Lifestyle factors and risk of multimorbidity of cancer and cardiometabolic diseases: a multinational cohort study

Background Although lifestyle factors have been studied in relation to individual non-communicable diseases (NCDs), their association with development of a subsequent NCD, defined as multimorbidity, has been scarcely investigated. The aim of this study was to investigate associations between five lifestyle factors and incident multimorbidity of cancer and cardiometabolic diseases. Methods In this prospective cohort study, 291,778 participants (64% women) from seven European countries, mostly aged 43 to 58 years and free of cancer, cardiovascular disease (CVD), and type 2 diabetes (T2D) at recruitment, were included. Incident multimorbidity of cancer and cardiometabolic diseases was defined as developing subsequently two diseases including first cancer at any site, CVD, and T2D in an individual. Multi-state modelling based on Cox regression was used to compute hazard ratios (HR) and 95% confidence intervals (95% CI) of developing cancer, CVD, or T2D, and subsequent transitions to multimorbidity, in relation to body mass index (BMI), smoking status, alcohol intake, physical activity, adherence to the Mediterranean diet, and their combination as a healthy lifestyle index (HLI) score. Cumulative incidence functions (CIFs) were estimated to compute 10-year absolute risks for transitions from healthy to cancer at any site, CVD (both fatal and non-fatal), or T2D, and to subsequent multimorbidity after each of the three NCDs. Results During a median follow-up of 11 years, 1910 men and 1334 women developed multimorbidity of cancer and cardiometabolic diseases. A higher HLI, reflecting healthy lifestyles, was strongly inversely associated with multimorbidity, with hazard ratios per 3-unit increment of 0.75 (95% CI, 0.71 to 0.81), 0.84 (0.79 to 0.90), and 0.82 (0.77 to 0.88) after cancer, CVD, and T2D, respectively. After T2D, the 10-year absolute risks of multimorbidity were 40% and 25% for men and women, respectively, with unhealthy lifestyle, and 30% and 18% for men and women with healthy lifestyles. Conclusion Pre-diagnostic healthy lifestyle behaviours were strongly inversely associated with the risk of cancer and cardiometabolic diseases, and with the prognosis of these diseases by reducing risk of multimorbidity.

: Scoring of the healthy lifestyle index (HLI) and its simplified version     Figure S1. Flowchart of exclusion steps in the EPIC (European Prospective Investigation into Cancer and Nutrition) multimorbidity data. Figure S2. Transitions from baseline to cancer, cardiovascular disease (CVD), type-2 diabetes (T2D) and specific multimorbidity patterns within the EPIC study State-specific number of events are in boxes, and transition-specific number of events and incidence rates per 1000 person-years are reported on arrows. Figure S3. Subgroup analysis of associations of a 3-unit increment (~1 standard deviation) in the healthy lifestyle index (HLI) with first cancer at any site, cardiovascular disease (CVD), type-2 diabetes (T2D), and transitions to cancer-cardiometabolic multimorbidity. Figure S4. Cumulative incidence functions to develop first cancer at any site, cardiovascular disease (CVD), type-2 diabetes (T2D), and subsequent cancer-cardiometabolic multimorbodity (MM) for 55 years old men (dotted) and women (continuous) for values of the healthy lifestyle index (HLI) of 15 (healthy, 85 th percentile in green) and 5 (unhealthy, 4 th percentile in red); the HLI ranges from 0 to 20 units, with greater scores reflecting healthy lifesyles.
Cancer refers to first malignant tumors at any site excl. non-melanoma skin cancer. Deaths were censored and not modelled as a separate outcome.
The model was stratified for centre, sex, and adjusted for education level (no schooling, primary, secondary and university or more), height (continuous), binary indicator of alcohol use (no/yes), total energy intake (kcal/day), and use of hormones and menopausal status in women. Figure S5. Sensitivity analysis excluding in turn one of the components of the healthy lifestyle index (HLI) of associations of a 3-unit increment (~1 standard deviation) in the HLI with first cancer at any site, cardiovascular disease (CVD), type-2 diabetes (T2D), and transitions to cancer-cardiometabolic multimorbidity. Figure S6. Sensitivity analysis comparing associations of 1 standard deviation (SD) increments in [A] the healthy lifestyle index (HLI), and [B] a simplified HLI, with first cancer at any site, cardiovascular disease (CVD), type-2 diabetes (T2D), and transitions to cancer-cardiometabolic multimorbidity.
9 Figure S7. Cumulative incidence functions to develop first cancer at any site, cardiovascular disease (CVD), type-2 diabetes (T2D), and subsequent cancer-cardiometabolic multimorbodity (MM) for 65 years old men (dotted) and women (continuous) for values of the simplified healthy lifestyle index (sHLI) of 4 (four healthy lifestyle habits, 90 th percentile in green) and 1 (unhealthy [one healthy lifestyle habit], 5 th percentile in red); the sHLI ranges from 0 to 5 units.
Cancer refers to first malignant tumors at any site excl. non-melanoma skin cancer. Deaths were censored and not modelled as a separate outcome.
The model was stratified for centre, sex, and adjusted for education level (no schooling, primary, secondary and university or more), height (continuous), binary indicator of alcohol use (no/yes), total energy intake (kcal/day), and use of hormones and menopausal status in women Figure S8. Cumulative incidence functions to develop cancer-cardiometabolic multimorbodity (MM) for 65 years old men (dotted) and women (continuous) after a less and more survivable cancer*, and after cardiovascular disease (CVD), and type-2 diabetes (T2D) for values of the healthy lifestyle index (HLI) of 15 (healthy, 85 th percentile in green) and 5 (unhealthy, 4 th percentile in red); the HLI ranges from 0 to 20 units, with greater scores reflecting healthy lifesyles.
*Cancer refers to first malignant tumors at any site excl. non-melanoma skin cancer, which were grouped according to 5-year relative survival rates into two distinct groups: (1) "less survivable cancer" with 5-year relative survival rates of less than 40% included pancreas, gallbladder, liver, lung, oesophagus, glioma, and stomach; (2) "more survivable cancer" with 5-year relative survival rates of equal or higher than 40% included ovarian, multiple myeloma, leukemia, colorectum, head and neck, non-Hodgkin's lymphoma, kidney, meningioma, bladder, endometrium, breast, thyroid, and prostate.
Deaths were censored and not modelled as a separate outcome.
The model was stratified for centre, sex, and adjusted for education level (no schooling, primary, secondary and university or more), height (continuous), binary indicator of alcohol use (no/yes), total energy intake (kcal/day), and use of hormones and menopausal status in women