In 2005, we recruited a convenience sample of postmenopausal women aged 60 to 80 years who were enrolled in a parent study in the Urbana-Champaign area of Illinois (10) (assessing body composition, bone health, physical activity, physical function, and self-efficacy) to complete additional nutritional assessments. Exclusion criteria for the parent study were neurologic illness, orthopedic limitations, or cognitive limitations that precluded completion of all study testing procedures. All those in the parent study were invited to participate in the current study of evaluation of calcium intake. The human subjects institutional review board of the University of Illinois approved the research protocol, and all participants completed an informed consent form before data collection began.

The CFFFQ and a self-reported 24-hour recall were completed on the same day to assess calcium intake and food source. The CFFFQ is a 46-item food frequency survey focusing on calcium-rich foods. Foods on the questionnaire were chosen based on calcium content and those found in previous studies to be commonly consumed. Foods on the CFFFQ were categorized and presented according to representative food groups: dairy (6 foods), foods with dairy (5 foods), fruits (6 foods), vegetables (3 foods), grains (10 foods), meats (8 foods), and other foods (8 foods) (Appendix). Inclusion of fortified foods was based on availability for purchase in area grocery stores (both local and chain stores). Fortified foods were noted as such on the CFFFQ. A standard portion size was included for each specific food item. Participants entered quantity of food relative to daily, weekly, monthly, or yearly consumption. "Not eaten" was given as an option. We entered foods and servings into a database (Microsoft Access 2003, Redmond, Washington) that calculated calcium content on a daily basis. Double entry was used for quality assurance.

Participants met with researchers to complete the 24-hour recall. The researchers used a multiple-pass interview style to elicit complete information (11). Foods from the 24-hour recall were categorized into the same food groups as on the CFFFQ and were then entered in Nutritionist Pro version 2.3 (First DataBank, San Bruno, California) to obtain calcium values per day.

We analyzed the data by using SPSS version 14.0 (SPSS, Inc, Chicago, Illinois). On the basis of the findings of the Kolmogorov-Smirnov test, the primary variables of interest were not normally distributed. Nonparametric tests were subsequently used to evaluate group or intake assessment differences (Spearman rank correlation and Mann-Whitney). We assessed the adequacy of calcium intake by comparing these 2 assessment methods to two-thirds of the Institute of Medicine's (12) adequate intake amount (used to determine inadequate intake) for women aged 51 years or older. Comparison of the adequacy classification as a measure of specificity of each calcium intake method was completed by using cross-tabulation to demonstrate the relationship between the 2 variables. Chi-square analysis was performed on the cross-tabulated variables. Significance was defined as α = .05.

Calcium from the CFFFQ was significantly correlated with 24-hour recall for all food groups (P < .001) except for vegetables (P = .08) (Table 1). Estimates obtained with the 24-hour recall method were significantly lower than CFFFQ estimates of total daily calcium, dairy, foods with dairy products, fruits, and vegetables (Table 2). The CFFFQ estimate of mean total calcium compared with the 24-hour recall was higher by 221 mg/d (P < .001), including within racial groups (195 mg/d for black women, P = .04, and 229 mg/d for white women, P < .001).

There was no significant difference between the CFFFQ and 24-hour recall methods for calculated calcium intake for grains, meats, and "other" foods. This pattern was consistent for the total group and for black and white women separately (Table 2). The Bland-Altman plot illustrates the lack of agreement between methods of assessment for total daily intake (Figure). The positive correlation between the lack of agreement (error score on the Y-axis) and the daily intake indicates that for increasing levels of daily calcium intake, the 24-hour recall of calcium intake increasingly underestimated calcium intake compared with the CFFFQ (r = 0.41, P < .001).

The prevalence of inadequate intake (<800 mg) was 56% (n = 103) using the 24-hour recall and 45% (n = 83) using the CFFFQ method. Examining the cases where intake would be inadequate as measured by both tools, 64 (35%) women would be classified as having inadequate calcium intake. However, 39 (21%) women whose intake would be classified as adequate by the CFFFQ would be classified as inadequate by the 24-hour recall method. Only 19 (10%) who would be classified as having inadequate intake by the CFFFQ would have adequate intake as classified by the 24-hour recall. Compared with the self-reported 24-hour recall, the CFFFQ indicates more specificity for lower intakes.

Regardless of dietary assessment method used, white women had higher calcium intakes than black women. When using the CFFFQ, white women reported consuming approximately 43% more calcium than did black women (mean [SD] 1,104 [632] mg for white women vs 768 [531] mg for black women, P < .001). When using the 24-hour recall method, mean calcium intake for white women was approximately 52% higher than intake for black women (875 [429] mg vs 573 [365] mg, P < .001).

From the CFFFQ, the primary calcium source was dairy products (55% for black women and 57% for white women). White women obtained more calcium (630 [423] mg) from dairy than did black women (424 [373] mg, P < .004). Grains were the second highest calcium source, although grains provided a much lower percentage than dairy (13% of total calcium for each racial group). Calcium from grains primarily came from fortified foods. Dairy was also the primary calcium source when the 24-hour recall data were analyzed, and a significant difference in mean (SD) dairy calcium intake between racial groups was found (243 [273] mg for black women, 444 [360] mg for white women, P < .001).

The values from the CFFFQ were typically higher than those from the 24-hour recall, with the exception of vegetables and "other" foods. This finding may be due in part to the limitations of using a 24-hour recall. Women may have identified foods in the CFFFQ but did not happen to consume those foods on the specific day of the recall, except for vegetables. However, Chee et al (24) found a similar trend in comparing a calcium-rich food frequency questionnaire with a 3-day diary for postmenopausal Malaysian women.

Assessing dietary intake by any method has inherent limitations. Validating food frequency questionnaires can be complicated without biomarkers for comparison. Because most nutrients do not yet have reliable biomarkers, 24-hour recalls or food records are the usual standard instruments. Correlations between these methods are considered adequate within the range of 0.4-0.7 (29). Although most of the correlations between total calcium and calcium from food groups when comparing CFFFQ to 24-hour recalls are significant, only total calcium intake, dairy calcium intake, and fruits fall within this acceptable statistical range.

Differences in reported calcium intake when comparing the CFFFQ and 24-hour recall could be attributed to a lack of consumer awareness of calcium fortification when responding to the 24-hour recall and a positive respondent bias on the CFFFQ. For example, participants could have responded more positively concerning calcium-fortified foods if consumption of that food was seen as a positive health behavior. Because calcium-fortified foods can essentially double calcium intake, consumer awareness of their own calcium-fortified food product consumption and interviewer probe for each of these food items are essential for an accurate estimate of calcium intake. To assist both the researcher and clinician, software for dietary assessment needs to be updated to include the calcium-fortified foods.

Several food frequency questionnaires that assess calcium intake have shown strong correlation with food actually consumed (19,22-25,30,31). However, many studies report correlations but no difference between means or report findings regarding only total calcium intake and not calcium intake by individual food groups. In addition, many rapid assessment tools include either no or limited sources of calcium-fortified foods, and these foods may or may not be probed for on 24-hour recalls.

Our results suggest that the CFFFQ could be used to determine inadequate intakes of calcium. Primary calcium sources for all women were dairy followed by grains, and calcium intake was higher in white women than in black women. The CFFFQ can be used to more accurately identify calcium intakes and usual calcium source, which is of interest because of the increasing availability of calcium-fortified foods. Calcium-fortified sources can go unreported in 24-hour recalls and diaries, leading to an underestimation of calcium intake. When 24-hour recalls are used to assess and quantify calcium intake, researchers and clinicians need to clarify with clients whether the foods they ate were fortified. The CFFFQ may be used to better quantify calcium intake, including both natural and fortified sources. Accurate assessment of calcium is critical in evaluating bone health risks. This information can aid in the development of effective interventions to increase calcium intake.