N UTR IT ION RE S EA RCH XX ( 2 0 14 ) XXX–X XX

Available online at www.sciencedirect.com

ScienceDirect www.nrjournal.com

Validity and reproducibility of an interviewer-administered food frequency questionnaire in Austrian adults at risk of or with overt diabetes mellitus☆ Michaela Farukuoye a , Klaus Strassburger b , Gertrud Kacerovsky-Bielesz a , Guido Giani b , Michael Roden a, c, d,⁎ st

a

Karl-Landsteiner Institute for Endocrinology and Metabolism at 1 Medical Department, Hanusch Hospital, Vienna, A-1140, Austria Institute of Biometrics and Epidemiology, German Diabetes Center, Leibniz Center for Diabetes Research at Heinrich-Heine University, Düsseldorf, D-40225, Germany c Institute for Clinical Diabetology, German Diabetes Center, Leibniz Center for Diabetes Research, Leibniz Center for Diabetes Research at Heinrich-Heine University, Düsseldorf, D-40225, Germany d Department of Endocrinology & Diabetology, Heinrich Heine University, Düsseldorf, D-40225, Germany b

ARTI CLE I NFO

A BS TRACT

Article history:

Food frequency questionnaires (FFQs) provide an inexpensive tool for dietary assessment.

Received 13 November 2013

Given the scarcity of data on their validity for nutritional analysis in persons with overt

Revised 6 April 2014

diabetes mellitus or with increased risk of diabetes (relatives of patients with diabetes), this

Accepted 8 April 2014

study tests the hypothesis that an FFQ, adapted to local dietary habits, yields a reliable estimate of nutrient intake when compared with 7-day food record (7DR) in healthy,

Keywords:

prediabetes, and diabetes cohorts. One hundred three volunteers (50 persons with overt

Nutrition assessment

diabetes mellitus, 24 relatives of patients with diabetes, and 29 nondiabetic individuals

Food frequency questionnaire (FFQ)

without a family history of diabetes) completed both FFQ and 7DR. A second FFQ was

Food record

completed by 100 of these volunteers after 3 months to evaluate its reproducibility. Data were

Validation

compared by correlation and Bland-Altman analyses. Across the entire group, estimates for

Diabetes mellitus

gram intakes of nutrients and total energy were associated with wide limits of agreement between FFQ and 7DR (correlation coefficients, 0.23-0.72; P < .02). Compared with 7DR, the FFQ overestimated intakes of saturated fat in the entire group (+6.6 ± 14 g; P < .001) and in persons with overt diabetes mellitus (+7.6 ± 15 g; P < .001) but underestimated protein intake in relatives of patients with diabetes (−16.36 ± 31 g; P = .01). The repeated FFQ revealed variable agreement (correlation coefficients, 0.34-0.72; P < .001) and underestimated (P < .01) macronutrient and

Abbreviations: FFQ, food frequency questionnaire; 7DR, 7-day food record; CON, nondiabetic individuals without family history of diabetes; T2D, type 2 diabetes mellitus; T1D, type 1 diabetes mellitus; BLS II, German Nutrient Data Base; BMI, body mass index; WHR, waist-to-hip ratio; %EI, percentage of daily energy intake. ☆ This work was supported in part by the European Foundation for the Study of Diabetes (Düsseldorf, Germany), Austrian National Bank (OENB11459) (Vienna, Austria), and the Austrian Diabetes Association (Vienna, Austria). The work of MR is supported by the Ministry of Science and Research of the State of North Rhine–Westphalia (Düsseldorf, Germany) and the German Federal Ministry of Health (Berlin, Germany) as well as in part by a grant of the German Federal Ministry for Research (Berlin, Germany) to the Centers for Diabetes Research and by a grant of the Helmholtz Alliance Imaging and Curing Environmental Metabolic Diseases. ⁎ Corresponding author. Institute for Clinical Diabetology, German Diabetes Center, Department of Endocrinology & Diabetology, Heinrich-Heine University, c/o Auf’m Hennekamp 65, D-40225 Düsseldorf, Germany. Tel.: +49 211 3382 201. E-mail address: [email protected] (M. Roden). http://dx.doi.org/10.1016/j.nutres.2014.04.004 0271-5317/© 2014 Elsevier Inc. All rights reserved.

Please cite this article as: Farukuoye M, et al, Validity and reproducibility of an interviewer-administered food frequency questionnaire in Austrian adults at risk of or with ove..., Nutr Res (2014), http://dx.doi.org/10.1016/j.nutres.2014.04.004

2

N UTR IT ION RE S EA RCH XX ( 2 0 14 ) XXX–X XX

total energy intakes, with slightly better performance in persons with overt diabetes mellitus and relatives of patients with diabetes than in nondiabetic individuals without a family history of diabetes. Hence, the FFQ allows measuring intakes of total energy and macronutrients in prediabetes and diabetes cohorts but reveals limitations when assessing dietary composition. © 2014 Elsevier Inc. All rights reserved.

1.

Introduction

Lifestyle factors are tightly related to the pathogenesis of diabetes mellitus, and lifestyle modification is the basis of treating type 2 diabetes mellitus (T2D) and related disorders such as obesity, hypertension, and coronary artery disease [1]. Interventions based on dietary modification and increased physical activity also reduce the risk of T2D by 58% in glucose intolerant persons, who are at increased risk of T2D [2,3]. The habitual diet also affects the development of T2D [4-6]. Not only the total amount of energy intake but also the percentage share of fat and protein has been linked to insulin resistance, which generally precedes T2D [7]. Analysis of dietary behavior is therefore essential not only for evaluating the impact of food composition as risk factors but also for developing and monitoring effective prevention and intervention strategies in large human trials related to T2D [8]. Food frequency questionnaires (FFQs) capture the habitual dietary intake [9] and are widely used in large epidemiologic studies [10-12]. As FFQs are low cost and easy-to-use tools, they could be also applied to monitor nutritional habits during the course of clinical intervention trials. For this purpose, the ideal FFQ should accurately reflect the individual’s food consumption over a defined period, independent of behavioral patterns or inaccurate memory [13]. Although FFQs have been validated against other methods such as weighed records or diet records in healthy humans [11], such validation is missing for assessing the complete dietary intake in individuals at risk of or with overt diabetes mellitus. We hypothesized that an interviewer-administered FFQ provided a reliable estimate of nutrient intake in healthy, prediabetes, and diabetes cohorts. This FFQ has been previously used along with 24-hour recalls for dietary assessment in lifestyle intervention trials [14,15]. For the present study, we adapted this FFQ for local dietary habits and determined its variability by comparison with a 7-day food record (7DR) and the reproducibility by comparison with a second food frequency questionnaire (FFQ2) after a 3-month interval. The study design aimed to examine the hypothesis in outpatients attending services for prediabetes and/or diabetes in a defined metropolitan area of Vienna, Austria.

2.

Methods and materials

2.1.

Participants

The volunteers were consecutively recruited from the Diabetes Outpatient Service of the 1st Medical Department of Hanusch Hospital, Teaching Hospital of Medical University of Vienna, and from a local Physiotherapy Service, Vienna, Austria, between January 2008 and August 2008. Persons aged

19 years or more were included when they agreed to complete both FFQ and 7DR, to be available within the area throughout the study, able to understand and willing to participate, and sign the consent forms. Persons reporting relevant changes in nutritional habits within the last year, recently diagnosed and treated diabetes [16]), and individuals with diseases affecting memory were excluded. All participants gave written informed consent after being explained the nature of the studies, which complied with the Declaration of Helsinki and were approved by the local ethics board. For the variability study, the population consisted of 123 volunteers, including 27 nondiabetic relatives of patients with T2D who had at least one parent or grandparent with overt T2D, 66 patients with diabetes (32 patients with T2D and 34 with type 1 diabetes [T1D]), and 30 nondiabetic healthy individuals serving as control (CON). Participants with missing 7DR (n = 18) or incomplete 7DR (n = 2) were excluded from further evaluation. The attrition rates were 16% in the entire group, 3% in CON, 11% in relatives of patients with diabetes, and 24% in persons with overt diabetes mellitus. For the reproducibility study, the population comprised 100 volunteers, including 23 relatives of patients with diabetes, 49 individuals with overt diabetes mellitus (26 T1D and 23 T2D), and 28 CON, after the exclusion of 23 cases due to missing FFQ2. The attrition rates were 19% in the entire group, 7% in CON, 15% in relatives of patients with diabetes, and 26% in individuals with overt diabetes mellitus. No participant of these studies attended the intervention trials, where this FFQ had been used [14,15].

2.2.

Study design and methods

An FFQ containing 128 items [17] was modified into an interviewer-administered 107-item open-ended questionnaire adapted to local dietary habits. The FFQ measures the participant’s usual food intake during a defined preceding period. For the 7DR, the participants recorded all food items and beverages according to a predefined protocol for 1 week. Each day, the participants had to fill in a structured one-page form and list all food items and beverages ingested at each meal and between meal times. The form did not provide food suggestions. Portion sizes were determined based on common household measuring units such as cups, bowls, spoons, and slices. Food frequency questionnaire and 7DR were analyzed using the German Nutrient Data Base (BLS II) [18]. One singletrained dietician performed both interviews. The calculation of nutrient and fluid intakes determined via FFQ was carried out as follows: every single item on the food listing was extrapolated by multiplying the portion sizes by the frequency of food consumption to obtain data in units per year and then entered into the BLS to yield mean values of energy intake (in kilo joules per day), carbohydrates, fat, protein, and saturated fat (grams per day and % of daily energy intake [%EI]), n-3 and n-6

Please cite this article as: Farukuoye M, et al, Validity and reproducibility of an interviewer-administered food frequency questionnaire in Austrian adults at risk of or with ove..., Nutr Res (2014), http://dx.doi.org/10.1016/j.nutres.2014.04.004

3

N UTR IT ION RE S EA RCH XX ( 2 0 14 ) XXX–X XX

Table 1 – Characteristics of participants in the comparison of FFQ and 7DR Entire group N (women/men) Age (y) Body weight (kg) BMI (kg/m2) Waist circumference (cm) Hip circumference (cm) WHR

103 (59/44) 43 ± 14 77 ± 17 26 ± 6 94 ± 17 106 ± 10 0.88 ± 0.10

Subgroups REL

DIAB

CON

24 (16/8) 36 ± 12 a 72 ± 15 a 25 ± 4 a 87 ± 14 a 104 ± 8 a 0.84 ± 0.10 a

50 (24/26) 50 ± 13 84 ± 17 29 ± 6 101 ± 17 110 ± 10 0.92 ± 0.11

29 (19/10) 37 ± 12 b 68 ± 13 b 23 ± 3 b 86 ± 11 b 101 ± 8 b 0.85 ± 0.08 b

Data are means ± SD. Abbreviations: N, number of subjects; DIAB, persons with overt diabetes mellitus; REL, relatives of patients with diabetes. a P ≤ .01; P value for difference between DIAB and CON derived from Wilcoxon-Mann-Whitney U test. b P ≤ .01; P value for difference between DIAB and REL derived from Wilcoxon-Mann-Whitney U test.

fatty acids (%EI), cholesterol (grams per day), fiber, and alcohol (grams per day). The calculation of nutrient and fluid intakes via 7DR was done as follows: all recorded food and beverage items were entered into the BLS to obtain mean values of energy intake and nutrient intake as described above. On the first day, the interviewer recorded the participants’ age (in years), medical history, anthropometric data such as body weight (in kilograms), height (in centimeters), waist and hip circumference (in centimeters), and habitual physical activity using the questionnaire of Baecke et al [19]. Body mass index (BMI) (calculated as body weight divided by the square of height and expressed as kilograms per square meter) and waist-to-hip ratio (WHR) (given as waist circumference divided by hip circumference) were assessed. Subsequently, all volunteers completed the FFQ in an interviewer-administered fashion. They were given instructions on how to perform the 7DR and were asked to maintain their usual lifestyle and return the 7DR by mail after 1 week in a prestamped addressed envelope. Assessment of body weight, waist and hip circumference, habitual physical activity, BMI, WHR, and the performance of FFQ were repeated after 3 months to ensure that participants would not recall their previous answers and to minimize effects of lifestyle changes [11]. Finally, they received their nutrition analysis and individual dietary counseling according to stateof-the art evidence-based recommendations [16].

2.3.

Statistical analyses

Normal distribution of the between-method differences was tested using the Kolmogorov-Smirnov Goodness-of-Fit Test based on the data of the entire group. Means and SD as well as Spearman correlation coefficients (r) were calculated. Between-method differences were analyzed by the Wilcoxon signed rank test. P ≤ .05 were considered to indicate statistically significant differences. Bland-Altman analyses [20] were carried out to produce quantitative estimates of the agreement between the 2 methods [11]. Limits of agreement (mean ± 1.96 SD/empirically 2.5% and 97.5% quantiles of the between-method difference) were given for normally/nonnormally distributed data for assessing validity. For assessing reproducibility, correlations were calculated as used in 90% of previous studies [11], and Bland-Altman analyses were carried out in addition. Limits of repeatability (mean ± 1.96 SD [20]/

median [empirically 2.5% and 97.5% quantiles of the betweenmethod difference]) were given for normally/nonnormally distributed data. The dependence of difference and mean was assessed via Spearman correlation test. The sample size of the present analysis is within the range of previous studies in this field [11]. All analyses were performed for the entire group and all subgroups using SAS for Windows version 9.2 (SAS Institute, Cary, NC, USA).

3.

Results

Table 1 summarizes the main characteristics of the study participants. Both CON and relatives of patients with diabetes were comparable (smallest P for difference > .23). Individuals with overt diabetes mellitus had approximately 37% higher age, 17% higher BMI, and 9% higher WHR than the other groups. This was due to 33% higher age, 28% higher BMI, and 15% higher WHR in T2D than in T1D subgroups.

3.1.

Validity of the FFQ compared with the 7DR

Across the entire group, between-method differences in the intake of the various nutrients were normally distributed except for total fat (%EI), n-3 fatty acids (%EI), and alcohol (grams per day) (Table 2). Results for total energy intake (kilojoules per day) correlated (r = 0.58) between FFQ and 7DR with no noticeable bias patterns between the 2 measurements (Fig. 1A). Furthermore, correlation coefficients were >0.50 for carbohydrate (%EI and grams per day), protein (grams per day), alcohol (grams per day), and potassium (milligrams per day). The difference and the mean of both methods were interdependent for alcohol (grams per day) and potassium (grams per day) intakes. Low correlation coefficients (r = 0.400.48) were found for comparing FFQ and 7DR with regard to protein (%EI), total and saturated fat (grams per day), cholesterol (grams per day), and fiber (grams per day). However, there was a systematic bias of FFQ results with 6% lower protein (%EI), 21% higher saturated fat (grams per day), and 9% lower cholesterol (grams per day) intakes. Estimates of total fat, n-6 fatty acids, saturated fat, and n-3 fatty acids (all in %EI) intakes exhibited a weak (r = 0.23-0.31) but significant

Please cite this article as: Farukuoye M, et al, Validity and reproducibility of an interviewer-administered food frequency questionnaire in Austrian adults at risk of or with ove..., Nutr Res (2014), http://dx.doi.org/10.1016/j.nutres.2014.04.004

4

N UTR IT ION RE S EA RCH XX ( 2 0 14 ) XXX–X XX

Table 2 – Daily energy and nutrient intakes of the entire group FFQ

7DR

Energy intake (kJ/day) 8587 ± 2781 8501 ± 2500 Carbohydrate (%EI) 44 ± 7 44 ± 6 Carbohydrate (g/day) 225 ± 81 222 ± 75 Fat (%EI) 36 ± 6 35 ± 5 Fat (g/day) 83 ± 32 80 ± 28 Protein (%EI) 16 ± 3 17 ± 3 Protein (g/day) 81 ± 29 84 ± 25 Saturated fat (%EI) 17 ± 4 15 ± 3 Saturated fat (g/day) 40 ± 16 33 ± 11 n-3 Fatty acids (%EI) 0.77 ± 0.23 0.84 ± 0.40 n-6 Fatty acids (%EI) 4.23 ± 1.42 5.04 ± 1.52 Cholesterol (g/day) 0.32 ± 0.15 0.35 ± 0.13 Fiber (g/day) 24 ± 11 24 ± 8 Alcohol (g/day) 7.48 ± 11 8.29 ± 12 Calciferol (μg/day) 2.90 ± 1.46 3.31 ± 3.14 Magnesium (mg/day) 406 ± 136 388 ± 121 Potassium (mg/day) 3374 ± 1183 3202 ± 812

Corr. coefficient

P

0.58 0.52 0.58 0.27 0.48 0.43 0.53 0.31 0.40 0.23 0.35 0.45 0.40 0.72 0.26 0.41 0.50

Validity and reproducibility of an interviewer-administered food frequency questionnaire in Austrian adults at risk of or with overt diabetes mellitus.

Food frequency questionnaires (FFQs) provide an inexpensive tool for dietary assessment. Given the scarcity of data on their validity for nutritional ...
769KB Sizes 2 Downloads 3 Views