Reliability And Validity Of Dietary Survey Methods In Anhui Region

ObjectiveHow to accurately and reliably assess long-term dietary intake is the greatest challenge facing nutritional epidemiology.The food frequency questionnaire（FFQ）is one of the most effective dietary survey methods for large-scale epidemiological studies.There is a unique food culture in the Anhui region,with Hui cuisine as the main feature.However,the reliability and validity of FFQ in Anhui region have not yet been evaluated.In addition,previous studies have mainly focused on total energy and nutrients,while studies of the reliability and validity of dietary patterns are limited.In addition to FFQ,24-hour dietary recall has also been widely used in large-scale health and nutrition surveys,with three commonly used ways including a random dietary recall（way-1）,2 random dietary recalls（way-2）,and 3 consecutive dietary recalls（2 weekdays and 1 weekend day,way-3）.However,the validity of the three ways of 24-hour dietary recall is unclear.Therefore,the objectives of this study were to 1)assess the reliability and validity of the FFQ compiled by our team for use in large-scale epidemiological studies in Anhui region,and2)evaluate validity of the three commonly used ways of 24-hour dietary recall.MethodsThe participants consisted of 620 volunteers from 13 prefecture-level cities in Southern Anhui,Central Anhui,and Northern Anhui,which constituted the Anhui Lifestyle Validation Study（ALVS）.Participants started the first FFQ（FFQ-1）on July 2021 and completed the second FFQ（FFQ-2）1 year later（from July to October 2022）.Meanwhile,the morning urine samples were collected before and after 1 year.To simultaneously account for the day-to-day variation in dietary intake including day of the week,seasonal effect,and random within-person variation,482 volunteers completed 4（1 time per season）consecutive 3-day（2 weekdays and 1 weekend day）dietary recalls（i.e.,12-day dietary recalls）,and 138 completed 2 consecutive 7-day dietary records 1 month apart.The reliability of FFQ was assessed by comparing the 2 FFQs.The validity of FFQ was evaluated by comparing the data of FFQ-2 and those from the average of 12-day dietary recalls or 14-day dietary records.Individual dietary components were also assessed by comparing the agreement between the FFQ measurements and the mean concentrations of biomarkers in the morning urine.The validity of 24-hour dietary recall was evaluated by comparing the consistency between the three survey ways and the average of 12-day dietary recalls.A total of 13 dietary patterns were constructed,including 5 low-carbohydrate diets（LCDs）,i.e.,overall LCD,plant-based and animal-based LCDs,and healthful and unhealthful LCDs;3 low-fat diets（LFDs）,i.e.,overall LFD and healthful and unhealthful LFDs;3 plant-based diet indices（PDIs）,i.e.,overall PDI,healthful PDI（h PDI）,and unhealthful PDI（h PDI）;alternative Mediterranean diet（AMED）,and dietary approaches to stop hypertension（DASH）.The reliability and/or validity of total energy,22 nutrients,16 foods/food groups,and 13 dietary patterns were validated in the present study.To exclude the interference of total energy intake,the residual method was used to adjust for total energy intake.The intraclass correlation coefficients（ICC）and energy-adjusted ICCs were used to evaluate the reliability.Validity was assessed by the Spearman correlation coefficients and energy-adjusted Spearman correlation coefficients.The deattenuated Spearman correlation coefficients were calculated to correct for random within-person error arising from multiple measurements.We also calculated the energy-adjusted,deattenuated Spearman correlation coefficients.Bland-Altman plots were used to visualize the mean of the difference between the two measures and 95%limits of agreement（LOA）,and to assess the agreement between the method to be evaluated and the reference method.The validity coefficients of soy isoflavones assessed by FFQ were evaluated using the triangulation analysis.Among them,the biomarker of soy isoflavone intake was the total concentration of soy isoflavone in morning urine（i.e.,the average of two morning urine concentrations before and after 1 year）.ResultsIn the study of the reliability and validity of FFQ,487 participants with a mean age of（45.6±15.9）years were included in the analysis.In the reliability of FFQ,ICC for total energy was 0.68.The mean energy-adjusted ICC for nutrients was 0.58,with the top three being carbohydrate（0.76）,vitamin B₃ and manganese（0.72）,phosphorus and potassium（0.70）,and the bottom three being retinol and vitamin A（0.44）,iron（0.45）,and cholesterol（0.46）.The mean energy-adjusted ICC for the foods/food groups was 0.56,with the top three being whole grains（0.68）,Chinese Baijiu（0.66）,white rice,steamed bun,and refined grains（0.64）,and the bottom three being red meat（0.44）,processed meat（0.46）,and aquatic products（0.47）.The mean energy-adjusted ICC for dietary patterns was 0.61,with the top three being u PDI（0.71）,overall PDI and h PDI（0.65）,and DASH（0.63）,and the bottom three being overall LFD（0.55）,unhealthful LCD and LFD（0.57）,AMED,healthful LFD,and animal-based and plant-based LCDs（0.58）.When the validity of FFQ-2 was assessed using the average of 12-day dietary recalls as the reference,the deattenuated Spearman correlation coefficient for total energy was0.55,and the mean energy-adjusted,deattenuated Spearman correlation coefficients for nutrients,foods/food groups,and dietary patterns were 0.27,0.29,and 0.22,respectively.The top three nutrients were vitamin B₂（0.35）,potassium（0.34）,fat,saturated fatty acids,carbohydrate,and selenium（0.32）,and the bottom three were copper（0.13）,zinc（0.16）,and vitamin B₃（0.19）.The top three foods/food groups were noodles（0.61）,Chinese Baijiu and milk and dairy products（0.47）,and total fruits（0.43）,and the bottom three were steamed bun（0.12）,legumes and soy products（0.14）,and aquatic products（0.18）.The top three dietary patterns were u PDI（0.33）,animal-based LCD（0.30）,healthful LFD and DASH（0.26）,and the bottom three were plant-based LCD（0.11）,AMED（0.12）,and healthful LCD（0.14）.Results of Bland-Altman analysis showed that FFQ-2overestimated the intake of total energy,15（68.2%）nutrients,and 8（50.0%）foods/food groups,and underestimated the intake of 4（25.0%）foods/food groups and AMED score（P<0.05）.When the validity of FFQ-2 was assessed using the average of 14-day dietary records as the reference,the deattenuated Spearman correlation coefficient for total energy was0.38,and the mean energy-adjusted,deattenuated Spearman correlation coefficients for nutrients,foods/food groups,and dietary patterns were 0.34,0.40,and 0.40,respectively.The top three nutrients were fat（0.60）,saturated fatty acids（0.57）,and carbohydrate（0.54）,and the bottom three were zinc（0.01）,vitamin B₁（0.02）,and copper and protein（0.14）.The top three foods/food groups were milk and dairy products（0.70）,white rice（0.68）,and sugar-sweetened beverages（0.62）,and the bottom three were whole grains（0.13）,total vegetables（0.21）,and aquatic products（0.26）.The top three dietary patterns were overall LFD（0.56）,unhealthful LCD（0.55）,and healthful LFD（0.51）,and the bottom three were overall PDI（0.18）,AMED（0.29）,and DASH（0.30）.Results of Bland-Altman analysis showed that FFQ-2 overestimated the intake of 5（22.7%）nutrients and6（37.5%）foods/food groups and underestimated the intake of 3（13.6%）nutrients and 2（12.5%）foods/food groups and AMED score（P<0.05）.In the triangulation analysis,the deattenuated Spearman correlation coefficients for soy isoflavone intake assessed by FFQ-1,FFQ-2,and mean of FFQ were 0.34,0.12,and0.32,respectively,when using the average of 12-day 24-hour dietary recalls as the reference method,and were 0.14,0.16,and 0.21,respectively,when using the average of morning urine concentrations as the reference method.Finally,the validity coefficients of soy isoflavones intake evaluated by FFQ-1,FFQ-2,and the mean of FFQ were 0.66,0.42,and 0.78,respectively.In the study of the validity of 24-hour dietary recall,a total of 327 participants with a mean age of（47.6±15.7）years were included in the analysis.For the validity of total energy and nutrients assessed by the three ways of 24-hour dietary recall,the deattenuated Spearman correlation coefficients for total energy were 0.49,0.83,and 0.82,respectively,and the mean energy-adjusted,deattenuated Spearman correlation coefficients for nutrients were 0.43,0.75,and 0.77,respectively.The top three nutrients of way-1 were vitamin B₃（0.58）,carbohydrate（0.52）,and fat（0.51）,and the bottom three were iron（0.26）,copper and retinol（0.32）,and vitamin A（0.33）.The top three nutrients of way-2were carbohydrate（0.86）,protein（0.84）,and selenium（0.83）,and the bottom three were zinc（0.57）,copper（0.58）,and iron（0.64）.The top three nutrients of way-3 were vitamin B₂（0.86）,vitamin B₃（0.85）,and saturated fatty acids（0.84）,and the bottom three were copper（0.61）,vitamin B₁（0.68）,and manganese（0.70）.When the correlation coefficients of 22 nutrients were compared among the three ways,all the correlation coefficients of way-1 were lower than those of both way-2 and way-3;12（54.5%）correlation coefficients of way-2 were lower,while 9（40.9%）were higher than those in way-3,and both of polyunsaturated fatty acids were 0.78.Results of Bland-Altman analysis showed that way-1 and way-2 underestimated the intake of 17（77.3%）and 5（22.7%）nutrients,respectively,while way-3 overestimated the total energy and carbohydrate intake（P<0.05）.When the range of 95%LOA was compared among the three ways,total energy and 19（86.4%）nutrients were larger in way-1 than those in way-2 and way-3;total energy and 9（40.9%）nutrients in way-2 were smaller,while 13（59.1%）nutrients were larger than those in way-3.For the validity of foods/food groups assessed by 24-hour dietary recall,energy-adjusted,deattenuated Spearman correlation coefficients of the top three foods/food groups in way-1 were refined grains（0.61）,white rice and whole grains（0.60）,and Chinese Baijiu（0.52）,and the bottom three were steamed bun,fried foods,and milk and dairy products（the correlation coefficients were not calculated because the energy-adjusted correlation coefficients for these three foods/food groups were negative or zero）.The top three foods/food groups in way-2 were white rice（0.92）,noodles,refined grains and total vegetables（0.85）,and total fruits（0.79）,and the bottom three were steamed bun（correlation coefficient was not calculated）,milk and dairy products（0.07）,and processed meat（0.46）.The top three foods/food groups in way-3 were white rice（0.84）,total vegetables（0.83）,noodles and refined grains（0.78）,and the bottom three were steamed bun（correlation coefficient was not calculated）,fried foods（0.05）,and whole grains（0.62）.When the energy-adjusted,deattenuated Spearman correlation coefficients（the deattenuated Spearman correlation coefficients were used if these correlation coefficients were not calculated）of 16 foods/food groups were compared among the three ways,all the correlation coefficients of way-1 were lower than those of both way-2 and way-3;6（37.5%）correlation coefficients of way-2 were lower than way-3 but 10（62.5%）were higher than way-3.Results of Bland-Altman analysis showed that the three ways of 24-hour dietary recalls underestimated the food/food group intake of 7（43.8%）,5（31.3%）,and 4（25.0%）foods/food groups,respectively.When the range of 95%LOA was compared among the three ways,14（87.5%）foods/food groups were larger in way-1 than those in way-2 and all were larger than those in way-3;8（50.0%）foods/food groups in way-2 were smaller,while 8（50.0%）were larger than those in way-3.For the validity of dietary patterns assessed by 24-hour dietary recall,the mean energy-adjusted,deattenuated Spearman correlation coefficients in the three ways were0.41,0.74,and 0.70,respectively.The top three dietary patterns in way-1 were overall and animal LCDs（0.55）,healthful LCD（0.50）,unhealthful LCD and LFD（0.46）,and the bottom three were AMED（0.15）,overall PDI（0.24）,and h PDI（0.36）.The top three dietary patterns in way-2 were overall LCD（0.87）,plant-based LCD（0.85）,and healthful LCD（0.82）,and the bottom three were AMED（0.40）,overall PDI（0.61）,DASH,and overall and healthful LFDs（0.73）.The top three dietary patterns in way-3 were unhealthful LFD（0.79）,animal-based LCD（0.76）,unhealthful LCD and overall LFD（0.75）,and the bottom three were AMED（0.48）,h PDI（0.64）,and overall PDI（0.65）.When the correlation coefficients of 13 dietary patterns were compared among the three ways,all the correlation coefficients of way-1 were lower than those of both way-2 and way-3;4（30.8%）correlation coefficients of way-2 were lower than way-3 but 7（53.8%）were higher than way-3,and both of healthful and unhealthful LFDs were 0.73 and 0.79,respectively.Results of Bland-Altman analysis showed that all three ways of 24-hour dietary recalls overestimated the AMED score while underestimated h PDI score（P<0.05）.In addition,way-1 overestimated the overall PDI score and way-2 underestimated the DASH score（P<0.05）.When the range of 95%LOA was compared among the three ways,all the dietary patterns in way-1 were lower than those in both way-2 and way-3;7（53.8%）dietary patterns in way-2 were smaller,while 6（46.2%）were larger than those in way-3.ConclusionThe FFQ developed by our team demonstrates good reliability（most reliability coefficients>0.50）and acceptable validity（most validity coefficients>0.20）in assessing total energy,and most nutrients,foods/food groups,and dietary patterns among the population in Anhui region.In addition,2 random days or 3 consecutive days（2 weekdays and 1 weekend day）of dietary recalls（most validity coefficients>0.50）seem superior to one-day dietary recall（most validity coefficients>0.20 but<0.50）.This study validates the reliability and validity of the FFQ developed by ourselves for the Anhui region and provides a scientific basis for guiding and optimizing large-scale nutrition surveys.