Development and Validation of a Food Frequency Questionnaire for Use in Epidemiological Studies Among North Karnataka Population

Background: Dietary records, such as food diaries (FD) and food frequency questionnaires (FFQs), are widely employed in large-scale epidemiological analyses. Given the regional diversity in dietary habits, this study aimed to develop and assess the reproducibility and construct validity of a novel FFQ tailored for northern Karnataka (FFQ-NK). Methodology: A detailed food list was created based on food use and market surveys in the region. Utilizing data from 24-hour diet recalls and a supplemental focus group discussion, a 116-item semi-quantitative FFQ was developed for adults in north Karnataka. Setting and Subjects: Involving 100 participants from north Karnataka, the FFQ was interviewer administered and the participants also completed three 24-hour dietary recalls (DR), serving as a reference for validity assessment. Results: Though the FFQ indicated higher food and nutrient intake compared to food records. Significant correlations were found for nutrient intake. Pearson’s correlation coefficients between FFQs and DRs ranged from 0.717-0.965 (isoflavonols to energy). Conclusions : The developed and validated FFQ is a valuable tool for epidemiological studies requiring nutrient intake estimates in the north Karnataka population. It serves as an effective dietary assessment tool for individuals aged 20 and older in large-scale epidemiological studies.


INTRODUCTION
4] Additionally, the FFQ facilitates the categorization of subjects into low, medium, and high consumption tertiles, streamlining the exploration of associations with disease prevalence and mortality data in specific populations. 5[11][12] Though the FFQ are not without limitations, they are popular as their feasibility is the major advantage for use in establishing long-term habitual dietary intake. 13Given the diverse dietary habits in the Indian population, it is crucial to develop and validate Food Frequency Questionnaires (FFQs) tailored to specific regions and research goals.5][16] while biomarkers offer independence from biases, they lack substitution for common dietary components and dietary recommendations.8][19] Despite their widespread use, the less accurate nature of FFQs necessitates data verification against reference techniques, like food diaries, to mitigate biases arising from portion size estimates and memory limitations.The Indian subcontinent is known for its diverse eating habits and the vast variety of foods consumed.[22] Dietary instruments like the FFQ play a vital role in epidemiological studies investigating the definitive role of nutrition on lifestyle diseases and formulating health policies.This study addresses the need for a validated FFQ tailored to diverse socio-demographic populations in the Indian subcontinent, particularly north Karnataka.The developed FFQ enables investigations into links between dietary patterns and health outcomes in the north Karnataka population, marking the first of its kind for the region, validated against a three-day 24-hour dietary recall. 5e study was motivated by the clear need to develop and validate an FFQ that is suitable to assess die-tary pattern across heterogeneous sociodemographic populations with varying incomes and categories in the regions of Indian subcontinent and of north Karnataka in particular.The study also aimed to explore differences in habitual dietary intakes that were attributed by age, gender, education, social class and body mass index (BMI).

METHODOLOGY
Population setting and cohort recruitment: The FFQ-NK was developed within the diverse population of Vijayapura district in north Karnataka, India, representing both rural and urban demographics.Vijayapura, a prototypical region in north Karnataka, exhibits a dietary pattern cantered around staples such as jowar and rice, complemented by a substantial intake of fresh vegetables and greens in the form of curries and spices.Ethical approval for the study was obtained from the Institutional Ethics Committee of BLDE (du) Shri B M Patil Medical College, Hospital, and Research Centre (Ref: BLDE(DU)/IEC/460/2020-21). Participants were provided with detailed information about the study through an information sheet and provided informed consent.The inclusion criteria encompassed healthy individuals above the age of 18, while those on extreme diets (anyone consuming more calories than the specified range (4200 kcal for men and 3500 kcal for women)) or fasting regimens were excluded.A total of 130 participants were recruited by a purposive sampling technique in which the participants were selected as per the maximum variation sampling technique.On the basis of the participant informant sheet and after exclusion on the predefined criteria the participants were selected based on urban/rural settings, gender, age group, physical activity and BMI.Additionally, their occupational categories (service and homemakers) and annual income of the household and their educational qualification were also considered but with an intention to ensure that the individuals belonged near equitable representation across the various factors which can affect their dietary choices and patterns.

Development of FFQ-NK:
The formulation of the FFQ-NK was meticulously conducted by leveraging information from the National Institute of Nutrition, Hyderabad, India's database, which includes nutritive values of commonly consumed food items. 23 Prior to developing FFQ-NK, extensive literature review, diet surveys, and FFQ development studies were examined.The preliminary stage involved formative fieldwork, which was qualitatively followed by a sequenced mixed-methods approach. 26od List Construction and Nutrient Alignment: The development of the FFQ-NK's food list involved a comprehensive approach, utilizing two primary methods: 1. Identification of Food Items: The food list was compiled through a dual approach: firstly, by identifying commonly consumed local, regional, and global foods using 24-Hour Diet Recalls of 50 representative participants aged 18-60 years.Secondly, written free lists were obtained from a purposive sample of adults, encompassing foods available in local stores and from vendors. 27This combination allowed for a diverse and representative compilation of food items.

Categorization and Review:
The resulting food list was categorized into fourteen groups based on their ingredients and preparation methods.This categorization facilitated a systematic understanding of dietary patterns within the study population.

Database Review for Target Nutrients:
To ensure the relevance of the food list to the study's nutritional goals, each listed food item underwent a meticulous review against a comprehensive nutrient database. 23This step aimed to align the food list with the intake of 32 nutrients, including energy, protein, fat, carbohydrate, fibre, various fatty acids (saturated, monounsaturated, polyunsaturated), essential minerals (iron, zinc, magnesium), and a spectrum of vitamins (B1, B2, B6, B12, C, A, D, E), among others.

Validation of Nutrient Profiles:
A rigorous validation process affirmed that the food list included items accounting for over 95% of the nutrient profiles used for the epidemiological studies. 28his validation was conducted through sample analysis, ensuring that the selected food items sufficiently represented the nutritional diversity of the study population.
Description and the Frequency response section of FFQ-NK: The FFQ-NK was developed with a 116item scale that measured food consumption over a year (96 composite food items (i.e., food items containing one or more ingredients) and 20 simple food items).The food list was built and categorized into fourteen food groups based on their ingredients and preparation methods: i.The semi-quantitative FFQ-NK was intervieweradministered to be more precise and had seven frequency categories for food consumption over 1 year, from 2-3 times per day to rarely/never as responses.
The participants were also tutored on the normal portion size typically eaten by them.As most of the common fruits are currently available throughout the year, the data capture was done as a single response and was noted as small, medium, and large.
The volume, diameter (2.4 to 3.7 cm) of the most commonly consumed size, and number of pieces were standardized for weight equivalence.They were also assisted in estimating their portion size as a multiple of the standard measure by using metal vessels and spoons, as described in earlier studies. 23ministration of FFQ: The FFQ-NK was administered to a representative sample of 130 individuals aged 20-70 years.For validation, 3-day, 24-hour diet recall interviews were employed as the reference method due to noted lower overall variance. 29rained nutritionists conducted open-ended, prompted interviews over three non-consecutive days, ensuring avoidance of days with significant dietary variations.Respondents provided chronological reports of all food encounters throughout the day.
A follow-up response rate of 100 participants was achieved when study's exclusion criteria were applied for data completeness.The resultant tabulation involved multiplying frequency of intakes by portion size and corresponding nutrient profiles reported by participants.
Statistical Analysis: All statistical analyses were conducted using GraphPad Prism (version 9.2, India) for descriptive statistics, frequency distributions, and graphical representations.For advanced statistical modelling and analyses, JMP 16.2 was utilized.The responses from FFQ were subjected to comprehensive analysis, encompassing mean, median, and standard deviation calculations for various nutrients.Descriptive statistics were employed to elucidate demographic details and frequency distribution, providing a foundational understanding of the acquired responses.Prior to analysis, normal distribution tests were conducted, revealing that most variables exhibited normal distribution.However, VIT E, Flavan-3-ol, Flavones, and Flavanones were logtransformed to enhance comparability.To assess the significance across different parameters, regression analysis was undertaken, focusing on gender, domicile, and per capita income.Associations between nutrient intakes from FFQ and the arithmetic mean of the 3-day DR were evaluated using Pearsons' correlations.One-way ANOVA tested for overall significance, followed by paired t-tests for specific comparisons.Simple linear regression was applied to assess the linear agreement between nutrient scores obtained from FFQ and the arithmetic average from 3day DRs.The agreement between the two methods was further evaluated using the Bland-Altman method, offering insights into potential bias, reproducibility and limits of agreement

RESULTS
Demographic characteristics (Table 1): BMI, basal metabolic rate; the study cohort demonstrated a balanced gender representation, with males constituting 51% and females 49%.The majority of partici-pants fell within the 30-40 years age group and urban residents (56) accounted for the majority of the study population.A substantial portion of participants exhibited a normal BMI (39%), while 47% were classified as overweight.Notably, the majority of individuals (68%) reported engaging in moderate physical activity.

Summary of the nutrient intakes:
The summarized findings, presented in Table 2, provide a detailed overview of the *statistically significant variations in nutrient intakes across gender, domicile, and per capita income groups.Significant variations in nutrient intake were identified between males and females.Total energy, net carbohydrate, total fat, MUFA, and folate were notably higher in males and the differences were statistically significant and also males reported increased mineral intake, particularly iron and calcium.Distinct differences in nutrient intake were observed between urban and rural dwellers.Notably, disparities in the intake of specific nutrients, including vitamin D, B12, folate and iron, were identified, highlighting the potential influence of domicile on dietary patterns.Participants were stratified based on per capita income levels using Prasad's classification.While the overall mean intake of energy, proteins, vitamin D, vitamin C, and minerals such as iron and calcium appeared lower in the lower economic class (EC) compared to middle and high-economic class (EC), these differences were statistically insignificant, except for vitamin C (P value: 0.04), indicating a potential association between income levels and vitamin C intake.
Validity of the Food frequency questionnaire: The validity of the FFQ was evaluated through Pearson correlations between responses from the FFQ and the mean of three 24-hour DR.The Pearson correlation coefficients were computed for each nutrient, providing a quantitative measure of the strength and direction of the linear relationship between the FFQ and the 3-day 24-hour DR.        ).The Bland-Altman plot for carbohydrate intake reveals a mean difference of 12.13, with a 95% CI spanning from 6.64 to 17.62 The systematic trend indicates that, on average, the FFQ tends to provide higher estimates for protein, carbohydrate, and total energy compared to the 3-day dietary records.
Bland-Altman Analysis: FFQ vs. 3-Day DR: To assess the agreement between reported nutrient intakes from the Food Frequency Questionnaire (FFQ) and the 3-Day DR, a Bland-Altman analysis was conducted.This method provides insights into the extent of agreement, potential biases, and variability between the two dietary assessment tools.Differences in nutrient intake between the FFQ and the 3-day FR were plotted against the mean nutrient intakes of the two methods.Positive differences in the average dis-crepancy indicated an overestimation of nutrient intake by the FFQ.A larger value of bias, reflected by wider limits of agreement (as indicated by the 95% Confidence Interval), suggested a greater extent of overestimation by the FFQ.For macronutrients such as carbohydrates and fats, as well as a range of micronutrients, there was a bias towards positive differences.This implies that the FFQ tended to provide higher estimates of certain macro-and micronutrients compared to the 3-day food diary.Bland-Altman plots indicated a systematic trend towards higher estimates with the FFQ for certain nutrients compared to the food diary records.This suggests a consistent pattern of overestimation by the FFQ.The summarized results of the Bland-Altman analysis, including bias, limits of agreement, and other relevant metrics, are tabulated in Table 5.

DISCUSSION
The authors have made a sincere effort to develop an FFQ to capture the dietary pattern, which can be representative of the one widely prevalent among the population of north Karnataka (FFQ-NK), also termed inland north Karnataka (Belgaum, Bagalkot, Bijapur, Gulbarga, Bidar, Gadag, Dharwad and Bellary).The pattern can also be extrapolated to the population of nearby regions like inland central Maharashtra (Nanded, Hingoli, Parbhani, Jalna, Aurangabad, Beed, Latur) and inland eastern Maharashtra (Buldana, Akola, Amravati, Wardha and Nagpur), where sorghum is still important in the consumption basket and a staple along with rice and wheat.A conscientious attempt was made to include a comprehensive list of the food items consumed among the sample population, which could be representative for group comparisons as they tend to capture the habitual intake of the population in north Karnataka over a given period.6 The above-validated FFQ-NK shall serve as an important tool to register dietary patterns and average energy and nutrient intake in the region of north Karnataka.It can further serve to design studies that investigate the association between the dietary pattern and common diseases of interest prevalent in this region.Registering energy and nutrient intake is key to obtaining more reliable evidence for diethealth relationships from nutritional cohort studies.
In the present study, 100 participants completed all of the questionnaires (i.e., 1 FFQ and three 24-hr dietary recalls).It was validated against a 24-hr diet record as a reference tool, as no truly gold standard exists. 29In comparison with the 24-hour diet record, the FFQ-NK tends to overestimate nutrients unadjusted, as in previous studies . 9,22 ioeconomic factors like the income of the family also had stronger associations, particularly with the average protein intake and nutrients like vitamin D, vitamin C, calcium, and zinc, than the total energy intake.Age and sex were the other determinants that significantly affected the intake of certain nutrients.
The present study also found that the high-and middle-income sectors had higher intakes of average total energy and most of the micronutrients in comparison to those in the lower income sector.Future researchers should be aware of the likely cumulative impact of the suggested differences and relationships.The majority of research conducted in India has hitherto generally explored pricing and expenditure as significant determinants of food consumption patterns, with socio-economic and regional factors receiving less attention and as possible determinants of variations in food consumption patterns.Higher income groups are generally linked to healthier eating habits, which include consuming a variety of fruits and vegetables and more sources of proteins like dairy, poultry, meat, and oil.[35] The average of the 3-day DR was considered to correlate with the FFQ derived nutrient values for validation.The correlation coefficient (energy adjusted) that was discerned between the FFQ and 3-day DR (0.45 to 0.93) was comparable to those reported by similar validation studies conducted in Trivandrum, South Kerala (0.34 to 0.72) 35 and also by studies set in rural villages in Ernakulum district, Kerala (ranging from 0.32 to 0.61) 5 and in Gujarat (ranging from 0.55 to 1.00) 9 Additionally, some other studies have also demonstrated a range of coefficients that appeared to be similar to our range. 26,34,36The findings of this study indicate that macronutrients like total fats and proteins have the largest association, whereas trace elements like iron, selenium, and calcium and vitamins like vitamin C have lesser correlations. 37,38and the flavonols (0.59) and flavanones (0.54) also had lesser correlations.As previously established, these nutrients may tend to have weaker correlation coefficients and significant within-person variability in validation studies since they are not concentrated in most diets. 5,37rtain macronutrients, such as carbohydrates and proteins, and others, such as folate, total carotenes, zinc, iron, selenium, and magnesium, were overestimated by the FFQ compared to the 24-hour dietary record.Also, vitamins like A, D, E, C, B1, and B6 were overestimated by the FFQ, whereas vitamins B2 and B12, along with copper and zinc, were underestimated by the FFQ compared to the food record.It is noted that the independent sociodemographic factors may account for 32% of the variation in the difference between the two assessment techniques.For the majority of them, gender was a significant explanatory factor, with men over reporting intakes in comparison to women for total energy, proteins, fibre, PUFA,vitamin B12, and zinc, whereas other researchers noted a reportedly increased intake by women for most of the nutrients. 39Dietary fibre intakes have been linked to lower family incomes in previous studies 40 that looked at the relationship be-tween income and dietary fibre intake.Similarly, our findings discerned an increased intake of dietary fibre by the middle-and higher-income groups and those belonging to the rural setting.Apart from being the first to develop a FFQ for use in the north Karnataka population, our study has some significant strengths.First, a wide range of tests were used to assess the validity of the FFQ, correlation coefficients in conjunction with the Bland-Altman approach.
When assessing the validity and reproducibility of a food frequency questionnaire, the Bland-Altman method has been found to be more effective than correlation analysis.Furthermore, the study's sample size was adequate to estimate the limits of agreement for the Bland-Altman analysis, which is a prerequisite for evaluating the validity of the FFQ. 29e FFQ was administered by the interviewers, who were registered and experienced nutritionists practicing at the institute, and we believe that this is the study's strength preventing underreporting and enhancing consistency.The participants were also trained to understand portion sizes.The guidance on portion size will be implemented as a pre-requisite for all future administrations of the current validated FFQ.A 116-item FFQ could have been lengthy, which may have led to participant enervation and also nutrient overestimation at times.Some of the recognized sources of error we would like the researchers to be wary of are predetermined and set food lists, intra-individual variations due to memory and interview fatigue, and also errors in interpreting the questions and portions of the food items listed. 41ditionally, though DR and 24-hour food diaries are often used as the most accepted reference tools for validating the FFQs, they are not known to be without limitations due to reporting errors attributable to gender and differences in BMI of the individuals. 36mission of validated biomarkers for nutrient assessment represents a notable limitation in our study, impacting the comprehensive validation of nutrient intake data.The absence of direct biomarker measurement may affect the accuracy and reliability of nutritional assessments, warranting cautious interpretation of the results.Future researchers can strengthen the findings of the current study by including biomarkers specific to the nutrients for cross-validation studies.In summary, the creation and validation of the present Food Frequency Questionnaire (FFQ) constitute a critical first measure that enables us to employ it as a key instrument in epidemiological research to evaluate food intakes, dietary patterns, and associations with disease attributes in the people of North Karnataka, India.It would be wise to evaluate the prospects of the application of this FFQ in the regions of north Karnataka where the populace has similar dietary habits.

CONCLUSION
In summary, the authors have successfully developed and validated the first Food Frequency Question-naire (FFQ-NK) for the population of north Karnataka, India, providing a valuable tool for assessing dietary patterns and nutrient intake in the region.The FFQ-NK, comprising 116 items, demonstrates associations with socioeconomic factors, age, and sex, highlighting disparities in nutrient intake among different income groups.Despite certain limitations, such as potential overestimation and reliance on selfreporting, the FFQ-NK offers a robust foundation for future epidemiological research in understanding dietary influences on health in this specific population.

Figure 1 :
Figure1: Bland-Altman Analysis for total energy and macronutrients (protein and carbohydrate): Each participant's (n=100) difference in nutrient intakes between the FFQ (Y-axis) and the average of the 3-day dietary records is plotted against the mean intake from the two methods (X-axis).

Figure 1 (
Figure 1(c): Carbohydrate(gm): Mean Difference = 12.13 (95% CI: 6.64-17.62).The Bland-Altman plot for carbohydrate intake reveals a mean difference of 12.13, with a 95% CI spanning from 6.64 to 17.62 The systematic trend indicates that, on average, the FFQ tends to provide higher estimates for protein, carbohydrate, and total energy compared to the 3-day dietary records.

Table 2 :
Mean nutrient intakes across demographic Factors SD, standard deviation; MUFA, monounsaturated fatty acid; PUFA, polyunsaturated fatty acids.Mean nutrient intakes were derived from the initial administration of the FFQ.

Table 4 :
Associations with reported intake differences: FFQ vs. 3-Day Dietary Record values for significance.'0.00' indicates p-values less than 0.001.'**'denotes significance at the 0.01 level, '*' at the 0.05 level.SD, standard deviation; MUFA, monounsaturated fatty acid; PUFA, polyunsaturated fatty acids.Positive correlation values indicate a positive linear relationship, while negative values suggest an inverse association.A high positive correlation would indicate strong agreement between nutrient esti-mates from the FFQ and the 3-day 24-hour DR, implying that the FFQ is a reliable tool for capturing dietary patterns.The correlation analysis serves as a validity assessment, offering insights into how well