Diabetes Care 1

Jieli Lu,1 Mian Li,1 Yu Xu,1 Yufang Bi,1 Early Life Famine Exposure, Ideal Yingfen Qin,2 Qiang Li,3 Tiange Wang,1 Ruying Hu,4 Lixin Shi,5 Qing Su,6 Min Xu,1 Cardiovascular Health Metrics, Zhiyun Zhao,1 Yuhong Chen,1 Xuefeng Yu,7 Li Yan,8 Rui Du,1 Chunyan Hu,1 Guijun Qin,9 and Risk of Incident Diabetes: Qin Wan,10 Gang Chen,11 Meng Dai,1 Di Zhang,1 Zhengnan Gao,12 Findings From the 4C Study Guixia Wang,13 Feixia Shen,14 Zuojie Luo,2 15 16 4 https://doi.org/10.2337/dc19-2325 Li Chen, Yanan Huo, Zhen Ye, Xulei Tang,17 Yinfei Zhang,18 Chao Liu,19 Youmin Wang,20 Shengli Wu,21 Tao Yang,22 Huacong Deng,23 Donghui Li,24 Shenghan Lai,25 Zachary T. Bloomgarden,26 Lulu Chen,27 Jiajun Zhao,28 Yiming Mu,29 Guang Ning,1 and Weiqing Wang,1 for the 4C Study Group

OBJECTIVE We aim to investigate the impact of ideal cardiovascular health metrics (ICVHMs) on the association between famine exposure and adulthood diabetes risk.

RESEARCH DESIGN AND METHODS This study included 77,925 participants from the China Cardiometabolic Disease and Cancer Cohort (4C) Study who were born around the time of the Chinese Great Famine and free of diabetes at baseline. They were divided into three famine exposure groups according to the birth year, including nonexposed (1963–1974), fetal exposed (1959–1962), and childhood exposed (1949–1958). Relative risk regression was used to examine the associations between famine exposure and 1Shanghai National Clinical Research Center for ICVHMs on diabetes. MetabolicDiseases,KeyLaboratoryforEndocrine and Metabolic Diseases of the National Health RESULTS Commission of the PR China, Shanghai National During a mean follow-up of 3.6 years, the cumulative incidence of diabetes was Center for Translational Medicine, Shanghai In- stitute of Endocrine and Metabolic Diseases, 4.2%, 6.0%, and 7.5% in nonexposed, fetal-exposed, and childhood-exposed RISK METABOLIC AND CARDIOVASCULAR Department of Endocrine and Metabolic Dis- participants, respectively. Compared with nonexposed participants, fetal-exposed eases, Ruijin Hospital, Shanghai Jiao Tong Uni- but not childhood-exposed participants had increased risks of diabetes, with versity School of Medicine, Shanghai, China 2 multivariable-adjusted risk ratios (RRs) (95% CIs) of 1.17 (1.05–1.31) and 1.12 (0.96– The First Affiliated Hospital of Guangxi Medical University, Nanning, China 1.30), respectively. Increased diabetes risks were observed in fetal-exposed individuals 3The Second Affiliated Hospital of Harbin Medical 2 with nonideal dietary habits, nonideal physical activity, BMI ‡24.0 kg/m , or blood University, Harbin, China pressure ‡120/80 mmHg, whereas significant interaction was detected only in BMI 4Zhejiang Provincial Center for Disease Control strata (P for interaction 5 0.0018). Significant interactions have been detected and Prevention, Hangzhou, China 5 fi P Af liated Hospital of Guiyang Medical College, between number of ICVHMs and famine exposure on the risk of diabetes ( for Guiyang, China interaction 5 0.0005). The increased risk was observed in fetal-exposed participants 6Xinhua Hospital Affiliated to Shanghai Jiao Tong with one or fewer ICVHMs (RR 1.59 [95% CI 1.24–2.04]), but not in those with two or University School of Medicine, Shanghai, China 7 more ICVHMs. Tongji Hospital, Tongji Medical College, Huaz- hong University of Science and Technology, Wu- CONCLUSIONS han, China 8Sun Yat-sen Memorial Hospital, Sun Yat-sen The increased risk of diabetes associated with famine exposure appears to be University, Guangzhou, China modified by the presence of ICVHMs. 9The First Affiliated Hospital of Zhengzhou Uni- versity, Zhengzhou, China 10The Affiliated Hospital of Southwest Medical Emerging evidence indicated that early life development was associated with the risk University, Luzhou, China 11Fujian Provincial Hospital, Fujian Medical Uni- of type 2 diabetes mellitus (T2DM) in adulthood (1,2). Low birth weight was associated versity, Fuzhou, China with a higher risk of diabetes in later life (3–5). In addition, studies of the Ukraine 12Dalian Municipal Central Hospital Affiliated famine and the Dutch “Hunger Winter” famine suggested that exposure to starvation to Dalian Medical University, Dalian, China

Diabetes Care Publish Ahead of Print, published online June 4, 2020 2 Famine Exposure, ICVHMs, and Risk of Diabetes Diabetes Care

in utero was associated with an elevated association between famine exposure The study was approved by the Med- risk of T2DM in later life (6,7). As one of and adulthood diabetes risk. Therefore, ical Ethics Committee of Ruijin Hospital, the largest catastrophes in human his- we conducted this prospective study in a Shanghai Jiao Tong University (Shanghai, tory, the Chinese Great Famine has nationwide large cohort of the China China). Written informed consent was aroused much attention from scholars Cardiometabolic Disease and Cancer Co- obtained from all participants. (8). The Chinese Great Famine had over- hort (4C) Study, with two aims: 1)to whelmingly cardiometabolic consequen- examine the association between early Data Collection ces, including increasing the risk of life famine exposure and risk of T2DM All questionnaire data collection and obesity (9,10) and metabolic syndrome later in life, and 2) to explore whether the anthropometric measurements were (11). Several previous epidemiological ICVHMs might modify the association performed by trained staff according to a studies have shown an association be- between famine exposure and risk of standard protocol at local health stations tween Chinese famine exposure and the diabetes. or community clinics at each study center. risk of T2DM (12–14). However, whether Using a detailed questionnaire, we col- andwhatfactorsinlater lifemight modify RESEARCH DESIGN AND METHODS lected information on sociodemographic this association have not been exten- Study Population characteristics, lifestyle factors, as well as sively investigated. The China Cardiometabolic Disease and medical history through personal inter- Rapid economic development and as- Cancer Cohort (4C) Study was a multi- views. Education levels were divided into sociated dramatic lifestyle changes have center, prospective, population-based highschool educationoraboveversusless led to a substantial increase in the prev- cohort study investigating the associa- than high school. The type and frequency alence of T2DM in China (15–17). The tions of glucose homeostasis with clinical of alcohol consumptions and smoking association between early life develop- outcomes, including diabetes, cardiovas- habits were recorded. Participants were ment and risk of T2DM may be modified cular disease, cancer, and all-cause mor- classified as never, former, or current by lifestyle in adulthood (5,14,18,19). In tality. A total of 20 communities from drinkers according to alcohol drinking 2010, the American Heart Association various geographic regions in China were habits. The information on intensity, du- proposed seven components critical to selected to represent the general pop- ration, and frequency of physical activity ideal cardiovascular health (CVH), includ- ulation in China. Eligible men and women was gathered using the short form of the ing four ideal health behaviors (nonsmok- aged $40 years were identified from International Physical Activity Question- ing within the last year, idealBMI,physical local resident registration systems. Trained naire, and the metabolic equivalent mi- activityatgoallevels,andadietarypattern community health workers visited eligible nutes per week were used to estimate recommended) and three biological fac- individuals’ homes and invited them to physical activities (one metabolic equiva- tors (ideal total cholesterol [TC], blood participate in the study. A total of 193,846 lent represents the energy expenditure for pressure (BP), and fasting plasma glucose) individuals were recruited for the study at an individual at rest) (27). A previously (20). There has been accumulating evi- baselinefrom2011to2012(24–26).During evaluated and validated dietary question- dence suggesting that ideal CVH might 2014–2016, all participants were invited to naire (25) was used to collect information be a marker of insulin sensitivity and participate in an in-person follow-up visit. on dietary intake over the past 12 months. related to lower risk of T2DM (21–23). Lifestyle risk factors and medical history The questionnaire was designed to cap- However, previous studies only exam- were queried by trained staff using the ture information on frequency and quan- ined the effect of diet and adult obesity same standard questionnaire as at base- tity of major food items such as red meat, on the association between famine ex- line. Anthropometric and BP measure- fruits and vegetables, dairy, and Chinese posure and the risk of T2DM (14,19). To ments, oral glucose tolerance tests, traditional food like pickles and salty the best of our knowledge, there have and blood samples were obtained using vegetables. been no studies to explore the impact of the same protocol that was used in the Height and weight were measured to these ideal CVH metrics (ICVHMs) on the baseline examination. the nearest 0.1 kg and 0.1 cm separately

13The First Hospital of Jilin University, Chang- 22The First Affiliated Hospital of Nanjing Medical Corresponding author: Weiqing Wang, wqingw61@ chun, China University, Nanjing, China 163.com, or Jieli Lu, [email protected] 14 fi 23 fi The First Af liated Hospital of Wenzhou Med- The First Af liated Hospital of Chongqing Med- Received 20 November 2019 and accepted 23 ical University, Wenzhou, China ical University, Chongqing, China April 2020 15Qilu Hospital of Shandong University, Jinan, 24Department of Gastrointestinal Medical On- China cology, The University of Texas MD Anderson This article contains supplementary material online fi 16Jiangxi Provincial People’s Hospital Affiliated to Cancer Center, Houston, TX at https://doi.org/10.2337/ gshare.12185469. Nanchang University, Nanchang, China 25Johns Hopkins University School of Medicine, J.L., M.L., Y.X., Y.B., Y.Q., Q.L., and T.W. contrib- 17The First Hospital of Lanzhou University, Lanz- Baltimore, MD uted equally to this work. 26 hou, China Icahn School of Medicine at Mount Sinai, © 2020 by the American Diabetes Association. 18 Central Hospital of Shanghai Jiading District, New York, NY Readers may use this article as long as the work is 27 Shanghai, China Union Hospital, Tongji Medical College, Huaz- properly cited, the use is educational and not for 19 Jiangsu Province Hospital on Integration of hong University of Science and Technology, Wu- profit, and the work is not altered. More infor- Chinese and Western Medicine, Nanjing, China han, China mation is available at https://www.diabetesjournals 20 28 The First Affiliated Hospital of Anhui Medical Shandong Provincial Hospital Affiliated to .org/content/license. University, Hefei, China Shandong University, Jinan, China 21Karamay Municipal People’s Hospital, Xinjiang, 29Chinese People’s Liberation Army General China Hospital, Beijing, China care.diabetesjournals.org Lu and Associates 3

with participants wearing lightweight Force of the American Heart Association discovery rate using the Benjamini- clothes and no shoes. BP was tested two (20): never smoked or quit smoking Hochberg method. We further compared times using an automated electronic .12 months prior, BMI ,24.0 kg/m2, the risk estimates in the strata of sex and device (Model HEM-752 with Fuzzy logic; physical activity at goal (at least 150 min/ famine exposure severity. To demonstrate Omron) in a seated position after at week of moderate-intensity physical ac- possible interactions of famine exposure least a 5-min quiet rest, and the means tivity, 75 min/week of vigorous-intensity andICVHMsinthedevelopmentofdi- of two measurements were used in the aerobic physical activity, or an equivalent abetes, we generated interaction terms final data analysis. combination of moderate- and vigorous- using the cross products of famine expo- intensity aerobic activities), dietary score sure with each component of the ICVHMs. Biochemical Evaluation $3 (including four components: fruits The interaction was tested using the likeli- A blood sample was collected in the and vegetables $4.5 cups/day; fish hood ratio test by comparing the full model morning after an overnight fast (at least $198 g/week; sweets/sugar-sweetened including the interaction term with the 10 h). Sera were aliquoted into 0.5-mL beverages #450 kcal/week; and soy pro- reduced model excluding the interaction Eppendorf tubes within 2 h after blood tein $25 g/day), TC ,200 mg/dL (un- term. The df for the P for interaction was collection and shipped in dry iceat 280°C treated; to convert to millimoles per liter, calculated based on the number of expo- to the central laboratory located at multiply by 0.0259), and BP ,120/80 mmHg sure groups (nonexposed, fetal exposed, Shanghai Institute of Endocrine and Met- (untreated) (Supplementary Table 1). Fast- and childhood exposed) and the number of abolic Diseases, which is certified by ingplasmaglucosewasnotincludedasa subgroups for each effect modifier in the the College of American Pathologists. All CVH metric in the main analysis because subgroup analysis. To reduce the bias re- participants underwent an oral glucose plasma glucose is used to define T2DM lated to age differences between famine tolerance test, and plasma glucose was (23). The number of ICVHMs was catego- and postfamine births, an “age-balanced” obtained at 0 and 2 h during the test. rized as #1, 2, 3, 4, and $5CVHmetrics method was used, in which both postfamine Plasma glucose concentrations were an- at the ideal level. and prefamine births were combined as alyzed locally using a glucose oxidase Diabetes unexposed control subjects (13,31). or hexokinase method within 2 h after Incident diabetes was defined as fasting All analyses were conducted using SAS bloodsample collectionunderastringent plasma glucose $7.0 mmol/L, and/or 2-h 9.2 (SAS Institute, Cary, NC), and a two- quality-control program. All regional lab- , postload plasma glucose $11.1 mmol/L, tailed P 0.05 was considered as sta- oratories passed a national standardiza- fi and/or a self-reported previous diagnosis tistically signi cant. tion program and study-specific quality by health care professionals during follow- assurance program. Serum TC was mea- RESULTS up among participants without diabetes at sured at the central laboratory at Ruijin baseline. Among 193,846 participants examined at Hospital using an autoanalyzer (ARCHITECT baseline, 170,240 (87.8%) were followed ci16200 analyzer; Abbott Laboratories, Statistical Analyses up in 2014–2016. Of them, participants Abbott Park, IL). The baseline characteristics of the study with missing baseline information on population by famine exposure groups plasmaglucosemeasurement(n56,074), fi De nitions were compared using the Pearson x2 test diagnosed or screen-detected diabetes at Famine Exposure for categorical variables and Student t baseline (n 5 34,497), missing data on According to the birth time during and test or Mann-Whitney U test for contin- BMI (n 5 2,497), smoking status (n 5 around the Chinese Great Famine, we uous variables. Cumulative incidence of 4,193), diet habits (n 5 4,429), and phys- fi de ned famine exposure subgroups as diabeteswascalculatedduring the follow- ical activity (n 5 2,583) were excluded. nonexposed (born between 1 January up. The association of incident diabetes Additionally, 26,166 participants born be- 1963 and 31 December 1974), fetal ex- withfamineexposurewasexaminedusing fore 31 December 1948 and 11,876 with- posed (born between 1 January 1959 and relative risk regression models (24,30). out glucose measurement at follow-up 31 December 1962), and childhood ex- Model 1 was unadjusted. Model 2 was visit were also excluded, leaving 77,925 posed (1 January 1949 and 31 December adjusted for sex and age (continuous for the current analysis (Supplementary 1958), as in previous studies (12,28). variable). Model 3 included variables in Fig. 1). Among them, 23,926 (30.70%) Famine severity was determined accord- model 2 plus family history of diabetes were men, and the mean age of the ing to the excess death rate for each (yes or no), drinking status (current drinker study participants was 54.5 6 7.6 years. province (12,29), which was calculated as or not), and education status (less than Baseline characteristics of participants the mortality rate change from the av- high school or high school or above). according to categories of famine expo- – erage level in 1956 1958 to the highest Individual CVH metrics were further ad- sure are presented in Table 1. In addition – rate during 1959 1962 (29). Based on the justed in model 4. to the obvious age difference between information, an excess mortality rate of The modifying effect of ICVHMs on the groups, the fetal-exposed group has a 100% was used as a threshold value to association of famine exposure and di- greater proportion of individuals with fi de ne severely and less severely affected abetes wasevaluatedinstratifiedanalyses higher education than the other two areas. by strata of the six individual components groups. There is an increasing trend of TC ICVHMs of ICVHMs and the number of ICVHMs. In and BP level and a decreasing trend of ICVHMs were adapted from the recom- the subgroup analyses, individual CVH four or five or more ICVHMs from non- mendations of the Goals and Metrics was mutually adjusted, and the P value exposed to fetal-exposed to childhood- Committee of the Strategic Planning Task was corrected for multiple testing via false exposed groups. 4 Famine Exposure, ICVHMs, and Risk of Diabetes Diabetes Care

– Table 1—Baseline characteristics of 77,925 participants according to famine 1.17 [95% CI, 1.05 1.31]), but not in child- exposure in early life hood-exposed participants (RR 1.12 Famine exposure [95% CI 0.96–1.30]). Moreover, the in- creased risk of diabetes associated with Nonexposed Fetal Childhood fetal famine exposure was observed in Number of participants (%) 23,582 (30.3) 13,195 (16.9) 41,148 (52.8) severely affected areas, but not in less Age at baseline, years 44.8 6 2.7 50.6 6 1.3 57.3 6 2.8 severely affected areas (P for interac- Male sex 7,185 (30.5) 3,559 (27.0) 13,182 (32.0) tion ,0.001) (Table 2). Furthermore, in BMI, kg/m2 24.2 6 3.6 24.4 6 3.4 24.4 6 3.5 the age-balanced analysis, compared High school education or above 10,518 (44.6) 7,547 (57.2) 13,906 (33.8) with the individuals combined of Current cigarette smoking 4,976 (21.1) 2,536 (19.2) 8,324 (20.2) prefamine and postfamine births as the Current alcohol drinking 2,448 (10.4) 1,290 (9.8) 4,347 (10.6) reference group, the increased risk of Moderate and vigorous physical activity 2,881 (12.2) 1,788 (13.6) 6,344 (15.4) diabetes in the fetal-exposed individuals Family history of diabetes 3,403 (14.4) 2,104 (16.0) 5,009 (12.2) remained statistically significant (RR 1.11 Healthy diet 13,867 (58.8) 7,662 (58.1) 22,745 (55.3) [95% CI 1.03–1.19]) (Supplementary TC, mmol/L 4.6 6 1.1 5.0 6 1.1 5.1 6 1.1 Table 2). Systolic BP, mmHg 123.0 6 17.1 127.0 6 18.0 131.9 6 19.6 Subgroup analysis stratified by the six Diastolic BP, mmHg 76.9 6 11.1 78.1 6 11.1 78.6 6 10.8 ICVHMs was further carried out, and Increased TC 6,550 (27.8) 5,393 (40.9) 18,856 (45.8) individual CVH metrics were mutually Increased BP 13,538 (57.4) 8,684 (65.8) 30,824 (74.9) adjusted in the regression model (Table 3). Compared with nonexposed individuals, Fasting plasma glucose, mmol/L 5.3 6 0.5 5.4 6 0.6 5.5 6 0.6 significantly increased diabetes risk was 2-h plasma glucose, mmol/L 6.5 6 1.6 6.7 6 1.6 6.9 6 1.7 observed among fetal famine-exposed ICVHMs #1 2,255 (9.6) 1,588 (12.0) 5,860 (14.2) individuals with nonideal dietary habits, 2 4,833 (20.5) 3,291 (24.9) 11,854 (28.8) nonideal physical activity, BMI $24 kg/ 2 3 6,909 (29.3) 4,044 (30.7) 12,645 (30.7) m ,orBP$120/80 mmHg, with RRs (95% 4 6,091 (25.8) 2,918 (22.1) 7,570 (18.4) CIs) of 1.29 (1.10–1.52), 1.16 (1.03–1.30), $5 3,494 (14.8) 1,354 (10.3) 3,219 (7.8) 1.20 (1.05–1.36), and 1.22 (1.08–1.38), Data are n (%) or mean 6 SD. respectively. Significant interaction was detected only in BMI strata and famine exposure on risk of diabetes (P for in- During up to 5 years of follow-up participants, both fetal-exposed (age- teraction 5 0.0018), but not across di- (mean 3.6 years), we identified a total of and sex-adjusted risk ratio [RR] 1.21 [95% etary habits, physical activity, and BP (all P 4,842 (6.21%) individuals with incident CI 1.09–1.35]) and childhood-exposed for interaction $0.05). The increased di- diabetes. The mean (SD) follow-up time (age- and sex-adjusted RR 1.20 [95% CI abetes risk associated with fetal famine was 3.67 (0.97), 3.63 (0.91), and 3.59 1.04–1.38]) participants had increased exposure is significant in both ideal (RR (0.90) and the cumulative incidence of risks of diabetes in adulthood. The risk of 1.14 [95% CI 1.01–1.30]) and nonideal diabetes was 4.15%, 6.0%, and 7.47% incident diabetes remained significantly smoking status groups (RR 1.28 [95% CI for the nonexposed, fetal-exposed, and increased after adjusting for all other 1.02–1.59]) (P for interaction 5 0.0621). childhood-exposed group to famine, re- covariates, including individual CVH met- The RR (95% CI) ofdiabetes risk associated spectively. Compared with nonexposed rics in fetal-exposed participants (RR with fetal exposure to famine among individuals with ideal and nonideal TC levels was 1.17 (1.01–1.35) and 1.15 Table 2—RRs (95% CIs) for incident T2DM according to famine exposure in early (0.97–1.37), respectively. No significant life among 77,925 participants interactions have been detected across Famine exposure TC strata (P for interaction 5 0.4103). Nonexposed Fetal Childhood When individual CVH metrics were not mutually adjusted, we observed signif- Case subjects/total number 978/23,582 792/13,195 3,072/41,148 icant differences in smoking status (P Cumulative incidence, % 4.15 6.00 7.47 for interaction 5 0.0013) and BMI (P – – Model 1 1.00 (reference) 1.45 (1.32 1.59) 1.80 (1.68 1.93) for interaction 5 0.0003) across strata Model 2 1.00 (reference) 1.21 (1.09–1.35) 1.20 (1.04–1.38) (Supplementary Table 3). Multiple test- Model 3 1.00 (reference) 1.19 (1.07–1.33) 1.11 (0.95–1.29) ing via false discovery rate analyses Model 4 1.00 (reference) 1.17 (1.05–1.31) 1.12 (0.96–1.30) and sensitivity analyses with further ad- Severely exposed areas 1.00 (reference) 1.20 (1.05–1.37) 1.14 (0.95–1.37) justment of area (rural/urban), marriage Less severely exposed areas 1.00 (reference) 1.07 (0.87–1.30) 1.05 (0.80–1.37) status, occupation, and economic status Model 1 was unadjusted; model 2 was adjusted for age and sex; and model 3 included model 2 plus showed similar results (Supplementary education attainment (less than high school or high school or above), drinking status (current Tables 4 and 5). In age-balanced anal- drinker or not), and family history of diabetes (yes or no). Model 4 included model 3 plus individual ysis, the risk estimates do not change CVH metrics. significantly (Supplementary Table 6). care.diabetesjournals.org Lu and Associates 5

Table 3—Multivariable-adjusted RRs (95% CIs) for incident T2DM according to famine exposure and combined ICVHMs Famine exposure Cumulative P for Case subjects/n incidence, % Nonexposed Fetal Childhood interaction Diet pattern 0.5392 Nonideal 2,219/33,651 6.59 1.00 (reference) 1.29 (1.10–1.52) 1.25 (1.00–1.56) Ideal 2,623/44,274 5.92 1.00 (reference) 1.07 (0.93–1.25) 1.01 (0.82–1.25) Physical activity 0.5766 Nonideal 4,181/66,912 6.25 1.00 (reference) 1.16 (1.03–1.30) 1.06 (0.90–1.25) Ideal 661/11,013 6.00 1.00 (reference) 1.25 (0.91–1.72) 1.52 (0.996–2.31) Smoking status 0.0621 Nonideal 1,110/15,836 7.01 1.00 (reference) 1.28 (1.02–1.59) 1.03 (0.75–1.42) Ideal 3,732/62,089 6.01 1.00 (reference) 1.14 (1.01–1.30) 1.14 (0.96–1.36) BMI 0.0018 Nonideal 3,307/40,399 8.19 1.00 (reference) 1.20 (1.05–1.36) 1.09 (0.91–1.31) Ideal 1,535/37,526 4.09 1.00 (reference) 1.10 (0.90–1.35) 1.17 (0.88–1.54) TC 0.4103 Nonideal 2,222/30,799 7.21 1.00 (reference) 1.15 (0.97–1.37) 1.17 (0.93–1.46) Ideal 2,620/47,126 5.56 1.00 (reference) 1.17 (1.01–1.35) 1.06 (0.86–1.30) BP 0.9710 Nonideal 3,920/53,046 7.39 1.00 (reference) 1.22 (1.08–1.38) 1.18 (1.00–1.40) Ideal 922/24,879 3.71 1.00 (reference) 0.97 (0.76–1.23) 0.87 (0.61–1.23) Number of ICVHMs 0.0005 #1 947/9,703 9.76 1.00 (reference) 1.59 (1.24–2.04) 1.28 (0.90–1.80) 2 1,553/19,978 7.77 1.00 (reference) 1.03 (0.85–1.26) 1.02 (0.78–1.33) 3 1,452/23,598 6.15 1.00 (reference) 1.00 (0.83–1.22) 1.01 (0.77–1.32) 4 658/16,579 3.97 1.00 (reference) 1.14 (0.84–1.55) 1.26 (0.82–1.93) $5 232/8,067 2.88 1.00 (reference) 1.09 (0.67–1.78) 0.79 (0.39–1.60) Adjusted for age, sex, education attainment (less than high school or high school or above), drinking status (current drinker or not), and family history of diabetes (yes or no). Individual CVH metrics were mutually adjusted.

The incidence of diabetes according to significant interactions have been de- Sex difference and the severity of famine exposure and the number of tected (P for interaction 5 0.0005). We famine exposure were further explored ICVHMs are displayed in Fig. 1. There was found the risk was significantly increased in stratified analysis (Fig. 2). We found an inverse relationship between the in participants with one or less ICVHMs that the association between famine number of ICVHMs and the incidence of (RR 1.59 [95% CI 1.24–2.04] for fetal- exposure and number of ICVHMs on the diabetes.Thehighestincidencewas found exposed), but not in those with two or risk of diabetes was present in both men in those with fetal famine exposure and more ICVHMs (Table 3). In sensitivity and women. The incremental risk of di- with no ICVHMs in adulthood. analyses and age-balanced analyses, the abetes associated with fetal famine ex- When analyzing the effect of the risk estimates do not change significantly posure was greatest among women with number of ICVHMs on the association (P for interaction 5 0.0003 and 0.0041, one or less ICVHMs (RR 1.66 [95% CI 1.11– between exposure to famine in early respectively) (Supplementary Tables 5 2.49]) and was lowest among men with life and the risk of incident diabetes, and 6). five or more ICVHMs (RR 0.64 [95% CI 0.18–2.22]) (Fig. 2A). Furthermore, the increased risk was significant in those with fetal exposure in severely affected areas and with one or less ICVHM (RR 1.52 [95% CI 1.14–2.02]), while in their counterparts, a marginal increased risk was observed (RR 1.61 [95% CI 0.96– 2.70]) (Fig. 2C). No significantly increased risk of T2DM was observed for partici- pants with childhood famine exposure in both sexes (Fig. 2B) and both famine- severity areas (Fig. 2D).

CONCLUSIONS In this large population-based study, we found that prenatal exposure to famine Figure 1—Cumulative incidence of diabetes according to famine exposure and the number of is associated with an increased risk of ICVHMs. T2DM in adulthood. More importantly, 6 Famine Exposure, ICVHMs, and Risk of Diabetes Diabetes Care

Figure 2—Multivariable-adjusted RRs (95% CIs) of incident diabetes for participants with famine exposure in relation to number of ICVHMs according to sex and area categories. A total of 77,925 participants (23,582 nonexposed, 13,195 fetal exposed, and 41,148 childhood exposed) were included in the analysis. The reference group is nonexposed individuals, with the same number of ICVHMs as the famine-exposed groups. RRs (95% CIs) were adjusted for age, sex, education attainment (less than high school or high school or greater), drinking status(current drinker or not), and family history of diabetes (yes or no). Interaction between the combination of famine exposure status with number of ICVHMs and sex on diabetes: P for interaction 5 0.863 (A) and P for interaction 5 0.484 (B). Interaction between the combination of famine exposure status with number of ICVHMs and level of affected area on diabetes: P for interaction 5 0.485 (C) and P for interaction 5 0.598 (D).

an interaction/effect modification be- the famine exposure and T2DM came life increased the risk of T2DM in adult- tween famine and number of ICVHMs from the 2002 China National Nutrition hood (hazard ratio 1.25 [95% CI 1.07– was observed, and this increased risk and Health Survey, which demonstrated 1.45]). However, most of these studies was diminished in individuals with two that fetal famine exposure in severely were limited, with a relatively small pop- or more ICVHMs. To our knowledge, this affected areas was associated with an ulation (28,32,33) or limited areas (32). is the first and largest epidemiological increased risk of hyperglycemia in adult- Our current study provided further study investigating the modifying effects hood (odds ratio 3.92 [95% CI 1.64–9.39]) supporting evidence that exposure to of ICVHMs on the association between (19). The effect of prenatal famine ex- famine in early life influences the risk of famine exposure and the risks of T2DM in posure was confirmed by other studies T2DM development later in life in a large adulthood. subsequently (12,14,28,32,33). Results prospective nationwide cohort. Partly Famine exposure in early life was from the Dongfeng-Tongji cohort (33) consistent with previous findings (12,19), shown to be related to the riskof T2DM in suggested that participants who were we found that individuals with fetal fam- epidemiologicalstudiespreviously(6,7,12). exposed to severe famine in childhood ine exposure appeared to have signifi- Findings from the Ukraine and Dutch had a 38% higher T2DM risk than those cantly increased risk of T2DM compared famines provide strong support for an exposed to less severe famine (odds ratio with nonexposed participants. Impor- association between famine exposure in 1.38 [95% CI 1.05–1.81]). The China Ka- tantly, only participants in severely af- early life and T2DM (6,7). In China, the doorie Biobank study (14) also con- fected areas with fetal famine exposure first evidence on the association between firmed that famine exposure in early have a higher risk of diabetes. care.diabetesjournals.org Lu and Associates 7

Previous epidemiologic studies on the programming hypothesis” suggested of the timing of diagnoses for diabetes. modifying factors and potential mecha- that the thrifty phenotype would reduce Fourth, we evaluated a diet score mainly nisms of famine exposure and diabetes b-cell function and is more prone to based on the information of fruit, vege- risk in later life are limited. Stratified develop T2DM under conditions of a table, soy protein, and level of caloric analysis from the 2002 China National sudden move to overnutrition. Our ob- intake, not including the sodium intake Nutrition and Health Survey has exam- servation supports an effect of a transform (40), which may underestimate the ac- ined dietary factors, economic status, lifestyle from starvation to overnutrition. tual effect of a healthy diet with T2DM. and BMI in 7,874 rural adults born be- Additionally, exposure to famine in early Fifth, the possibility of residual confound- tween 1954 and 1964 and concluded that life increases the susceptibility to chronic ing due to unmeasured or poorly mea- the increased risk of hyperglycemia in diseases in adulthood potentially through a sured confounders such as maternal adulthood related to fetal exposure to memory of the effects of early nutritional health and maternal child-feeding be- the severe Chinese famine appears to be environments, which was also called “met- haviors could not be ruled out. Finally, exacerbated by a nutritionally rich en- abolic imprinting” (39). although age-balanced analysis has been vironment in later life (19). The China It is worth mentioning that aging effect recommended by previous studies, there Kadoorie Biobank study (14) reported is an unavoidable issue for the analysis might exist limitations of combining the that coexistence of prenatal experience between famine exposure and health out- age group. Famine exposure has been of undernutrition and abdominal obe- comes. The age gap may account for dif- shown to be associated with cardiome- sity in adulthood was associated with a ference in the physical activity, employment, tabolic risk and adult mortality; thus, sur- higher risk of T2DM. In the current study, residence (urban/rural), and economic sta- vival bias might be possible. for the first time, we examined the effect tus. As the incidence of diabetes is highly In conclusion, we found that famine of ICVHMs and the number of these correlated with aging, this age difference exposure in early life significantly in- metrics on the association between fam- between individuals born during the fam- creased risk of T2DM in later life. This ine exposure in early life and the risk of ine and postfamine control individuals association could be attenuated by two T2DM in a large nationwide Chinese co- can introduce substantial bias in analysis. or more ICVHMs in adulthood, and an hort. Interestingly, we found that the To overcome this issue, we applied age- interaction/effect modification between increased risk of diabetes due to famine balanced analysis and stratified analysis famine and number of ICVHMs was ob- exposure might bemodified by a healthy with the severityleveloffamineexposure, served. Our findings emphasize the im- lifestyle or metabolic metrics (such as as proposed by the previous work (7) and portance of a healthy lifestyle in adulthood ideal BMI) in adulthood. These findings recent re-examination of available studies in prevention of T2DM even in presence imply the importance of a healthy life- (13,31). We demonstrated that fetal fam- of the adverse prenatal or early life style in the prevention of T2DM among ine exposure was associated with increased factors. individuals who experienced the fetal diabetes risk using age-balanced analy- undernutrition (5). In addition, unlike sis. This association was seen in severely the heterogeneity of men and women affected areas. reported previously (28,34), our study The strengths of this study include a Acknowledgments. Theauthorsthankallof the found that fetal famine exposure was nationwide prospective cohort design, study participants. associated with increased risk of diabe- the large sample size, and the detailed Funding. The research reported in this publication was supported by the National Basic Research Pro- tes in both male and female. Further- information about lifestyle factors. The gram of China (973 Program) (award 2015CB553601), more, fetal famine exposure significantly diagnosis of diabetes was not self- the Ministry of Science and Technology of the increased the risk of diabetes only in reported, but based on oral glucose tol- People’s Republic of China (awards 2016YFC1305600, severely affected areas but not less se- erance test at both baseline and follow-up 2016YFC1305202, 2016YFC1304904, and verelyaffectedareas,whichreinforcesour visit. Our study does have a number of 2017YFC1310700), the National Natural Science Foundation of China (awards 81700764, 81670795, conclusions on famine exposure and di- important limitations. First, the Chinese 81621061, and 81561128019), National Major abetes risk. famine did not have a definite beginning Scientific and Technological Special Project There are several potential mecha- or ending time, making it difficult to for “Significant New Drugs Development” (award nisms underlying the association of famine precisely define the famine exposure. 2017ZX09304007), and the Innovative Re- exposure in early life and diabetes in Misclassification of famine exposure was search Team of High-Level Local Universities in Shanghai. adulthood. One is the developmental inevitable. However, using birth date to Duality of Interest. No potential conflicts of origin hypothesis (i.e., early nutrition define famine exposure was the most interest relevant to this article were reported. status influences the epigenetic changes) common method in studies on the Chi- Author Contributions. J.L., Y.B., G.N., and W.W. (35,36). Epigenetic dysregulation was re- nese famine (11,19). Second, the study conceived and designed the study. J.L., M.L., and ported in diabetic islets in a study of participants were only followed for a R.D. analyzed data. Y.Q., Q.L., R.H., L.S., Q.S., Z.Z., comprehensive DNA methylation in di- mean of 3.6 years. This relatively short X.Y.,L.Y.,G.Q.,Q.W.,G.C.,Z.G.,G.W.,F.S.,Z.L.,LiC., Y.H., Z.Y., X.T., Y.Z., C.L.,Y.W.,S.W.,T.Y.,H.D.,Lu.C., abetic and nondiabetic pancreatic islets follow-up duration reduced the number J.Z., and Y.M. collected data. All authors were (37). Furthermore, differential DNA methyl- of incident diabetes and the study’s sta- involved in writing and revising the paper and had ations were reported in those exposed to tistical power. Third, the study partic- final approval of the submitted and published Dutch famine, which suggested that pre- ipants only had one follow-up visit, and versions. Y.B., G.N., and W.W. are the guarantors ofthisworkand,assuch,hadfullaccesstoallofthe natal starvation might promote an ad- glycemic measures were obtained at only data in the study and take responsibility for the verse metabolic phenotype in later life by two time points (the baseline and fol- integrity of the data and the accuracy of the data epigenetic modulation (38). The “fetal low-upvisits).Thiscouldlimittheaccuracy analysis. 8 Famine Exposure, ICVHMs, and Risk of Diabetes Diabetes Care

References 15. Zhang N, Du SM, Ma GS. Current lifestyle 28. Wang N, Wang X, Han B, et al. Is exposure to 1. Gluckman PD, Hanson MA, Bateson P, et al. factors that increase risk of T2DM in China. Eur J famine in childhood and economic development Towards a new developmental synthesis: adap- Clin Nutr 2017;71:832–838 in adulthood associated with diabetes? J Clin tive developmental plasticity and human disease. 16. Li Y, Wang DD, Ley SH, et al. Time trends of Endocrinol Metab 2015;100:4514–4523 Lancet 2009;373:1654–1657 dietary and lifestyle factors and their potential 29. Luo ZMR, Zhang X. Famine and overweight in 2. BerendsLM, OzanneSE.Early determinantsof impact on diabetes burden in China. Diabetes China. Rev Agric Econ 2006;28:296–304 type-2 diabetes. Best Pract Res Clin Endocrinol Care 2017;40:1685–1694 30. McNutt LA, Wu C, Xue X, Hafner JP. Esti- Metab 2012;26:569–580 17. Lv J, Yu C, Guo Y, et al.; China Kadoorie mating the relative risk in cohort studies and 3. Ruiz-Narvaez´ EA, Palmer JR, Gerlovin H, et al. Biobank Collaborative Group. Adherence to a clinical trials of common outcomes. Am J Epi- Birth weight and risk of type 2 diabetes in the healthy lifestyle and the risk of type 2 diabetes in demiol 2003;157:940–943 Black Women’s Health Study: does adult BMI Chinese adults. Int J Epidemiol 2017;46:1410– 31. Li C, Tobi EW, Heijmans BT, Lumey LH. The play a mediating role? Diabetes Care 2014;37: 1420 effect of the Chinese Famine on type 2 diabetes 2572–2578 18. Li Y, Ley SH, Tobias DK, et al. Birth weight and mellitus epidemics. Nat Rev Endocrinol 2019;15: 4. Xiao X, Zhang ZX, Cohen HJ, et al. Evidence of a later life adherence to unhealthy lifestyles in 313–314 relationship between infant birth weight and predicting type 2 diabetes: prospective cohort 32. Li J, Liu S, Li S, et al. Prenatal exposure to later diabetes and impaired glucose regulation study. BMJ 2015;351:h3672 famine and the development of hyperglycemia in a Chinese population. Diabetes Care 2008;31: 19. Li Y, He Y, Qi L, et al. Exposure to the Chinese and type 2 diabetes in adulthood across con- 483–487 famine in early life and the risk of hyperglycemia secutive generations: a population-based cohort 5. Hu C, Mu Y, Wan Q, et al.; REACTION Study and type 2 diabetes in adulthood. Diabetes 2010; study of families in Suihua, China. Am J Clin Nutr Group. Association between birth weight and 59:2400–2406 2017;105:221–227 diabetes: role of body mass index and lifestyle in 20. Lloyd-Jones DM, Hong Y, Labarthe D, et al.; 33. Wang J, Li Y, Han X, et al. Exposure to the later life. J Diabetes 2020;12:10–20 American Heart Association Strategic Planning Chinese famine in childhood increases type 2 6. Ravelli AC, van der Meulen JH, Michels RP, Task Force and Statistics Committee. Defining diabetes risk in adults. J Nutr 2016;146:2289– et al. Glucose tolerance in adults after prenatal and setting national goals for cardiovascular health 2295 exposure to famine. Lancet 1998;351:173–177 promotion and disease reduction: the American 34. Sun Y, Zhang L, Duan W, Meng X, Jia C. 7. Lumey LH, Khalangot MD, Vaiserman AM. Heart Association’s strategic Impact Goal through Association between famine exposure in early Association between type 2 diabetes and pre- 2020 and beyond. Circulation 2010;121:586–613 life and type 2 diabetes mellitus and hypergly- natal exposure to the Ukraine famine of 1932-33: 21. Joseph JJ, Echouffo-Tcheugui JB, Carnethon cemia in adulthood: results from the China a retrospective cohort study. Lancet Diabetes MR, et al. The association of ideal cardiovascular Health And Retirement Longitudinal Study (CHARLS). Endocrinol 2015;3:787–794 health with incident type 2 diabetes mellitus: the J Diabetes 2018;10:724–733 8. Smil V. China’s great famine: 40 years later. Multi-Ethnic Study of Atherosclerosis. Diabeto- 35. Gluckman PD, Hanson MA, Cooper C, Thornburg BMJ 1999;319:1619–1621 logia 2016;59:1893–1903 KL. Effect of in utero and early-life conditions on 9. Liu D, Yu DM, Zhao LY, et al. Exposure to 22. Fretts AM, Howard BV, McKnight B, et al. adult health and disease. N Engl J Med 2008;359: famine during early life and abdominal obesity in Life’s Simple 7 and incidence of diabetes among 61–73 adulthood: findings from the great Chinese fam- American Indians: the Strong Heart Family Study. 36. Uauy R, Kain J, Corvalan C. How can the ine during 1959-1961. Nutrients. 2019;11:903 Diabetes Care 2014;37:2240–2245 Developmental Origins of Health and Disease 10. Liu L, Pang ZC, Sun JP, et al. Exposure to 23. Climie RE, van Sloten TT, Perier´ MC, et al. (DOHaD) hypothesis contribute to improving famine in early life and the risk of obesity in Change in cardiovascular health and incident health in developing countries? Am J Clin Nutr adulthood in Qingdao: evidence from the 1959- type 2 diabetes and impaired fasting glucose: 2011;94(Suppl.):1759S–1764S 1961 Chinese famine. Nutr Metab Cardiovasc Dis the Whitehall II study. Diabetes Care 2019;42: 37. Volkmar M, Dedeurwaerder S, Cunha DA, 2017;27:154–160 1981–1987 et al. DNA methylation profiling identifies epi- 11. Li Y, Jaddoe VW, Qi L, et al. Exposure to the 24. Lu J, He J, Li M, et al.; 4C Study Group. genetic dysregulation in pancreatic islets from chinese famine in early life and the risk of met- Predictive value of fasting glucose, postload type 2 diabetic patients. EMBO J 2012;31:1405– abolic syndrome in adulthood. Diabetes Care 2011; glucose, and hemoglobin A1c on risk of diabetes 1426 34:1014–1018 and complications in Chinese adults. Diabetes 38. Tobi EW, Goeman JJ, Monajemi R, et al. DNA 12. Wang N, Cheng J, Han B, et al. Exposure to Care 2019;42:1539–1548 methylation signatures link prenatal famine ex- severe famine in the prenatal or postnatal period 25. Wang T, Lu J, Su Q, et al.; 4C Study Group. posure to growth and metabolism [published and the development of diabetes in adulthood: Ideal cardiovascular health metrics and major correction appears in Nat Commun 2015;6:7740]. an observational study. Diabetologia 2017;60: cardiovascular events in patients with prediabe- Nat Commun 2014;5:5592 262–269 tes and diabetes. JAMA Cardiol. 2019;4:874–883 39. Waterland RA, Garza C. Potential mecha- 13. Li C, Lumey LH. Exposure to the Chinese 26. Lu J, Wang W, Li M, et al. Associations of nisms of metabolic imprinting that lead to famine of 1959-61 in early life and long-term hemoglobin A1c with cardiovascular disease and chronic disease. Am J Clin Nutr 1999;69:179– health conditions: a systematic review and meta- mortality in Chinese adults with diabetes. J Am 197 analysis. Int J Epidemiol 2017;46:1157–1170 Coll Cardiol 2018;72:3224–3225 40. Bi Y, Jiang Y, He J, et al.; 2010 China 14. Meng R, Lv J, Yu C, et al.; China Kadoorie 27. Craig CL, Marshall AL, Sjostr¨ om¨ M, et al. Noncommunicable Disease Surveillance Group. Biobank Collaborative Group. Prenatal famine International physical activity questionnaire: 12- Status of cardiovascular health in Chinese exposure, adulthood obesity patterns and risk of country reliability and validity. Med Sci Sports adults. J Am Coll Cardiol 2015;65:1013– type 2 diabetes. Int J Epidemiol 2018;47:399–408 Exerc 2003;35:1381–1395 1025