not provided for the DHQ, which was filled out at home, where other household members may have aided in completing this FFQ. It should also be noted that differences in the time periods reflected by the three FFQs varied somewhat. Directions printed on the DHQ and Calcium Questionnaire were to report intake “during the past year”; instructions printed on the Short Calcium Questionnaire were to indicate servings consumed “during a typical week.” Because the design excluded subjects with calcium intakes that exceeded 3,500 mg/day, the results of the study may not apply to individuals with very high calcium intakes.
Lastly, subjects’ knowledge that they were participating in a study addressing calcium intake and were weighed during their clinic visits may have affected their reports of intake and possibly improved their recording of calcium-containing foods.