Cox regression models were stratified by 5 year groups and 10 survey sites, with age as the underlying time scale.

Multivariable models were adjusted for sex (men or women); level of education (no formal school, primary school, middle school, high school, college or university or higher); occupation (agricultural, industrial, administrative or managerial, professional or technical, sales or service, retired, house wife or husband, self-employed, unemployed, other); household income (<2500, 2500 to 4999, 5000 to 9999, 10 000 to 19 999, 20 000 to 34 999, ≥35 000 RMB per year); marital status (married, widowed, divorced or separated or never married); family history of certain diseases (presence, absence or unclear; for corresponding disease only; no adjustment for heart failure, COPD and CKD); smoking status (non-smoker, former smoker, current smoker who smoked <15, 15 to 24 or ≥25 cigarettes or equivalents per day; those who had stopped smoking because of ill health were included in current smokers to avoid bias); total physical activity level (MET-hours/day); alcohol consumption (non-drinker, former weekly drinker, weekly drinker, daily drinker with an intake of <15, 15 to 29, 30 to 59 or ≥60 g/day of pure alcohol); intake frequency of fresh vegetables, fresh fruit and red meat (daily, 4 to 6 days/week, 1 to 3 days/week, monthly or rarely or never); BMI (kg/m2); waist circumference (cm); prevalent hypertension and diabetes at baseline (presence or absence, no adjustment for T2DM).

*Assessed by assigning consecutive integers to four BMF categories in a separate model. A Bonferroni corrected threshold was used, α=0.00625.

BMF, bowel movement frequency;BMI, body mass index; CKD, chronic kidney disease; COPD, chronic obstructive pulmonary disease; IHD, ischaemic heart disease; MCEs, major coronary events;PYs, person-years; T2DM, type 2 diabetes mellitus.