Journal of Human Nutrition and Dietetics
RESEARCH PAPER Validation of a food frequency questionnaire to assess food group intake by pregnant women P. Barbieri,1 L. C. Crivellenti,1 R. Y. Nishimura1 & D. S. Sartorelli1,2 Graduate Program in Community Health, Riber~ao Preto Medical School, University of S~ ao Paulo, Ribeir~ ao Preto, SP, Brazil Department of Social Medicine, Riber~ao Preto Medical School, University of S~ ao Paulo, Ribeir~ ao Preto, SP, Brazil
1 2
Keywords food consumption, food frequency questionnaire, food group, nutritional epidemiology, pregnant women, validation studies. Correspondence D. S. Sartorelli, Department of Social Medicine, Riber~ ao Preto Medical School, University of S~ao Paulo. Av. Bandeirantes, 3900 – 14049-900 – Ribeir~ ao Preto, SP, Brazil. Tel: +55 (16) 3602 2712 Fax: +55 (16) 3633 1386 E-mail: [email protected]. How to cite this article Barbieri P., Crivellenti L.C., Nishimura R.Y. & Sartorelli D.S. (2015) Validation of a food frequency questionnaire to assess food group intake by pregnant women. J Hum Nutr Diet. 28 (Suppl. 1), 38–44 doi:10.1111/jhn.12224
Abstract Background: Previous studies conducted among pregnant women to test the accuracy of food frequency questionnaires (FFQ) for estimating food group intake were restricted to one specific trimester of pregnancy. The present study aimed to validate a FFQ for assessing the intake of food groups throughout pregnancy. Methods: In total, 75 adult pregnant Brazilian women were evaluated. Dietary intake was assessed by the FFQ (completed at the third trimester of pregnancy) and by three 24-h dietary recalls; one in each trimester of pregnancy. Results: The food items were classified into 20 groups. Adequate deatenuatted Pearson correlation coefficients (>0.4) were observed for the intake of bread/cake, butter/margarine; milk/dairy products; soft drinks/artificial juices; coffee/tea; and pastries/sandwiches. The FFQ served poorly for estimating fruit and vegetable intake. A high percentage (>70%) of women were classified into the same or adjacent quartiles for estimates of cookies/ crackers, butter/margarine, milk/dairy products, fruit juices, soft drinks/artificial juices, coffee/tea, roots, rice, beans, meat/chicken/sausages, fried foods, fish, eggs, sweets/sugars, and pastries/sandwiches. Nevertheless, the agreement of joint classification between the dietary methods was mostly into adjacent quartiles, rather than in the same quartile, and low values of kappa were found. Conclusions: The data reported in the present study suggest that the FFQ might not be an appropriate dietary method for evaluating food group intake throughout pregnancy. The joint classification between methods by categories of intake of food groups was mostly into adjacent quartiles, which could lead to attenuated associations when investigating diet–disease relationships during pregnancy.
Introduction Studies suggest a compelling link between the exposure of a foetus to micronutrients from the maternal diet during gestation and the occurrence of chronic illness in adulthood (Scientific Advisory Committee on Nutrition, 2011). The consumption of specific food items by pregnant women has also been associated with the development of both maternal (Borgen et al., 2012) and foetal 38
health problems (Brantsaeter et al., 2012; Englund-Ogge et al., 2012). Research on these issues requires accurate methods for estimating food intake during pregnancy, which poses a challenge for studies of nutritional epidemiology (Meltzer et al., 2008). The food frequency questionnaire (FFQ) is considered an accurate method of estimating dietary intake during pregnancy (Erkkola et al., 2001; Pinto et al., 2010), and is capable of detecting changes in dietary ª 2014 The British Dietetic Association Ltd.
P. Barbieri et al.
intake during both the periconceptional and gestational periods (Brown et al., 1996). However, questionnaires developed and validated in other countries are not applicable to the Brazilian population because diet varies with ethnicity, customs, dietary taboos, food availability and socioeconomic conditions (Willett, 1998). Women’s energy and nutrient requirements, as well as food intake, undergo changes during pregnancy as a result of physiological and metabolic alterations, and cultural practices. For this reason, the use of dietary intake assessment methods that are not specifically designed for pregnancy may weaken correlations between diet and health outcomes. Studies from many parts of the world report on testing the validity of FFQ to estimate nutrient intake during pregnancy (Erkkola et al., 2001; Mouratidou et al., 2006; Brantsaeter et al., 2008; Pinto et al., 2010; Marhazlina et al., 2011; Barbieri et al., 2013). Nevertheless, studies testing the accuracy of the FFQ to estimate food group intake by pregnant women are scarce, and are restricted to one specific trimester of pregnancy (Erkkola et al., 2001; Mouratidou et al., 2006; Brantsaeter et al., 2008; Marhazlina et al., 2011). Testing the accuracy of a FFQ to estimate food group intake throughout pregnancy is therefore crucial. The present study aimed to test the relative validity of an FFQ for estimating food group intake by Brazilian women throughout pregnancy. Materials and methods Study design and subjects This was a prospective study conducted with 103 pregnant women attending for prenatal care at the Brazilian National Health Service in the city of Ribeir~ao Preto, S~ao Paulo State. A convenience sample was used, and the sample size was based on the recommendation that 100 individuals are sufficient to test for agreement between food intake assessment methods (Cade et al., 2002). Inclusion criteria were: age 18–35 years, prepregnancy body mass index (BMI) of 18.5–25.0 kg m–² Institute of Medicine (IOM) (2009), gestation period of no more than 14 weeks, and no report of pathologies known to alter food intake during pregnancy (e.g. gestational diabetes, heart disease, kidney disease and hypertension). Data were collected in four health clinics in the southern, eastern and western regions of the city by previously trained nutritionists. Pregnant women were assessed at the beginning of the study (at which time their gestation period was no more than 14 weeks), in the second trimester (between the weeks 14 and 28) and in the third trimester (after week 28). The first assessment was carried out at the first prenatal care visit, between September 2009 and May 2010. The second and third assessments ª 2014 The British Dietetic Association Ltd.
Validation of a FFQ for pregnant women
were carried out during prenatal care visits or interviews at home. Characteristics of the pregnant women Data on age, schooling level and socioeconomic condition were obtained with a structured questionnaire, and the socioeconomic level of subjects was determined using the Brazilian Economic Classification Criteria (ABEP, 2007), which defines classes from A (highest socioeconomic level) to E (lowest socioeconomic level). Date of last menstruation and ultrasonography were used to calculate gestational periods. Prepregnancy weight was provided by the subjects and the classification of nutritional status followed the criteria of the IOM (2009). The food frequency questionnaire A FFQ was developed for pregnant women attending public health clinics in Ribeir~ao Preto (Oliveira et al., 2010). Briefly, a 24-h dietary recall was obtained for 150 pregnant women (50 in each trimester of pregnancy). The study included healthy adult pregnant women with prepregnancy BMI between 18.5 and 26 kg m–2. The list of foods reported on the dietary recalls was systematically shortened using stepwise multiple regression analysis in which the nutrients of interest (adjusted for energy and variability) were the dependent variables. Eighty-five foods that explained the greatest proportions of interpersonal variation in nutrient intake (74–99%) were identified. The 25%, 50%, 75% and 100% quartiles were used to determine portion sizes. The final version of the FFQ contains 85 food items and four options of portion sizes, as well as detailed information on food preparation (fried, cooked) and type of oils used for cooking and salad dressings. The FFQ was tested for reproducibility and its estimates of nutrient intake were validated. Reproducibility was assessed in a study involving 95 pregnant women. The FFQ was applied twice, with 15–45 days between applications, and the mean raw intraclass correlation coefficient for estimating nutrient intake was 0.81 (0.70 after adjusting for energy). For 82.4% of pregnant women, the two completed FFQs placed them in the same quartile or in adjacent quartiles (Isobe et al., 2013). A prospective study was conducted among 103 women to validate the FFQ estimates of nutrient intake, in which adequate Pearson’s correlation coefficient (adjusted for energy and deattenuated; r > 0.35) were found for estimates of calcium, potassium, zinc, magnesium, fibre, vitamins C, niacin and folic acid compared to nutrient estimates from three 24-h recalls performed throughout the gestation. A high proportion of pregnant women 39
P. Barbieri et al.
Validation of a FFQ for pregnant women
(>70%) were classified in the same quartile or in adjacent quartiles for estimates of energy, carbohydrates, calcium, potassium, fibre, zinc and cholesterol, as well as vitamins A, riboflavin, niacin, C, E and folic acid, compared to nutrient estimates from the 24-h recalls (Barbieri et al., 2013). Validation of the food frequency questionnaire for food groups Validation studies are carried out to assess the accuracy of a FFQ to measure what it should measure, or to assess the agreement between the FFQ with a gold standard (Cade et al., 2002). To test the accuracy of the FFQ for estimating the intake of various food groups, the reported foods on food recalls (three 24-h recalls, one in each trimester of pregnancy) and the FFQ (obtained at the third trimester of pregnancy) were grouped into 20 food groups based on their nutrient content per 100 g of the food item. The food groups considered were: bread and cake; cookies and crackers; butter and margarine; milk and dairy products; soft drinks and artificial fruit juices; natural fruit juice; coffee and tea; fruits; vegetables; roots; rice; beans; pasta; red meat, chicken, and sausages; fried foods; fish; eggs; sweets; salty snacks and popcorn; and pastries and sandwiches. In the present study, the FFQ was applied in the third trimester of pregnancy and takes the confirmed date of pregnancy as the time reference. Three 24-h recalls were obtained for all women; one in each trimester of pregnancy. The multiple-pass technique was used to obtain the 24-h recalls into three steps: quick list (individuals were required to report all food and beverages consumed without interferences of the interviewer), detail cycle (details on time of consumption, how the food was prepared, and added items) and final revision (look over the food recall to determine whether the participant can remember anything else) (Johnson et al., 1998). To convert portion sizes to grams of food consumed, manuals developed for Brazil were used (Fisberg & Slater, 2002; Pinheiro et al., 2005). The FFQ and the 24-h recall data were analysed with DIETSYS, version 4.02 (National Cancer Institute, Bethesda, MD, USA) and NutWin, version 1.5 (Programa de Apoio a Nutricß~ao, Paulista School of Medicine, S~ao Paulo, Brazil), respectively. Statistical analysis The means (SD) of the continuous variables and the frequencies of the categorical variables were calculated. Median values (and interquartile ranges) were used to describe the dietary data. Dietary variables were tested for 40
normality with the Kolmogorov–Smirnov test and variables that were not normally distributed were natural logtransformed before statistical analysis. Between-person and within-person variances were obtained using analysis of variance. Raw and deattenuated Pearson’s correlation coefficients were used to validate the FFQ. The deattenuated Pearson’s correlation coefficient was calculated as: [1 + (r2 intra/r2 inter)/n]½, where n is the number of replicates of the 24-h recalls, r2 intra is the between-person variance, and r2 inter is the within-person variance between the 24-h recalls (Willett, 1998). Validation of the FFQ also included assessments of the degree of agreement between the methods: joint classification of estimated food group intake quartiles and weighted quadratic kappa. Statistical analyses were performed using SPSS, version 17.0 (SPSS Inc., Chicago, IL, USA). Ethical considerations The project was approved by the Research Ethics Committee of the Riber~ao Preto Medical School (Centro de Sa ude Escola Division) of the University of S~ao Paulo (permit number 239) and its execution was authorised by the Municipal Health Department of the Riber~ao Preto city government. All subjects who agreed to participate in the study provided their written informed consent. Results A total of 247 pregnant women were contacted. Of these, five (2%) did not agree to participate and 139 (56%) were omitted as a result of the exclusion criteria: 62 (25%) because of age, 45 (18%) as a result of inadequate prepregnancy nutritional status, and 32 (12.9%) because they were in the advanced stages of pregnancy. Thus, 103 pregnant women were included in the study, of whom 88 (85.4%) participated in the second assessment and 75 (70%) participated in the third. Of the subjects who were lost to follow-up, 10 had spontaneous abortions, five gave birth prematurely, five moved to another city and 11 were not located. Mean (SD) weight gain was 5.2 (2.8) kg between the first and second assessments and 9 3.1 kg between the first and third assessments. The mean time interval was 82 days (approximately 12 weeks) between the first and second assessments and 47 days (approximately 7 weeks) between the second and third assessments. Table 1 shows data on age, schooling level, socioeconomic class and pregestation BMI of the women studied. Sixty-five percent of women reported more than 8 years of schooling and 71% belonged to the ‘C’ socioeconomic ª 2014 The British Dietetic Association Ltd.
P. Barbieri et al.
Validation of a FFQ for pregnant women
Table 1 Age, education, socioeconomic class and nutritional status of pregnant women in Ribeir~ao Preto, S~ao Paulo, Brazil, at the initial evaluation in 2009–2010 (n = 75) Characteristics
Mean
(SD)
Age (years) Weeks of gestation Prepregnancy body mass index (kg m–²) Monthly family income (US$)
24 11 21.9 680
4.5 2.1 1.9 340
16 34 52
15.7 33 51
8 73 22
7.8 70.9 21.4
72 11 20 87 15 44 17
69.9 10.7 19.4 84.5 14.6 50 22.7
Education (years of schooling), n (%) Up to 4 4–8 >8 Socioeconomic class* A+B C D+E Head of family Subject’s partner/husband Subject Other Never consumes alcoholic beverages Smokes during pregnancy Takes folic acid daily Reports weekly practice of physical activity
*As defined by the Brazilian Economic Classification Criteria, which defines classes from A (highest socioeconomic level) to E (lowest socioeconomic level).
class, as defined by the Brazilian Economic Classification Criteria. Table 2 shows the median intake (and interquartile range) of food groups according to the FFQ and the 24-h recalls, as well as the corresponding Pearson’s correlation coefficient. The mean raw Pearson’s correlation coefficient was 0.27, ranging from 0.005 for vegetables to 0.51 for coffee and tea. The mean deattenuated correlation coefficient was 0.30, ranging from 0.005 for vegetables to 0.54 for coffee and tea. Deattenuated Pearson’s correlation coefficients considered to be adequate (>0.40) were found for the groups: bread/cake, butter/margarine; milk/dairy products; soft drinks/artificial juices; coffee/tea; and pastries/sandwiches. Joint classification of food group into quartiles of intake for the different methods and the weighted quadratic kappa are shown in Table 3. On average, 29% of pregnant women were classified in the same quartile (ranging from 21% for salty snacks to 40% for butter/ margarine) and 43% were classified into adjacent quartiles (ranging from 32% for pasta to 55% for milk/dairy products) by the FFQ and the 24-h recalls. The mean proportion of women classified in opposite extreme quartiles was 6%, ranging from 0% for fish to 11% for rice and sweets/sugars. The mean quadratic kappa coefficient was 0.26, ranging from 0.07 for eggs to 0.46 for soft drinks/artificial juices.
Table 2 Food intake as estimated by the food frequency questionnaire (FFQ) and 24-h dietary recalls and Pearson correlation coefficients between results obtained with the two methods, for pregnant women in Ribeir~ ao Preto, S~ ao Paulo, Brazil, 2009–2010 (n = 75) FFQ
24-h recalls
Food group
Median
Interquartile range
Median
Interquartile range
Raw correlation coefficient
Deattenuated correlation coefficient
Bread and cake Cookies and crackers Butter and margarine Milk and dairy products Natural fruit juice Soft drinks and artificial juices Coffee and tea Fruits Vegetables Roots Rice Beans Pasta Red meat, chicken, and sausages Fried food Fish Eggs Sweets and sugars Salty snacks and popcorn Pastries and sandwiches Mean
65.7 30.7 6.4 424.0 107.1 142.9 25.0 243.5 124.4 7.1 257.2 156.0 44.6 100.4 6.7 3.7 7.0 79.5 3.2 24.8 –
(46.5–108.6) (11.7–60.0) (4.3–10.0) (264.5–598.6) (8.3–214.3) (71.4–428.8) (3.6–71.4) (117.4–409.0) (71.4–180.6) (1.2–1.4.3) (170.0–340.0) (78.6–161.2) (21.6–62.9) (63.1–168.7) (0.6–20.0) (0.4–7.1) (1.6–13.7) (42.8–120.4) (0.0–13.7) (11.6–60.9) –
75.1 0.2 5.0 226.6 0.0 258.3 13.3 120.5 58.0 0.0 224.5 86.3 20.8 130.8 0.0 0.0 0.0 34.0 0.0 26.6 –
(48.7–106.0) (0.0–18.6) (1.6–10.0) (140.8–344.0) (0.0–166.6) (100.0–475.0) (0.0–47.5) (26.0–150.5) (19.3–109.0) (0.0–19.0) (141.6–312.0) (37.5–123.4) (0.0–94.6) (92.5–175.5) (0.0–8.0) (0.0–0.0) (0–10.4) (14.6–68.4) (0.0–0.0) (0.0–66.0) –
0.39 0.21 0.44 0.38 0.26 0.48 0.51 0.28 0.005 0.24 0.30 0.22 0.23 0.17 0.26 0.30 0.05 0.20 0.13 0.43 0.27
0.42 0.24 0.47 0.40 0.28 0.52 0.54 0.31 0.005 0.27 0.33 0.23 0.26 0.18 0.29 0.35 0.06 0.22 0.15 0.48 0.30
ª 2014 The British Dietetic Association Ltd.
41
P. Barbieri et al.
Validation of a FFQ for pregnant women
Table 3 Joint classification into quartiles for food group intake and the quadratic kappa statistic for the two methods, for pregnant women in Ribeir~ ao Preto, S~ ao Paulo, Brazil, 2009–2010 (n = 75).
Food group
Same quartile (%)
Adjacent quartiles (%)
Same quartile and adjacent quartiles (%)
Opposite quartiles (%)
Quadratic Kappa
Bread and cake Cookies and crackers Butter and margarine Milk and dairy products Natural fruit juice Soft drinks and artificial juices Coffee and tea Fruits Vegetables Roots Rice Beans Pasta Red meat, chicken, and sausages Fried food Fish Eggs Sweets and sugars Salty snacks and popcorn Pastries and sandwiches Mean
31 29 40 28 28 36 39 31 25 37 31 39 27 31 32 28 31 33 21 37 29
39 41 43 55 45 48 39 33 44 40 51 35 32 45 47 48 43 44 51 36 43
69 71 83 83 73 84 77 64 69 77 81 73 59 76 79 76 73 77 72 73 75
7 1 3 5 1 4 9 9 7 5 11 5 8 9 3 0 7 11 3 8 6
0.22 0.17 0.43 0.39 0.22 0.46 0.44 0.12 0.20 0.22 0.30 0.31 0.19 0.24 0.29 0.21 0.07 0.24 0.14 0.38 0.26
Discussion Previous studies that tested the accuracy of the FFQ to estimate food intake by pregnant women were restricted to dietary intake at the second (Mouratidou et al., 2006; Brantsaeter et al., 2008) or third trimester of pregnancy (Erkkola et al., 2001; Marhazlina et al., 2011). The present study was the first FFQ validation study to consider food intake throughout pregnancy. On average, Pearson’s correlation coefficients in the present study were low. Satisfactory correlation coefficients (>0.4) were observed for the estimates of six food groups (30% of the tested food groups): bread/cake, butter/margarine; milk/dairy products; soft drinks/artificial juices; coffee/tea; and pastries/sandwiches. In previous studies conducted in Norway (Brantsaeter et al., 2008) and the UK (Mouratidou et al., 2006), the FFQ was accurate in estimating 63% and 82% of the tested food groups compared to weighted food diaries and two food recalls at the second trimester of pregnancy, respectively. In a validation study conducted on Finland, the FFQ, as obtained at the third trimester of pregnancy, was accurate in estimating 72% of the tested food groups compared to two food records of five consecutive days at the eighth month of pregnancy (Erkkola et al., 2001). However, in a study conducted in Malaysia at the third trimester of pregnancy, the FFQ was accurate in estimating only 30% 42
of the tested food groups (Marhazlina et al., 2011). The discrepancies between our results and those reported in previous studies might be partly explained by the high within-person variation in dietary intake throughout pregnancy (Nyambose et al., 2002), thereby reducing the correlation between the different dietary methods. According to Beaton (1991), the fact that FFQs overestimate food intake estimates is not a problem in epidemiological studies aiming to identify diet–disease associations as long as the classification of individuals into intake levels is accurate. Considering the joint classification into categories of food intake by the two methods, the data obtained in the studies conducted in Finland (Erkkola et al., 2001) and Norway (Brantsaeter et al., 2008), showed that, for 53% and 45% of the tested food groups, respectively, more than 70% of women were classified into the same or adjacent quintiles of intake. In the study conducted in Malaysia, 99% of the tested food groups were classified into the same or adjacent quartiles for both methods (Marhazlina et al., 2011). In the present study, a high agreement (≥70%) into the same or adjacent quartiles was found for 85% of the tested food groups. Nevertheless, the agreement was mostly explained by the classification into the adjacent quartile than into the same quartile by two dietary methods. Unfortunately, the previous studies conducted among pregnant women did not provide such details, limiting any comparisons. ª 2014 The British Dietetic Association Ltd.
P. Barbieri et al.
Previous studies aiming to validate questionnaires designed and validated for pregnant women suggest that this method is a valuable tool for assessing nutrient intake by pregnant women (Erkkola et al., 2001; Baer et al., 2005; Mouratidou et al., 2006; Pinto et al., 2010; Barbieri et al., 2013) because comparisons with other methods yield acceptable correlation coefficient values. However, recent studies recommend against using the Pearson’s coefficient in isolation to test for agreement between methods, and argue that it should be used in association with other types of analysis, such as agreement into the same categories, the weighted kappa statistic and the Bland–Altman method (Cade et al., 2002). In the present study, analyses with the Bland–Altman method were not possible because the final sample size was insufficient (Bland & Altman, 1995). The mean quadratic kappa value in the present study was low, in agreement with FFQ validation of food groups in an earlier study of Brazilian students, in which kappa values ranged from 0.15 for pasta to 0.56 for beans (Voci et al., 2008). The FFQ was not effective at estimating intake of fruits and vegetables, as observed in other FFQ validation studies for food groups (Matarazzo et al., 2006). The intake of these foods is highly recommended during pregnancy, and the FFQ may overestimate their consumption, which could increase disagreement between the tests. The most important limitation of the present study was the limited number of 24-h recalls. Given the high within-person variation in diet among the women, a large number of 24-h recall replicates are recommended. Moreover, a higher number of replicates of food recalls in each trimester of pregnancy could provide information about the agreement between the FFQ and food recalls from each trimester individually. Another limitation of the present study was the use of a convenience sample and restriction to women with normal BMI in the prepregnancy period. The study focused on normal weight women during the prepregnancy period as a result of the observed tendency for under-reporting by overweight women, which would have compromised the results of the present study. Another limitation is the fact that the FFQ and 24-h recalls contain similar errors, such as recall bias. The main contribution of the present study was to evaluate the food intake throughout pregnancy, including data with respect to dietary intake from each trimester of pregnancy. Our data do not support using the FFQ to estimate food intake during pregnancy taking the gestational period as a whole. Combined data from short-term methods and the frequency of intake have been suggested recently (Harttig et al., 2011) and should be tested when investigating diet–disease relationships during pregnancy. ª 2014 The British Dietetic Association Ltd.
Validation of a FFQ for pregnant women
Conflict of interests, source of funding and authorship The authors declare that they have no conflict of interests. The project received financial support from the Coordenacß~ao de Aperfeicßoamento de Pessoal de Nıvel Superior (CAPES, master’s grant to PB), Fundacß~ao de Amparo a Pesquisa do Estado de S~ao Paulo [FAPESP, master’s grant to LCC (2011/-03781-8) and RYN (2010/12320-1)] and the Teaching and Research Support Foundation at the Clinical Hospital of the Ribeir~ao Preto Medical School of the University of S~ao Paulo (FAEPA, research grant). DSS and PB were responsible for designing the research project, analysing and interpreting the data, and writing the manuscript. LCC and RYN were responsible for data collection, analysing and interpreting the data, and reviewing the manuscript. All authors approved the final version submitted for publication.
References Associacß~ao Brasileira de Empresa de Pesquisas [Brazilian Economic Classification Criteria]. (2007) Dados com Base no Levantamento Socioecon^omico. Available at: http://www.abep. org (Accessed at January 2012). Baer, H.J., Blum, R.E., Rockett, H.R., Leppert, J., Gardner, J.D., Suitor, C.W. & Colditz, G.A. (2005) Use of a food frequency questionnaire in American Indian and Caucasian pregnant women: a validation study. BMC Public Health 5, 135. Barbieri, P., Nishimura, R.Y., Crivellenti, L.C. & Sartorelli, D.S. (2013) Relative validation of a quantitative FFQ for use in Brazilian pregnant women. Public Health Nutr. 16, 1419– 1426. Beaton, G.H. (1991) Interpretation of results from diet history studies. In The diet history method. ed L. Kohlmeier. Smith-Gordon: London. Bland, J.M. & Altman, D.G. (1995) Comparing methods of measurement: why plotting difference against standard method is misleading. Lancet 345, 1085–1087. Borgen, I., Aamodt, G., Harsem, N., Haugen, M., Meltzer, H.M. & Brantsaeter, A.L. (2012) Maternal sugar consumption and risk of preeclampsia in nulliparous Norwegian women. Eur. J. Clin. Nutr. 66, 920–925. Brantsaeter, A.L., Birgisdottir, B.E., Meltzer, H.M., Kvalem, H.E., Alexander, J., Magnus, P. & Haugen, M. (2012) Maternal seafood consumption and infant birth weight, length and head circumference in the Norwegian Mother and Child Cohort Study. Br. J. Nutr. 107, 436–444. Brantsaeter, A.L., Haugen, M., Alexander, J. & Meltzer, H.M. (2008) Validity of a new food frequency questionnaire for 43
Validation of a FFQ for pregnant women
pregnant women in the Norwegian Mother and Child Cohort Study (MoBa). Matern. Child Nutr. 4, 28–43. Brown, J.E., Buzzard, I.M., Jacobs, D.R. Jr, Hannan, P.J., Kushi, L.H., Barosso, G.M. & Schmid, L.A. (1996) A food frequency questinnaire can detect pregnancy-related changes in diet. J. Am. Diet. Assoc. 96, 262–266. Cade, J., Thompson, R., Burley, V. & Warm, D. (2002) Development, validation and utilization of food-frequency questionnaires – a review. Public Health Nutr. 5, 567–587. Englund-Ogge, L., Brantsaeter, A.L., Haugen, M., Sengpiel, V., Khatibi, A., Myhre, R., Myking, S., Meltzer, H.M., Kacerovsky, M., Nilsen, R.M. & Jacobsson, B. (2012) Association between intake of artificially sweetened and sugar-sweetened beverages and preterm delivery: a large prospective cohort study. Am. J. Clin. Nutr. 96, 552–559. Erkkola, M., Karppinen, M., Javanainen, J., R€as€anen, L., Knip, M. & Virtanen, S.M. (2001) Validity and reproducibility of a food frequency questionnaire for pregnant Finnish women. Am. J. Epidemiol. 154, 466–476. Fisberg, R.F. & Slater, B. (2002) Manual de receitas e medidas caseiras para calculo de inqueritos alimentares: manual elaborado para auxiliar o processamento de inqueritos alimentares. S~ao Paulo: Signus. Harttig, U., Haubrock, J., Knuppel, S. & Boeing, H. & EFCOVAL Consortium (2011). The MSM program: web-based statistics package for estimating usual dietary intake using the Multiple Source Method. Eur. J. Clin. Nutr., 65: 87–91. Institute of Medicine (IOM). (2009) Weight Gain During Pregnancy: Reexamining the Guidelines. Washington, DC: National Academy Press. Isobe, M.T., Bertola, M.R., Zucolotto, D.C.C. & Sartorelli, D.S. (2013) The influence of the educational status on the reproducibility of a quantitative food frequency questionnaire for pregnant women. Rev. Bras. Saude Mater Infant 13, 23–28. Johnson, R.K., Soultanakis, R.P. & Matthews, D.E. (1998) Literacy and body fatness are associated with underreporting of energy intake in US low income women using the multiple-pass 24 hours recall: a doubly labeled water study. J. Am. Diet. Assoc. 98, 1136–1140. Marhazlina, L.S.L., Azwany, N. & Hamid Jan, J.M. (2011) Development, validity and reproducibility of a food frequency questionnaire in pregnancy for the
44
P. Barbieri et al.
Universiti Sains Malaysia Birth Cohort Study. Malays. J. Nutr. 1, 1–18. Matarazzo, H.C.Z., Marchioni, D.M.L., Figueiredo, R.A.O., Slater, B., Eluf Neto, J. & W€ unsch Filho, V. (2006) Reprodutibilidade e validade do questionario de freq€ u^encia de consumo alimentar utilizado em estudo caso-controle de c^ancer oral. Rev. Bras. Epidemiol. 9, 316–324. Meltzer, H.M., Brantsaeter, A.L., Ydersbond, T.A., Alexander, J. & Haugen, M. (2008) Methodological challenges when monitoring the diet of ant women in a large study: experiences from the Norwegian Mother and Sy (MoBa). Matern. Child Nutr. 4, 14–27. Mouratidou, T., Ford, F. & Fraser, R.B. (2006) Validation of a food-frequency questionnaire for use in pregnancy. Public Health Nutr. 9, 515–522. Nyambose, J., Koshi, K.G. & Tucker, K.L. (2002) High intra/ interindividual variance ratios for energy intakes of pregnant women in rural Malawi show that many days are required to estimate usual intake. J. Nutr. 132, 1313–1318. Oliveira, T., Marquitti, F.D., Carvalhaes, M.A. & Sartorelli, D.S. (2010) Desenvolvimento de um Questionario Quantitativo de Freq€ u^encia Alimentar (QQFA) para gestantes usuarias de unidades basicas de sa ude de Ribeir~ao Preto, S~ao Paulo, Brasil. Cad. Sa ude P ublica 26, 2296–2306. Pinheiro, A.B.V., Lacerda, E.M.A., Benzecry, E.H., Gomes, M.C.S. & Costa, V.M. (2005) Tabela para avaliacß~ao de consumo alimentar em medidas caseiras, 5th edn. S~ao Paulo: Editora Atheneu. Pinto, E., Severo, M., Correia, S., Silva, I.S., Lopes, C. & Barros, H. (2010) Validity and reproducibility of a semi-quantitative food frequency questionnaire for use among portuguese pregnant women. Matern. Child Nutr. 6, 105–119. Scientific Advisory Committee on Nutrition. (2011) The Influence of Maternal, Fetal and Child Nutrition on the Development of Chronic Disease in Later Life. Available at: http://www.sacn.gov.uk/pdfs/sacn_early_nutrition_final_ report_20_6_11.pdf. (Accessed on 5 November 2012). Voci, S.M., Enes, C.C. & Slater, B. (2008) Validacß~ao do Questionario de Freq€ u^encia Alimentar para Adolescentes (QFAA) por grupos de alimentos em uma populacß~ao de escolares. Rev. Bras. Epidemiol. 11, 561–572. Willett, W.C. (1998) Nutritional Epidemiology, 2nd edn. New York, NY: Oxford University Press.
ª 2014 The British Dietetic Association Ltd.
RESEARCH PAPER Validation of a food frequency questionnaire to assess food group intake by pregnant women P. Barbieri,1 L. C. Crivellenti,1 R. Y. Nishimura1 & D. S. Sartorelli1,2 Graduate Program in Community Health, Riber~ao Preto Medical School, University of S~ ao Paulo, Ribeir~ ao Preto, SP, Brazil Department of Social Medicine, Riber~ao Preto Medical School, University of S~ ao Paulo, Ribeir~ ao Preto, SP, Brazil
1 2
Keywords food consumption, food frequency questionnaire, food group, nutritional epidemiology, pregnant women, validation studies. Correspondence D. S. Sartorelli, Department of Social Medicine, Riber~ ao Preto Medical School, University of S~ao Paulo. Av. Bandeirantes, 3900 – 14049-900 – Ribeir~ ao Preto, SP, Brazil. Tel: +55 (16) 3602 2712 Fax: +55 (16) 3633 1386 E-mail: [email protected]. How to cite this article Barbieri P., Crivellenti L.C., Nishimura R.Y. & Sartorelli D.S. (2015) Validation of a food frequency questionnaire to assess food group intake by pregnant women. J Hum Nutr Diet. 28 (Suppl. 1), 38–44 doi:10.1111/jhn.12224
Abstract Background: Previous studies conducted among pregnant women to test the accuracy of food frequency questionnaires (FFQ) for estimating food group intake were restricted to one specific trimester of pregnancy. The present study aimed to validate a FFQ for assessing the intake of food groups throughout pregnancy. Methods: In total, 75 adult pregnant Brazilian women were evaluated. Dietary intake was assessed by the FFQ (completed at the third trimester of pregnancy) and by three 24-h dietary recalls; one in each trimester of pregnancy. Results: The food items were classified into 20 groups. Adequate deatenuatted Pearson correlation coefficients (>0.4) were observed for the intake of bread/cake, butter/margarine; milk/dairy products; soft drinks/artificial juices; coffee/tea; and pastries/sandwiches. The FFQ served poorly for estimating fruit and vegetable intake. A high percentage (>70%) of women were classified into the same or adjacent quartiles for estimates of cookies/ crackers, butter/margarine, milk/dairy products, fruit juices, soft drinks/artificial juices, coffee/tea, roots, rice, beans, meat/chicken/sausages, fried foods, fish, eggs, sweets/sugars, and pastries/sandwiches. Nevertheless, the agreement of joint classification between the dietary methods was mostly into adjacent quartiles, rather than in the same quartile, and low values of kappa were found. Conclusions: The data reported in the present study suggest that the FFQ might not be an appropriate dietary method for evaluating food group intake throughout pregnancy. The joint classification between methods by categories of intake of food groups was mostly into adjacent quartiles, which could lead to attenuated associations when investigating diet–disease relationships during pregnancy.
Introduction Studies suggest a compelling link between the exposure of a foetus to micronutrients from the maternal diet during gestation and the occurrence of chronic illness in adulthood (Scientific Advisory Committee on Nutrition, 2011). The consumption of specific food items by pregnant women has also been associated with the development of both maternal (Borgen et al., 2012) and foetal 38
health problems (Brantsaeter et al., 2012; Englund-Ogge et al., 2012). Research on these issues requires accurate methods for estimating food intake during pregnancy, which poses a challenge for studies of nutritional epidemiology (Meltzer et al., 2008). The food frequency questionnaire (FFQ) is considered an accurate method of estimating dietary intake during pregnancy (Erkkola et al., 2001; Pinto et al., 2010), and is capable of detecting changes in dietary ª 2014 The British Dietetic Association Ltd.
P. Barbieri et al.
intake during both the periconceptional and gestational periods (Brown et al., 1996). However, questionnaires developed and validated in other countries are not applicable to the Brazilian population because diet varies with ethnicity, customs, dietary taboos, food availability and socioeconomic conditions (Willett, 1998). Women’s energy and nutrient requirements, as well as food intake, undergo changes during pregnancy as a result of physiological and metabolic alterations, and cultural practices. For this reason, the use of dietary intake assessment methods that are not specifically designed for pregnancy may weaken correlations between diet and health outcomes. Studies from many parts of the world report on testing the validity of FFQ to estimate nutrient intake during pregnancy (Erkkola et al., 2001; Mouratidou et al., 2006; Brantsaeter et al., 2008; Pinto et al., 2010; Marhazlina et al., 2011; Barbieri et al., 2013). Nevertheless, studies testing the accuracy of the FFQ to estimate food group intake by pregnant women are scarce, and are restricted to one specific trimester of pregnancy (Erkkola et al., 2001; Mouratidou et al., 2006; Brantsaeter et al., 2008; Marhazlina et al., 2011). Testing the accuracy of a FFQ to estimate food group intake throughout pregnancy is therefore crucial. The present study aimed to test the relative validity of an FFQ for estimating food group intake by Brazilian women throughout pregnancy. Materials and methods Study design and subjects This was a prospective study conducted with 103 pregnant women attending for prenatal care at the Brazilian National Health Service in the city of Ribeir~ao Preto, S~ao Paulo State. A convenience sample was used, and the sample size was based on the recommendation that 100 individuals are sufficient to test for agreement between food intake assessment methods (Cade et al., 2002). Inclusion criteria were: age 18–35 years, prepregnancy body mass index (BMI) of 18.5–25.0 kg m–² Institute of Medicine (IOM) (2009), gestation period of no more than 14 weeks, and no report of pathologies known to alter food intake during pregnancy (e.g. gestational diabetes, heart disease, kidney disease and hypertension). Data were collected in four health clinics in the southern, eastern and western regions of the city by previously trained nutritionists. Pregnant women were assessed at the beginning of the study (at which time their gestation period was no more than 14 weeks), in the second trimester (between the weeks 14 and 28) and in the third trimester (after week 28). The first assessment was carried out at the first prenatal care visit, between September 2009 and May 2010. The second and third assessments ª 2014 The British Dietetic Association Ltd.
Validation of a FFQ for pregnant women
were carried out during prenatal care visits or interviews at home. Characteristics of the pregnant women Data on age, schooling level and socioeconomic condition were obtained with a structured questionnaire, and the socioeconomic level of subjects was determined using the Brazilian Economic Classification Criteria (ABEP, 2007), which defines classes from A (highest socioeconomic level) to E (lowest socioeconomic level). Date of last menstruation and ultrasonography were used to calculate gestational periods. Prepregnancy weight was provided by the subjects and the classification of nutritional status followed the criteria of the IOM (2009). The food frequency questionnaire A FFQ was developed for pregnant women attending public health clinics in Ribeir~ao Preto (Oliveira et al., 2010). Briefly, a 24-h dietary recall was obtained for 150 pregnant women (50 in each trimester of pregnancy). The study included healthy adult pregnant women with prepregnancy BMI between 18.5 and 26 kg m–2. The list of foods reported on the dietary recalls was systematically shortened using stepwise multiple regression analysis in which the nutrients of interest (adjusted for energy and variability) were the dependent variables. Eighty-five foods that explained the greatest proportions of interpersonal variation in nutrient intake (74–99%) were identified. The 25%, 50%, 75% and 100% quartiles were used to determine portion sizes. The final version of the FFQ contains 85 food items and four options of portion sizes, as well as detailed information on food preparation (fried, cooked) and type of oils used for cooking and salad dressings. The FFQ was tested for reproducibility and its estimates of nutrient intake were validated. Reproducibility was assessed in a study involving 95 pregnant women. The FFQ was applied twice, with 15–45 days between applications, and the mean raw intraclass correlation coefficient for estimating nutrient intake was 0.81 (0.70 after adjusting for energy). For 82.4% of pregnant women, the two completed FFQs placed them in the same quartile or in adjacent quartiles (Isobe et al., 2013). A prospective study was conducted among 103 women to validate the FFQ estimates of nutrient intake, in which adequate Pearson’s correlation coefficient (adjusted for energy and deattenuated; r > 0.35) were found for estimates of calcium, potassium, zinc, magnesium, fibre, vitamins C, niacin and folic acid compared to nutrient estimates from three 24-h recalls performed throughout the gestation. A high proportion of pregnant women 39
P. Barbieri et al.
Validation of a FFQ for pregnant women
(>70%) were classified in the same quartile or in adjacent quartiles for estimates of energy, carbohydrates, calcium, potassium, fibre, zinc and cholesterol, as well as vitamins A, riboflavin, niacin, C, E and folic acid, compared to nutrient estimates from the 24-h recalls (Barbieri et al., 2013). Validation of the food frequency questionnaire for food groups Validation studies are carried out to assess the accuracy of a FFQ to measure what it should measure, or to assess the agreement between the FFQ with a gold standard (Cade et al., 2002). To test the accuracy of the FFQ for estimating the intake of various food groups, the reported foods on food recalls (three 24-h recalls, one in each trimester of pregnancy) and the FFQ (obtained at the third trimester of pregnancy) were grouped into 20 food groups based on their nutrient content per 100 g of the food item. The food groups considered were: bread and cake; cookies and crackers; butter and margarine; milk and dairy products; soft drinks and artificial fruit juices; natural fruit juice; coffee and tea; fruits; vegetables; roots; rice; beans; pasta; red meat, chicken, and sausages; fried foods; fish; eggs; sweets; salty snacks and popcorn; and pastries and sandwiches. In the present study, the FFQ was applied in the third trimester of pregnancy and takes the confirmed date of pregnancy as the time reference. Three 24-h recalls were obtained for all women; one in each trimester of pregnancy. The multiple-pass technique was used to obtain the 24-h recalls into three steps: quick list (individuals were required to report all food and beverages consumed without interferences of the interviewer), detail cycle (details on time of consumption, how the food was prepared, and added items) and final revision (look over the food recall to determine whether the participant can remember anything else) (Johnson et al., 1998). To convert portion sizes to grams of food consumed, manuals developed for Brazil were used (Fisberg & Slater, 2002; Pinheiro et al., 2005). The FFQ and the 24-h recall data were analysed with DIETSYS, version 4.02 (National Cancer Institute, Bethesda, MD, USA) and NutWin, version 1.5 (Programa de Apoio a Nutricß~ao, Paulista School of Medicine, S~ao Paulo, Brazil), respectively. Statistical analysis The means (SD) of the continuous variables and the frequencies of the categorical variables were calculated. Median values (and interquartile ranges) were used to describe the dietary data. Dietary variables were tested for 40
normality with the Kolmogorov–Smirnov test and variables that were not normally distributed were natural logtransformed before statistical analysis. Between-person and within-person variances were obtained using analysis of variance. Raw and deattenuated Pearson’s correlation coefficients were used to validate the FFQ. The deattenuated Pearson’s correlation coefficient was calculated as: [1 + (r2 intra/r2 inter)/n]½, where n is the number of replicates of the 24-h recalls, r2 intra is the between-person variance, and r2 inter is the within-person variance between the 24-h recalls (Willett, 1998). Validation of the FFQ also included assessments of the degree of agreement between the methods: joint classification of estimated food group intake quartiles and weighted quadratic kappa. Statistical analyses were performed using SPSS, version 17.0 (SPSS Inc., Chicago, IL, USA). Ethical considerations The project was approved by the Research Ethics Committee of the Riber~ao Preto Medical School (Centro de Sa ude Escola Division) of the University of S~ao Paulo (permit number 239) and its execution was authorised by the Municipal Health Department of the Riber~ao Preto city government. All subjects who agreed to participate in the study provided their written informed consent. Results A total of 247 pregnant women were contacted. Of these, five (2%) did not agree to participate and 139 (56%) were omitted as a result of the exclusion criteria: 62 (25%) because of age, 45 (18%) as a result of inadequate prepregnancy nutritional status, and 32 (12.9%) because they were in the advanced stages of pregnancy. Thus, 103 pregnant women were included in the study, of whom 88 (85.4%) participated in the second assessment and 75 (70%) participated in the third. Of the subjects who were lost to follow-up, 10 had spontaneous abortions, five gave birth prematurely, five moved to another city and 11 were not located. Mean (SD) weight gain was 5.2 (2.8) kg between the first and second assessments and 9 3.1 kg between the first and third assessments. The mean time interval was 82 days (approximately 12 weeks) between the first and second assessments and 47 days (approximately 7 weeks) between the second and third assessments. Table 1 shows data on age, schooling level, socioeconomic class and pregestation BMI of the women studied. Sixty-five percent of women reported more than 8 years of schooling and 71% belonged to the ‘C’ socioeconomic ª 2014 The British Dietetic Association Ltd.
P. Barbieri et al.
Validation of a FFQ for pregnant women
Table 1 Age, education, socioeconomic class and nutritional status of pregnant women in Ribeir~ao Preto, S~ao Paulo, Brazil, at the initial evaluation in 2009–2010 (n = 75) Characteristics
Mean
(SD)
Age (years) Weeks of gestation Prepregnancy body mass index (kg m–²) Monthly family income (US$)
24 11 21.9 680
4.5 2.1 1.9 340
16 34 52
15.7 33 51
8 73 22
7.8 70.9 21.4
72 11 20 87 15 44 17
69.9 10.7 19.4 84.5 14.6 50 22.7
Education (years of schooling), n (%) Up to 4 4–8 >8 Socioeconomic class* A+B C D+E Head of family Subject’s partner/husband Subject Other Never consumes alcoholic beverages Smokes during pregnancy Takes folic acid daily Reports weekly practice of physical activity
*As defined by the Brazilian Economic Classification Criteria, which defines classes from A (highest socioeconomic level) to E (lowest socioeconomic level).
class, as defined by the Brazilian Economic Classification Criteria. Table 2 shows the median intake (and interquartile range) of food groups according to the FFQ and the 24-h recalls, as well as the corresponding Pearson’s correlation coefficient. The mean raw Pearson’s correlation coefficient was 0.27, ranging from 0.005 for vegetables to 0.51 for coffee and tea. The mean deattenuated correlation coefficient was 0.30, ranging from 0.005 for vegetables to 0.54 for coffee and tea. Deattenuated Pearson’s correlation coefficients considered to be adequate (>0.40) were found for the groups: bread/cake, butter/margarine; milk/dairy products; soft drinks/artificial juices; coffee/tea; and pastries/sandwiches. Joint classification of food group into quartiles of intake for the different methods and the weighted quadratic kappa are shown in Table 3. On average, 29% of pregnant women were classified in the same quartile (ranging from 21% for salty snacks to 40% for butter/ margarine) and 43% were classified into adjacent quartiles (ranging from 32% for pasta to 55% for milk/dairy products) by the FFQ and the 24-h recalls. The mean proportion of women classified in opposite extreme quartiles was 6%, ranging from 0% for fish to 11% for rice and sweets/sugars. The mean quadratic kappa coefficient was 0.26, ranging from 0.07 for eggs to 0.46 for soft drinks/artificial juices.
Table 2 Food intake as estimated by the food frequency questionnaire (FFQ) and 24-h dietary recalls and Pearson correlation coefficients between results obtained with the two methods, for pregnant women in Ribeir~ ao Preto, S~ ao Paulo, Brazil, 2009–2010 (n = 75) FFQ
24-h recalls
Food group
Median
Interquartile range
Median
Interquartile range
Raw correlation coefficient
Deattenuated correlation coefficient
Bread and cake Cookies and crackers Butter and margarine Milk and dairy products Natural fruit juice Soft drinks and artificial juices Coffee and tea Fruits Vegetables Roots Rice Beans Pasta Red meat, chicken, and sausages Fried food Fish Eggs Sweets and sugars Salty snacks and popcorn Pastries and sandwiches Mean
65.7 30.7 6.4 424.0 107.1 142.9 25.0 243.5 124.4 7.1 257.2 156.0 44.6 100.4 6.7 3.7 7.0 79.5 3.2 24.8 –
(46.5–108.6) (11.7–60.0) (4.3–10.0) (264.5–598.6) (8.3–214.3) (71.4–428.8) (3.6–71.4) (117.4–409.0) (71.4–180.6) (1.2–1.4.3) (170.0–340.0) (78.6–161.2) (21.6–62.9) (63.1–168.7) (0.6–20.0) (0.4–7.1) (1.6–13.7) (42.8–120.4) (0.0–13.7) (11.6–60.9) –
75.1 0.2 5.0 226.6 0.0 258.3 13.3 120.5 58.0 0.0 224.5 86.3 20.8 130.8 0.0 0.0 0.0 34.0 0.0 26.6 –
(48.7–106.0) (0.0–18.6) (1.6–10.0) (140.8–344.0) (0.0–166.6) (100.0–475.0) (0.0–47.5) (26.0–150.5) (19.3–109.0) (0.0–19.0) (141.6–312.0) (37.5–123.4) (0.0–94.6) (92.5–175.5) (0.0–8.0) (0.0–0.0) (0–10.4) (14.6–68.4) (0.0–0.0) (0.0–66.0) –
0.39 0.21 0.44 0.38 0.26 0.48 0.51 0.28 0.005 0.24 0.30 0.22 0.23 0.17 0.26 0.30 0.05 0.20 0.13 0.43 0.27
0.42 0.24 0.47 0.40 0.28 0.52 0.54 0.31 0.005 0.27 0.33 0.23 0.26 0.18 0.29 0.35 0.06 0.22 0.15 0.48 0.30
ª 2014 The British Dietetic Association Ltd.
41
P. Barbieri et al.
Validation of a FFQ for pregnant women
Table 3 Joint classification into quartiles for food group intake and the quadratic kappa statistic for the two methods, for pregnant women in Ribeir~ ao Preto, S~ ao Paulo, Brazil, 2009–2010 (n = 75).
Food group
Same quartile (%)
Adjacent quartiles (%)
Same quartile and adjacent quartiles (%)
Opposite quartiles (%)
Quadratic Kappa
Bread and cake Cookies and crackers Butter and margarine Milk and dairy products Natural fruit juice Soft drinks and artificial juices Coffee and tea Fruits Vegetables Roots Rice Beans Pasta Red meat, chicken, and sausages Fried food Fish Eggs Sweets and sugars Salty snacks and popcorn Pastries and sandwiches Mean
31 29 40 28 28 36 39 31 25 37 31 39 27 31 32 28 31 33 21 37 29
39 41 43 55 45 48 39 33 44 40 51 35 32 45 47 48 43 44 51 36 43
69 71 83 83 73 84 77 64 69 77 81 73 59 76 79 76 73 77 72 73 75
7 1 3 5 1 4 9 9 7 5 11 5 8 9 3 0 7 11 3 8 6
0.22 0.17 0.43 0.39 0.22 0.46 0.44 0.12 0.20 0.22 0.30 0.31 0.19 0.24 0.29 0.21 0.07 0.24 0.14 0.38 0.26
Discussion Previous studies that tested the accuracy of the FFQ to estimate food intake by pregnant women were restricted to dietary intake at the second (Mouratidou et al., 2006; Brantsaeter et al., 2008) or third trimester of pregnancy (Erkkola et al., 2001; Marhazlina et al., 2011). The present study was the first FFQ validation study to consider food intake throughout pregnancy. On average, Pearson’s correlation coefficients in the present study were low. Satisfactory correlation coefficients (>0.4) were observed for the estimates of six food groups (30% of the tested food groups): bread/cake, butter/margarine; milk/dairy products; soft drinks/artificial juices; coffee/tea; and pastries/sandwiches. In previous studies conducted in Norway (Brantsaeter et al., 2008) and the UK (Mouratidou et al., 2006), the FFQ was accurate in estimating 63% and 82% of the tested food groups compared to weighted food diaries and two food recalls at the second trimester of pregnancy, respectively. In a validation study conducted on Finland, the FFQ, as obtained at the third trimester of pregnancy, was accurate in estimating 72% of the tested food groups compared to two food records of five consecutive days at the eighth month of pregnancy (Erkkola et al., 2001). However, in a study conducted in Malaysia at the third trimester of pregnancy, the FFQ was accurate in estimating only 30% 42
of the tested food groups (Marhazlina et al., 2011). The discrepancies between our results and those reported in previous studies might be partly explained by the high within-person variation in dietary intake throughout pregnancy (Nyambose et al., 2002), thereby reducing the correlation between the different dietary methods. According to Beaton (1991), the fact that FFQs overestimate food intake estimates is not a problem in epidemiological studies aiming to identify diet–disease associations as long as the classification of individuals into intake levels is accurate. Considering the joint classification into categories of food intake by the two methods, the data obtained in the studies conducted in Finland (Erkkola et al., 2001) and Norway (Brantsaeter et al., 2008), showed that, for 53% and 45% of the tested food groups, respectively, more than 70% of women were classified into the same or adjacent quintiles of intake. In the study conducted in Malaysia, 99% of the tested food groups were classified into the same or adjacent quartiles for both methods (Marhazlina et al., 2011). In the present study, a high agreement (≥70%) into the same or adjacent quartiles was found for 85% of the tested food groups. Nevertheless, the agreement was mostly explained by the classification into the adjacent quartile than into the same quartile by two dietary methods. Unfortunately, the previous studies conducted among pregnant women did not provide such details, limiting any comparisons. ª 2014 The British Dietetic Association Ltd.
P. Barbieri et al.
Previous studies aiming to validate questionnaires designed and validated for pregnant women suggest that this method is a valuable tool for assessing nutrient intake by pregnant women (Erkkola et al., 2001; Baer et al., 2005; Mouratidou et al., 2006; Pinto et al., 2010; Barbieri et al., 2013) because comparisons with other methods yield acceptable correlation coefficient values. However, recent studies recommend against using the Pearson’s coefficient in isolation to test for agreement between methods, and argue that it should be used in association with other types of analysis, such as agreement into the same categories, the weighted kappa statistic and the Bland–Altman method (Cade et al., 2002). In the present study, analyses with the Bland–Altman method were not possible because the final sample size was insufficient (Bland & Altman, 1995). The mean quadratic kappa value in the present study was low, in agreement with FFQ validation of food groups in an earlier study of Brazilian students, in which kappa values ranged from 0.15 for pasta to 0.56 for beans (Voci et al., 2008). The FFQ was not effective at estimating intake of fruits and vegetables, as observed in other FFQ validation studies for food groups (Matarazzo et al., 2006). The intake of these foods is highly recommended during pregnancy, and the FFQ may overestimate their consumption, which could increase disagreement between the tests. The most important limitation of the present study was the limited number of 24-h recalls. Given the high within-person variation in diet among the women, a large number of 24-h recall replicates are recommended. Moreover, a higher number of replicates of food recalls in each trimester of pregnancy could provide information about the agreement between the FFQ and food recalls from each trimester individually. Another limitation of the present study was the use of a convenience sample and restriction to women with normal BMI in the prepregnancy period. The study focused on normal weight women during the prepregnancy period as a result of the observed tendency for under-reporting by overweight women, which would have compromised the results of the present study. Another limitation is the fact that the FFQ and 24-h recalls contain similar errors, such as recall bias. The main contribution of the present study was to evaluate the food intake throughout pregnancy, including data with respect to dietary intake from each trimester of pregnancy. Our data do not support using the FFQ to estimate food intake during pregnancy taking the gestational period as a whole. Combined data from short-term methods and the frequency of intake have been suggested recently (Harttig et al., 2011) and should be tested when investigating diet–disease relationships during pregnancy. ª 2014 The British Dietetic Association Ltd.
Validation of a FFQ for pregnant women
Conflict of interests, source of funding and authorship The authors declare that they have no conflict of interests. The project received financial support from the Coordenacß~ao de Aperfeicßoamento de Pessoal de Nıvel Superior (CAPES, master’s grant to PB), Fundacß~ao de Amparo a Pesquisa do Estado de S~ao Paulo [FAPESP, master’s grant to LCC (2011/-03781-8) and RYN (2010/12320-1)] and the Teaching and Research Support Foundation at the Clinical Hospital of the Ribeir~ao Preto Medical School of the University of S~ao Paulo (FAEPA, research grant). DSS and PB were responsible for designing the research project, analysing and interpreting the data, and writing the manuscript. LCC and RYN were responsible for data collection, analysing and interpreting the data, and reviewing the manuscript. All authors approved the final version submitted for publication.
References Associacß~ao Brasileira de Empresa de Pesquisas [Brazilian Economic Classification Criteria]. (2007) Dados com Base no Levantamento Socioecon^omico. Available at: http://www.abep. org (Accessed at January 2012). Baer, H.J., Blum, R.E., Rockett, H.R., Leppert, J., Gardner, J.D., Suitor, C.W. & Colditz, G.A. (2005) Use of a food frequency questionnaire in American Indian and Caucasian pregnant women: a validation study. BMC Public Health 5, 135. Barbieri, P., Nishimura, R.Y., Crivellenti, L.C. & Sartorelli, D.S. (2013) Relative validation of a quantitative FFQ for use in Brazilian pregnant women. Public Health Nutr. 16, 1419– 1426. Beaton, G.H. (1991) Interpretation of results from diet history studies. In The diet history method. ed L. Kohlmeier. Smith-Gordon: London. Bland, J.M. & Altman, D.G. (1995) Comparing methods of measurement: why plotting difference against standard method is misleading. Lancet 345, 1085–1087. Borgen, I., Aamodt, G., Harsem, N., Haugen, M., Meltzer, H.M. & Brantsaeter, A.L. (2012) Maternal sugar consumption and risk of preeclampsia in nulliparous Norwegian women. Eur. J. Clin. Nutr. 66, 920–925. Brantsaeter, A.L., Birgisdottir, B.E., Meltzer, H.M., Kvalem, H.E., Alexander, J., Magnus, P. & Haugen, M. (2012) Maternal seafood consumption and infant birth weight, length and head circumference in the Norwegian Mother and Child Cohort Study. Br. J. Nutr. 107, 436–444. Brantsaeter, A.L., Haugen, M., Alexander, J. & Meltzer, H.M. (2008) Validity of a new food frequency questionnaire for 43
Validation of a FFQ for pregnant women
pregnant women in the Norwegian Mother and Child Cohort Study (MoBa). Matern. Child Nutr. 4, 28–43. Brown, J.E., Buzzard, I.M., Jacobs, D.R. Jr, Hannan, P.J., Kushi, L.H., Barosso, G.M. & Schmid, L.A. (1996) A food frequency questinnaire can detect pregnancy-related changes in diet. J. Am. Diet. Assoc. 96, 262–266. Cade, J., Thompson, R., Burley, V. & Warm, D. (2002) Development, validation and utilization of food-frequency questionnaires – a review. Public Health Nutr. 5, 567–587. Englund-Ogge, L., Brantsaeter, A.L., Haugen, M., Sengpiel, V., Khatibi, A., Myhre, R., Myking, S., Meltzer, H.M., Kacerovsky, M., Nilsen, R.M. & Jacobsson, B. (2012) Association between intake of artificially sweetened and sugar-sweetened beverages and preterm delivery: a large prospective cohort study. Am. J. Clin. Nutr. 96, 552–559. Erkkola, M., Karppinen, M., Javanainen, J., R€as€anen, L., Knip, M. & Virtanen, S.M. (2001) Validity and reproducibility of a food frequency questionnaire for pregnant Finnish women. Am. J. Epidemiol. 154, 466–476. Fisberg, R.F. & Slater, B. (2002) Manual de receitas e medidas caseiras para calculo de inqueritos alimentares: manual elaborado para auxiliar o processamento de inqueritos alimentares. S~ao Paulo: Signus. Harttig, U., Haubrock, J., Knuppel, S. & Boeing, H. & EFCOVAL Consortium (2011). The MSM program: web-based statistics package for estimating usual dietary intake using the Multiple Source Method. Eur. J. Clin. Nutr., 65: 87–91. Institute of Medicine (IOM). (2009) Weight Gain During Pregnancy: Reexamining the Guidelines. Washington, DC: National Academy Press. Isobe, M.T., Bertola, M.R., Zucolotto, D.C.C. & Sartorelli, D.S. (2013) The influence of the educational status on the reproducibility of a quantitative food frequency questionnaire for pregnant women. Rev. Bras. Saude Mater Infant 13, 23–28. Johnson, R.K., Soultanakis, R.P. & Matthews, D.E. (1998) Literacy and body fatness are associated with underreporting of energy intake in US low income women using the multiple-pass 24 hours recall: a doubly labeled water study. J. Am. Diet. Assoc. 98, 1136–1140. Marhazlina, L.S.L., Azwany, N. & Hamid Jan, J.M. (2011) Development, validity and reproducibility of a food frequency questionnaire in pregnancy for the
44
P. Barbieri et al.
Universiti Sains Malaysia Birth Cohort Study. Malays. J. Nutr. 1, 1–18. Matarazzo, H.C.Z., Marchioni, D.M.L., Figueiredo, R.A.O., Slater, B., Eluf Neto, J. & W€ unsch Filho, V. (2006) Reprodutibilidade e validade do questionario de freq€ u^encia de consumo alimentar utilizado em estudo caso-controle de c^ancer oral. Rev. Bras. Epidemiol. 9, 316–324. Meltzer, H.M., Brantsaeter, A.L., Ydersbond, T.A., Alexander, J. & Haugen, M. (2008) Methodological challenges when monitoring the diet of ant women in a large study: experiences from the Norwegian Mother and Sy (MoBa). Matern. Child Nutr. 4, 14–27. Mouratidou, T., Ford, F. & Fraser, R.B. (2006) Validation of a food-frequency questionnaire for use in pregnancy. Public Health Nutr. 9, 515–522. Nyambose, J., Koshi, K.G. & Tucker, K.L. (2002) High intra/ interindividual variance ratios for energy intakes of pregnant women in rural Malawi show that many days are required to estimate usual intake. J. Nutr. 132, 1313–1318. Oliveira, T., Marquitti, F.D., Carvalhaes, M.A. & Sartorelli, D.S. (2010) Desenvolvimento de um Questionario Quantitativo de Freq€ u^encia Alimentar (QQFA) para gestantes usuarias de unidades basicas de sa ude de Ribeir~ao Preto, S~ao Paulo, Brasil. Cad. Sa ude P ublica 26, 2296–2306. Pinheiro, A.B.V., Lacerda, E.M.A., Benzecry, E.H., Gomes, M.C.S. & Costa, V.M. (2005) Tabela para avaliacß~ao de consumo alimentar em medidas caseiras, 5th edn. S~ao Paulo: Editora Atheneu. Pinto, E., Severo, M., Correia, S., Silva, I.S., Lopes, C. & Barros, H. (2010) Validity and reproducibility of a semi-quantitative food frequency questionnaire for use among portuguese pregnant women. Matern. Child Nutr. 6, 105–119. Scientific Advisory Committee on Nutrition. (2011) The Influence of Maternal, Fetal and Child Nutrition on the Development of Chronic Disease in Later Life. Available at: http://www.sacn.gov.uk/pdfs/sacn_early_nutrition_final_ report_20_6_11.pdf. (Accessed on 5 November 2012). Voci, S.M., Enes, C.C. & Slater, B. (2008) Validacß~ao do Questionario de Freq€ u^encia Alimentar para Adolescentes (QFAA) por grupos de alimentos em uma populacß~ao de escolares. Rev. Bras. Epidemiol. 11, 561–572. Willett, W.C. (1998) Nutritional Epidemiology, 2nd edn. New York, NY: Oxford University Press.
ª 2014 The British Dietetic Association Ltd.
