The Kailuan Study (trial registration number ChiCTR-TNC-11001489) is an ongoing longitudinal prospective study conducted in the Kailuan community in Tangshan, Republic of China. A total of 101,510 employees (including the retired, 81,110 men and 20,400 women, mean age = 51.9 y in 2006) of the Kailuan Group were invited and agreed to participate in the Kailuan Study between June 2006 and October 2007 (referred to as the “baseline survey”). Detailed study design and procedures have been described in previous studies (13, 14). All participants completed questionnaire assessments, clinical examinations, and laboratory tests in 11 hospitals responsible for the health care of the Kailuan general hospital. Participants then took physical examination every 2 years, and the incidences of chronic diseases (e.g., stroke, MI, cancer, et al.) were recorded annually. In the current study, data analysis was performed from baseline survey to December 31, 2019.

In the present study, according to previous Chinese famine studies (4, 12), 65,105 participants born between January 1, 1949, and December 31, 1974, were recruited from the Kailuan Study. Participants (n = 1,060) who had a history of MI or stroke at the baseline survey were excluded. Moreover, we excluded participants (n = 2,518) who had incomplete data on health factors or behaviors, leaving 61,527 participants (47,549 men and 13,978 women) available for analyses (Supplementary Figure S1).

We collected information on sociodemographic characteristics (e.g., age, gender, birth date, and education level), lifestyle factors (e.g., smoking, alcohol intake, salt intake, and physical activity), medical history (e.g., CVD, hypertension, diabetes mellitus, family history of CVD), and active treatments such as hypoglycemic, antihypertensive, and lipid-lowering medications through the self-reported questionnaire in the Kailuan Study since the baseline survey, as detailed previously (13). Education was classified as less than high school and high school or above. Drinking status was stratified into two levels: current or never/former. Weight and height were measured, and BMI was calculated as weight (kg)/height (m)². Moreover, BP was measured on the left arm using an appropriate cuff size after the participant had a rest in a chair for at least 5 min. The average values of at least two readings each of systolic and diastolic BPs were used for further analysis.

The blood sample of each participant was collected on the morning of the survey after at least a 12-h fast. All samples were measured by a Hitachi 747 autoanalyzer at the central laboratory of the Kailuan General Hospital. The estimated glomerular filtration rate (eGFR) was calculated using the Chronic Kidney Disease Epidemiology Collaboration creatinine equation (15).

Because the Chinese Famine occurred from 1959 to 1962, consistent with previous Chinese famine studies (4, 12, 16, 17), birth year was taken as the proxy variable of exposure to famine in this study. We defined those born between January 1, 1959, and December 31, 1962, as fetal exposure group, those born between January 1, 1949, and December 31, 1958, as childhood-exposed group, and those born between January 1, 1963, and December 31, 1974, as a nonexposed group. As a large enterprise group, the employees in the Kailuan group were nationwide. Therefore, the severity of the famine was determined according to the excess death rate for each province, which was calculated as this rate change from the average level in 1956–1958 to the highest value in 1959–1962 (7). Based on the information, an excess mortality rate of 50% was used as a threshold value to define severely and less severely affected areas in the present study.

ICVHMs were adapted from the recommendations of American Heart Association (8): nonsmoking, BMI <24 kg/m², physical activity at goal (at least 80 min per week of moderate-intensity physical activity), ideal diet (daily salt intake <6 g) (18); TC <5.2 mmol/L (untreated), BP <120/80 mmHg (untreated), FBG <5.6 mmol/L (untreated). Cardiovascular health status was then categorized as high (5–7 metrics), moderate (3–4 metrics), or low (0–2 metrics) levels (19).

The main outcome of this study was the first occurrence of CVD events, including MI, severe coronary artery disease, intracerebral hemorrhagic stroke, and ischemic stroke. CVD events were defined as described (20, 21). Briefly, the Municipal Social Insurance Institution database and Hospital Discharge Register were linked to identify the incidence of CVD based on The International Statistical Classification of Diseases and Related Health Problems 10th Revision (ICD-10) (I61 for intracerebral hemorrhagic stroke, I63 for ischemic stroke, and I21 for MI, I25.1 for coronary heart disease). These two databases covered all the Kailuan Study participants. For all suspected CVD cases, a panel of three experienced physicians, consisting of neurologists, cardiologists, and radiologists, reviewed the medical records, blind to exposure status. We defined coronary artery disease and MI cases according to the Multinational Monitoring of Trends and Determinants in CVD criteria of WHO on basis of clinical symptoms and dynamic changes in cardiac enzymes and/or biomarker concentrations and electrocardiogram results. Severe coronary artery disease was defined as coronary heart disease treated with coronary bypass surgery or stent placement. Incident stroke was diagnosed according to the WHO criteria, based on symptoms, neuroimages (from CT or MRI), and other diagnostic reports. Mortality data were collected from provincial vital statistics offices, as described previously (14).

All analyses were performed using SAS, version 9.4 (SAS Institute, Inc, Cary, NC). Two-sided values of P < 0.05 were regarded as significant.

The baseline characteristics of the study population were described by famine exposure status. Continuous variables with normal distribution were expressed as means ± SDs and compared using one-way ANOVA analysis, while those with skewed distribution were expressed as medians and interquartile range and compared by Kruskal-Wallis test. Categorical variables were shown in proportions and compared by Pearson's Chi-Square test.

We computed the person-year of follow-up for each participant from the date of the baseline survey to the date of the CVD onset, death, loss to follow-up (n=2,156, 3.5%), or the end of follow-up (December 31, 2019), whichever came first. The Cox regression model was used to predict the risk of CVD and to estimate the hazard ratio (HR) and 95% CI, after adjustments for age, sex, education attainment (less than high school, high school, or above), drinking (current, never/former), eGFR (<30, 30 ≤ eGFR <60, or ≥ 60 mL/min/1.73 m²), and high-sensitivity C-reactive protein (Hs-CRP) (<1.0, 1.0 ≤ Hs-CRP ≤ 3.0, or > 3.0 mg/L), triglycerides (TG) (<1.7, 1.7 ≤ TG <2.3, or ≥ 2.3 mmol/L), severity famine exposed, family history of CVD, using of antihypertensive, hypoglycemic, lipid-lowering medications (yes/no for each), and individual ICVHMs. Cochran-Armitage trend test was used to investigate the association between the increasing number of ICVHMs and the cumulative incidence of CVD. Missing covariates (drinking, eGFR, Hs-CRP, TG) will be imputed by means of multiple imputations using the Fully Conditional Specification method computing 10 imputed datasets and to prevent case-wise deletion of missing data; 60,874 (98.9%) participants had complete data for all covariates.

To explore the modifying effect of ICVHMs, we evaluated the association between famine exposure and CVD by strata of the individual ICVHMs items and the number of ICVHMs. Other ICVHMs items were adjusted when evaluating individual ICVHMs items. P-value was corrected for multiple testing via false discovery rate using the Benjamini-Hochberg method. To demonstrate potential interactions of famine exposure and ICVHMs on the development of CVD, we generated interaction terms using the cross products of famine exposure and ICVHMs in Cox regression models. To address specific questions on age differences between famine and postfamine births, an “age-balanced” method was used, in which both prefamine and postfamine births were combined as unexposed control subjects, as detailed previously (12, 22).

We further conducted several sensitivity analyses. To assess the impact of imputing missing covariates, we repeated the main analyses in those without imputation. To test the robustness of our finding, we repeated the main analyses excluding those with antihypertensive, hypoglycemic, or lipid-lowering medications treatment, excluding those lost to follow-up, and excluding those who had CVD events or death within the first 2 years of follow-up.

A total of 61,527 participants (mean age 47.9 ± 6.6 y) were recruited in the present analyses, including 47,549 (77.3%) men and 13,978 (22.7%) women. There were 13.9% (n = 8,572), who were exposed to the Great Chinese Famine in the fetal stage. Compared with the nonexposed group (born between January 1, 1963, and December 31, 1974), participants with fetal famine exposure were more likely to be with a family history of CVD, or use antihypertensives or hypoglycemics, had higher BMI, TC, TG, systolic or diastolic BPs, FBG level, Hs-CRP level, had a greater proportion of individuals with ideal physical activity, as well as lower eGFR, education level, less proportion of individuals with other six individual ideal items, and less drinkers (Table 1). Supplementary Table S1 showed baseline information for the age-balanced group with combining of prefamine and postfamine births as the reference group.

During a median of 13.0 (12.7–13.2) years follow-up, 4,814 incident CVD events were identified. Age of CVD onset was 49.5±4.3, 54.2±3.8, 60.8±4.4 for the nonexposed, fetal-exposed, or childhood-exposed group, respectively (Supplementary Table S2). As shown in Table 2, both the incidences of CVD for fetal famine-exposed (5.95/1000 pys) or childhood famine-exposed (8.31/1000 pys) group were greater than that for the nonexposed group (3.10/1000 pys). After adjusting for covariates, the increased CVD risk remained significant in fetal famine-exposed participants (HR: 1.21; 95% CI: 1.07–1.37), but not in childhood famine-exposed participants (HR: 1.00; 95% CI: 0.85–1.17). Furthermore, in the sex stratified analysis (Table 2) and age-balanced analysis (Supplementary Table S3), the results were similar.

The incidence of CVD according to famine exposure and the number of ICVHMs were displayed in Figure 1. There was a reverse correlation between the number of ICVHMs and the incidence of CVD. With the increasing number of ICVHMs, the cumulative incidence of CVD decreased gradually for participants in nonexposed or fetal-exposed groups (all P for trend < 0.001).

After stratifying participants by the number of ICVHMs, we found that in participants with less ICVHMs ( ≤ 2), fetal famine exposure significantly increased the CVD risk after adjusting for all covariates (HR: 1.30; 95% CI: 1.11–1.52, P for interaction=0.008), comparing with a nonexposed group (Table 3). However, we do not observe the association in participants with three or more ICVHMs (HR: 1.14; 95% CI: 0.92–1.40, HR: 0.89; CI: 0.37–2.11). In sensitivity analyses, consistent results were observed without imputing covariates, excluding participants using antihypertensive, hypoglycemic, or lipid-lowering medications, participants lost to follow-up, and who had events or death within 2 years of follow-up (Table 3). Similar results were observed in the age-balanced analysis (Supplementary Table S4).

The sex difference was further explored in stratified analysis. Due to the small sample size of participants with ≥5 ICVHMs items in men or women, we combined and divided men or women into two groups according to the number of ICVHMs: participants with less ICVHMs ( ≤ 2) items or more ICVHMs (≥3) items. Similar results were observed in men, that the association between fetal famine exposure and CVD risk was significant in men with less ICVHMs (HR: 1.26, 95% CI: 1.07–1.49) rather than those with three or more ICVHMs (HR: 1.05, 95% CI: 0.84–1.31). However, results were inverse in women, that the association between fetal famine exposure and CVD risk was significant in women with three or more ICVHMs (HR: 1.65, 95% CI: 1.01–2.69) rather than those with less ICVHMs (HR: 1.58, 95% CI: 0.92–2.70). The interaction between sex and famine exposure was significant (P = 0.031) (Table 4).

We further analyzed the association between famine exposure and CVD risk according to individual ICVHMs items. The associations between famine exposure and CVD risk were observed in participants with nonideal ICVHMs items, except for TC (in ideal TC) and smoking items (in both ideal and nonideal) (Table 5). Multiple testing via false discovery rate analyses showed similar results (Supplementary Table S5). In age-balanced analysis, the risk estimates did not change dramatically (Supplementary Table S6).