Wen-Harn Pan, PhD Institute of Biomedical Sciences Academia Sinica & Institute of Population Health Sciences National Health Research Institutes Taiwan On April 10, 2014 & in Singapore
Collaborators and coworkers Hsing-Yi Chang: PhD/ biostatistician, NHRI Measurement error correction for prevalence estimation Chau-JengPeng: PhD candidate/nutritionist, Nat l Chang-Kung University Food security study Yeh, Naihua MS/nutritionist, Academia Sinica 24-h recall methodology Jung-Chang Lee: PhD/ EnvirHealth Scientist, Nat l Chang-Kung University Food security study
Topics of interest Food/nutrition guide Food /nutrient supply Food/ nutrient consumption Food /nutrient adequacy
Content Dietary assessment methods & Nutrition and Health Survey in Taiwan (NAHSIT) Food models (24-h recall) Data capturing system (24-h recall) Monitoring food security,nutrient/food intake adequacy, & food safety Food/nutrient availability compared to DRI Estimate intake inadequacy using 24-h recall data -Method to remove day-to-day variations -Method to handle a large amount of zero s
Nutrition and Health Survey in Taiwan since 1993 Food frequency questionnaire 24 hour-recall 1993-1996 1999-2000 2001-2002 2005-2008 2010 2011 2012 2013~now 13-64 years 64 years+ 6-12 years 1-6 & 19 years+ 13-15 years 16-18 years 7-12 years All ages 5
Nutrition and Health Survey in Taiwan (NAHSIT) NAHSIT I 4+ Elderly NAHSIT II Elementary School Children NAHSIT III Household <6 & >18 NAHSIT school students Year 1993-96 1999-2000 2001-02 2005-08 2010 12-16 yr n=1971 Age >=4 yr >=65 yr 6-12 yr 0-6 yr & >=19 yr Sample size Strata 9962 1937 2419 6189 Metropolitan areas, Province & urbanization Class I, Urbanization Class II, Eastern areas, Hakka areas, Mountainous areas, Penghu islands 1 st -3 rd northern areas, 1 st -3 rd central areas, 1 st -3 rd southern areas, Eastern areas, Hakka areas, Mountainous areas, Penghu islands 1 st Northern areas, 2 nd Northern areas, Central areas, Southern areas, Eastern areas, Hakka areas, Mountainous areas, Penghu islands 2011 15-19 yr n=1279 2012 6-12 yr n=2000 1 st Northern areas, 2 nd Northern areas, Central areas, Southern areas, Eastern areas, Mountainous areas 6
問卷訪視
體檢 -DEXA
Questionnaires Socio-demographics Disease history & related questionnaires Allergy, weight reduction, bone health, irritable bowl syndrome,bsrs5,anorexiz/bulimia, sleep quality, cognitive function) 24-H recall FFQ, dietary habits & food/nutrient supplements Physical activity KAP & health belief
Physical examination Anthropometrics Blood chemistry & nutritional biochemistry Urinary electrolytes and iodine Bone density and Body composition by DEXA Toe nails Ca/Zn/Cu/Se etc Physical fitness Others Blood pressure/central BP, ECG
Food consumption trends at national level Food balance sheet data from FAO
Dietary assessment methods at individual level Methods Merits Faults 24-h recall Good for quantifying mean intake Not representing usual intakes for individual Food frequency questionnaire Good forranking people to study diet-disease associations Can not provide accurate estimates of nutrient /or food intake
Features of the 24-hour recall in NAHSIT Data record form onsite entry to computerized system Food piece models and validations Details on foods and recipes Any potential regrouping of foods is possible
Food preparation place Saltiness Oiliness Cooking method Cooking DATE TIME Dish name
Cooking method Food shapes Meaures for qantitation Dish 1 Dish 2 Foods name (Dish content) Model s & states
Amount consumed Consumpation Date Time Dish 1 Dish 2
Smart searching for food/dish names
Food models 1. Abstract models 14 hollow hemispheres, a deck of 0.3cm-thick cards), tape ruler, transparent board with 2cmx2cm squares 2. Measuring cup(240cc), measuring spoon(4) 3. Portable electric balance 4. Salt & water 5. Food piece models (12) 玉米粒 白色滾刀塊 肉末 橘色小方丁 綠豌豆仁 四季豆 肉絲 米色粗絲 綠色粗絲 肉片 米色薄片 大葉片 18
Food Piece Models Examples:
玉米粒 用於顆粒狀之食物, 如 : 玉米粒 米飯 炒蛋 花生等 20
玉米粒模型 21
白色滾刀塊 用於較大塊且佔體積之食物, 不論是否為滾刀塊, 如 : 筍塊 紅蘿蔔塊 馬鈴薯塊 芋頭塊 菜心 花椰菜 雞塊 松茸 ( 整棵 ) 等 22
白色滾刀塊 - 可排成雞腿等大塊食物 23
肉末 用於細末或顆粒較小之食物, 如 : 絞肉 蔥花 芹菜小段 香菜 蝦米 魚肉鬆 炒蛋等 24
橘色小方丁 用於任何顏色丁狀的食物, 如 : 紅蘿蔔丁 豆乾丁 馬鈴薯小丁等 25
綠豌豆仁 用於較大顆粒形的食物, 大小不一定要和此綠豌豆仁一樣, 如 : 豌豆仁 蓮子 花生 毛豆等 26
四季豆 用於長形且立體的食物, 如 : 四季豆 豆乾條 西洋芹段 炸薯條 芥藍菜 甜豆 肉豆等 27
肉絲 用於絲狀的食物, 如 : 肉絲 紅蘿蔔絲 豆乾絲 韭菜段 芹菜段 青椒絲 小魚乾等 28
米色粗絲 輔助不足的咖啡色粗絲 如 : 干絲 豆芽菜 麵條 米粉 麵線 麵條 油麵 冬粉等 29
綠色粗絲 可用於韭菜花 韭菜 韭黃或可當作有梗葉菜類梗之部份的模型, 也可輔助絲狀模型之不足 30
肉片 直接代表肉片, 或用於所有薄片形的食物, 不論是圓形或橢圓形, 如 : 麵筋 豆乾片 紅蘿蔔片 肝臟片 豌豆片 松茸片等, 可與米色薄片相輔相成 31
米色薄片 用於輔助咖啡色小長方薄片不足時用, 或白色之薄片, 如 : 雞肉片 豆腐片 筍片 菜心片等 32
大葉片 除梗較粗的葉菜類以外的所有葉菜類, 如 : 茼蒿 高麗菜 蕃薯葉等 33
Abstract Models Examples:
鹽 用於評估鹽與粉狀類食品 ( 如 : 味精 糖 奶粉 麥片等 ), 記錄鹽重量 ( 克 ) 因鹽容易潮解並凝聚成塊, 請隨時注意更換 35
半球型模型 為 14 個大小不同之半球型, 直徑依序為 2 至 15 公分 適用於估計圓球形圓球形的食物, 如 : 整棵之蔬菜或水果 洋蔥 蘋果 柳丁 橘子 龍眼 荔枝 柚子 美濃瓜 貢丸 魚丸等 36
厚度尺 伸縮尺 厚度尺 ( 高度尺 )-- 為 10 片重疊塑膠片, 每片厚度固定為 0.3 cm 用於估計食物之厚度 由受訪者疊出厚度後, 將片數乘以 0.3, 即為厚度 ( 公分 ), 記錄此數據 伸縮尺 -- 當長度較長時, 厚度尺不適用, 可改用伸縮尺估計食物的長度 如 : 甘蔗的長度 油條長度 請受訪者以手比出長度, 訪視者再以伸縮尺丈量其長度 ; 或請受訪者直接以伸縮尺表示長度 37
方格板 用於計算底面積 量白紙上所畫食物之格數, 再換算成面積, 以平方公分為單位記錄 格數應算至小數點下一位, 再乘上 4 即可代表有多少平方公分 38
Some topics of interest for food security issues 1.Does country provide sufficient food to citizens comparing to other countries? 2.What are trends of food and nutrient supply? Any signs need attention? 3.Do people in Taiwan consume food and nutrients to their needs?
Preliminary data 1. Does country provide sufficient food to citizens comparing to other countries?
Preliminary data 2.1 What are trends of food supply? Any signs need attention? (Taiwanese FBS data) A. Trends in servings of daily food availabilities from Taiwan food balance sheets, 1991 to 2010 16.0 Cereal/roots_supply 3.0 Vegetables_supply 2.6 Fruits_supply 0.9 Dairy_supply 9.0 Soy/fish/meat/egg_supply 16 Oil/nuts_supply 15.5 15.0 2.8 2.4 0.8 8.6 15 servings 14.5 14.0 2.6 2.4 2.2 0.7 0.6 8.2 7.8 14 13 13.5 13.0 Slope since 1992 y =15.789-0.119x R² = 0.842 trend p<0.001 2.2 Slope since 1997 y =2.946-0.021x R² = 0.297 trend p=0.044 2.0 Slope since 1997 y =2.482-0.022x R² = 0.437 trend p=0.010 0.5 Slope since 1997 y =0.937-0.019x R² = 0.882 trend p<0.001 7.4 Slope since 1997 y =9.130-0.085x R² = 0.718 trend p<0.001 12 Slope since 1997 y =15.443-0.133x R² = 0.424 trend p=0.012 12.5 2.0 1.8 0.4 7.0 11 B. Trends in the ratios of the food availabilities to weighted food needs derived from the composite Taiwan Food Guides, 1991 to 2010 2.6 Cereal/roots_supply 2.6 Vegetables_supply 2.6 Fruits_supply 2.6 Dairy_supply 2.6 Soy/fish/meat/egg_supply 2.6 Oil/nuts_supply 2.4 2.4 2.4 2.4 2.4 2.4 2.2 2.2 2.2 2.2 2.2 2.2 Ratio of supply to food needs 2.0 1.8 1.6 1.4 1.2 1.0 0.8 Slope since 1992 y =1.236-0.008x R² = 0.815 trend p<0.001 2.0 1.8 1.6 1.4 1.2 1.0 0.8 Slope since 1997 y =1.091-0.01x R² = 0.421 trend p=0.012 2.0 1.8 1.6 1.4 1.2 1.0 0.8 Slope since 1997 y =1.266-0.012x R² = 0.465 trend p=0.007 2.0 1.8 1.6 1.4 1.2 1.0 0.8 Slope since 1997 y =0808-0.012x R² = 0.808 trend p<0.001 2.0 1.8 1.6 1.4 1.2 1.0 0.8 Slope since 1997 y =2.435-0.026x R² = 0.777 trend p<0.001 2.0 1.8 1.6 1.4 1.2 1.0 0.8 Slope since 1997 y =2.579-0.023x R² = 0.454 trend p=0.008 0.6 0.6 0.6 0.6 0.6 0.6 0.4 0.4 0.4 0.4 0.4 0.4
Preliminary data 2.2 What are trends of nutrient supply? Any signs need attention? (Taiwanese FBS data) C. Trends in the ratios of available nutrients (using Taiwan food balance sheet data) to their DRIs, 1991 to 2010 Energy Protein Calcium Phosphrous Iron ratio of available nutrients to DRIs 2.5 2.0 1.5 1.0 0.5 Slope since 1997 y =+ 1.851-0.013x R² = 0.684 trend p<0.001 2.5 2.0 1.5 1.0 0.5 Slope since 1997 y =2.102-0.020x R² = 0.805 trend p<0.001 2.5 2.0 1.5 1.0 0.5 Slope since 1997 y =0.699-0.008x R² = 0.726 trend p<0.001 2.5 2.0 1.5 1.0 0.5 Slope since 1997 y =1.892-0.023x R² = 0.861 trend p<0.001 2.5 2.0 1.5 1.0 0.5 Vitamin A Vitamin C Vitamin B1 Vitamin B2 Niacin 2.0 1.5 1.0 0.5 ratio of available nutrients to DRIs2.5 Slope since 1997 y =2.790-0.022x R² = 0.662 trend p<0.001 2.5 2.0 1.5 1.0 0.5 Slope since 1997 y =1.675-0.018x R² = 0.825 trend p<0.001 2.5 2.0 1.5 1.0 0.5 Slope since 1997 y =1.844-0.018x R² = 0.495 trend p=0.005 2.5 2.0 1.5 1.0 0.5 Slope since 1997 y =1.974-0.013x R² = 0.649 trend p<0.001 2.5 2.0 1.5 1.0 0.5 Slope since 1997 y =1.852-0.022x R² = 0.829 trend p<0.001
Often asked questions for monitoring food intakes, nutritional status and environmental pollutant exposure Prevalence of nutrient deficiencies or excess? Proportion of people who eat 5 V+F or more a day? Proportion of people who eat beef (or peanuts)? Corrected distribution for various food groups for TDS (Total Diet Study)?
Different Number of Days assessed
Estimating Population Distribution from information of 1 day obtained via 24h recall? 24h recall data subject to within individual variations Nusseret. Al, 1997, JASA Transform data to normally distributed via Box-Cox transformation Estimate the within individual variation Remove the within individual variation from the total variance Back transform the data to original scale based on normal with the new variance
Our approach using Gamma distribution HY Chang, C Suchindran, WH Pan. Stat in Med. 2001
Estimation Steps Estimate the within individual variance and the ratio of within to among individual variance Sampling weights Estimate the parameters of the observed distribution accounting for sampling weights Estimate the adjusted variance by applying the results of 1 to the results of 3 Reconstruct the adjusted distribution using the results of 4 SAS program
Ratio of Within to Among Variation
Estimating Within Individual Variation Repeated Measures are needed NutrientY = i µ + subjec + ε where ε represents day-to-day variation The ratio of within to among individual variance (or 2 intra/inter variability) is σ w ρ = 2 σ 1 γ It can be estimated by γ where γ is the average correlation coefficient between repeated measurements. b ( K Liu et al, Journal of Chronic Disease 1978)
Nutrient 1% 25% 50% 75% 99% Energy (KCAL) 1662.1 (552.9) 1996.0 (1602.2) 2368.1 (2164.8) 2975.4 (2857.2) 4004.1 (8502.7) Protein (g) 52.2 ( 21.9) 78.2 ( 60.2) 90.9 ( 81.6) 105.4 ( 108.3) 145.7 ( 411.0) Fat (g) 39.7 ( 5.7) 66.7 ( 31.6) 80.7 ( 58.0) 97.0 ( 105.6) 145.4 ( 636.8) Carbohydrate (g) 151.0 ( 49.0) 252.0 ( 213.2) 304.5 ( 292.3) 360.4 ( 370.4) 548.3 ( 814.3) Fiber (g) 1.7 ( 0.2) 3.8 ( 2.0) 5.0 ( 3.2) 6.5 ( 5.8) 11.2 ( 46.9) Calcium (mg) 220.7 ( 23.2) 382.2 ( 209.3) 467.9 ( 369.5) 564.0 ( 600.9) 851.4 (1951.0) Phosphorus (mg) 630.9 (198.7) 1007.4 ( 764.4) 1192.6 (1039.7) 1423.6 (1475.2) 2022.4 (5830.9) Iron (mg) 8.2 ( 1.7) 13.4 ( 8.4) 16.0 ( 11.9) 18.9 ( 18.7) 28.3 ( 117.7) Vitamin A (I. U.) 3251.7 ( 80.5) 5630.0 (1533.3) 6843.7 (3188.7) 8260.3 (6885.6) 12932.0 (56173.6) Vitamin B (mg) 0.8 ( 0.2) 1.2 ( 0.8) 1.5 ( 1.2) 1.7 ( 1.7) 2.5 ( 7.9) Vitamin B2 (mg) 0.7 ( 0.2) 1.2 ( 0.8) 1.4 ( 1.1) 1.7 ( 1.6) 2.4 ( 6.7) Niacin (mg) 9.4 ( 4.0) 15.7 ( 10.8) 19.0 ( 15.9) 23.0 ( 22.0) 34.5 ( 95.5) Vitamin C (mg) 47.2 ( 0.5) 119.5 ( 42.6) 161.9 ( 92.3) 214.7 ( 175.0) 400.0 (1425.9) SFA (g) 11.8 ( 1.7) 21.9 ( 10.4) 27.1 ( 17.9) 33.3 ( 34.9) 52.6 ( 218.5) PFA (g) 11.4 ( 1.0) 17.5 ( 7.8) 21.2 ( 14.4) 24.2 ( 28.1) 34.0 ( 110.0) Oleic acid (g) 11.9 ( 1.9) 23.7 ( 11.4) 30.3 ( 19.6) 37.6 ( 39.6) 61.3 ( 281.6) Cholesterol (mg) 169.2 ( 3.7) 330.8 ( 175.1) 417.1 ( 393.8) 521.5 ( 620.9) 837.8 (1240.0)
% Not reaching DRI for 2013NAHSIT Based on corrected 24-h recall Ca P Fe
% Not reaching DRI for 2013NAHSIT Based on corrected 24-h recall Mg Zn
% Not reaching DRI for 2013NAHSIT Based on corrected 24-h recall
% Not reaching DRI for 2013NAHSIT Based on corrected 24-h recall
Problem With Large Number of 0 s There are two kinds of zero s in the 24 hr. recall data. One is that the individual never eats the food. The other is the individual did not eat the food 24 hours before the interview. The Food Frequency Questionnaire provides information on whether an individual consume certain food.
Method x We use the Zero-Inflated Model to model usual food intakes with large number of zero s. The distribution function of the amount of usual food intakes is following: F( x) = ( 1 p) + pf ( x) if x > 0 Otherwise F( 0) =1 p is a random variable that represents the amount of usual intake for some kind of food is the probability that an individual consume this food p a
24 hr recall Repeated 24 hr recall FFQ Estimation of population mean (X) Estimation of correlation coefficients (r) Estimation of the proportion of eating certain food group (p) Excel can be used to obtain the percentile. For example, the 25 th percentile is calculated using the Excel comment, GAMMAINV((.25-(1-p))/p, α, β).
Example (estimation of the distribution) 利用 Excel 函數 GAMMAINV(probability,alpha,beta) 畫出食用量的分配圖形 相關參數 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 50 100 150 200 250 300 350
Example (estimation of the quantiles) 利用 Excel 函數 GAMMAINV(probability,alpha,beta) 計算分配 quantiles 的估計值
Rice G
Green Vegetable G
Berries age Proportion of zero s Grams mean p25 p50 p75 p90 Males 19-44 0.20 68.24 0.04 9.51 71.83 209.58 Males 45-64 0.17 97.76 3.91 41.19 132.81 273.50 Males >=65 0.19 73.38 0.16 13.83 83.18 222.86 Females 19-44 0.13 55.41 0.00 2.10 38.57 166.69 Females 45-64 0.10 64.69 0.09 6.44 59.26 197.46 Females >=65 0.24 49.99 0.00 0.90 30.66 149.94
For Total Diet Study
National Food Consumption Database in Taiwan- Food classification system Nutritional databases, laws, and regulations were compared and incorporated into this classification system Nutrition and Health Survey in Taiwan (NAHSIT) Total Diet Study in Taiwan (established by the China Medical University) Standards for antibiotics, pesticides, and drugs residue levels in animals Standards for fish and aquatic animals classification Standards for pesticides and drugs residue levels in vegetables and fruits Other food classification requirements from health risk assessment experts
Food code Nutrition and Health Survey Food Group National Food Consumption Database Food Group A Grain, legumes, roots, and tubers Grain and legumes B Cooking oil Cooking oil C Poultry and it's products Poultry and it's products D Live stock and other meat Live stock and other meat E Fish and aquatic animal Fish and aquatic animal F Protein food and product (including egg, milk, and soybean and their products) G Vegetables Milk and milk products (including cow and goat milk) H Fruits Fruits I Snack and beverages Vegetables J Wine, beer, and other alcoholic products Snacks (including candy, chocolate, and other snacks) K Seasonings and condiments (including sugar, salt, soy sauce, and other condiment product) Egg Beverages L Other food Wine, beer, and other alcoholic products M N Seasonings and condiments (including sugar, salt, soy sauce, and other condiment product) Other food (including composite food) O Infant food (under age 3)
National Food Consumption Database in Taiwan Website: http://intakes.nhri.org.tw/ About us Website navigation: Database and Downloads Food consumption data request application Ask a question Related topics website link Contact us
National Food Consumption Database in Taiwan http://intakes.nhri.org.tw/ 73
Health risk for general population Monte Carlo Simulation Phthalates concentration ADD = Each group Mean(min-max) 100% C IR AF BW Each group Mean(min-max) HQ = ADD TDI HI = i= 1 HI > 1 described as indicating that a potential may exist for adverse human health HI < 1 indicates that no adverse human health effects are expected to occur TDI : Tolerable Daily Intakes (µg/kg-bw/day) n: Number of chemicals For the adverse hepatic effects For the anti-androgenic effects n HQ ADD: Average Daily Dose (µg/kg/day). C: Concentration of PAEs weight exposure (µg/kg) IR: Ingestion rate (g/day), considering gender, age, food items. AF: Absorption Fraction, assumed to 100% in potential dose BW: Body weight (kg), considering gender, age. (National food consumption Database) PAEs Endpoint RfD a Endpoint TDI b DEP Organ weight 800 Organ weight 500 c DBP Increased mortality 100 Reproductive effect 10 BBP Liver effect 200 Reproductive effect 500 DEHP Liver weight 20 Reproductive effect 50 DIDP - - Liver effect 150 DINP - - Liver effect 150 a USEPA RfD, Reference Dose (µg/kg-bw/day) b EFSA TDI, Tolerable Daily Intake (µg/kg-bw/day) c WHO TDI (µg/kg-bw/day) 74
Conclusion 1. 24-h recall data assisted by FFQ is important in monitoring food adequacy and food safety. 2. Food security issue can be monitored via combining with food balance sheet data. 3. Comprehensive database is worthy the efforts. 4. Methodological advancement is warranted.