Table 2. Methods/results for influence of SES on association between combinations of unhealthy LFs and outcomes.

Study	Methods	Covariates (n)	Interaction tests between combinations of unhealthy LFs and SES	Main interaction results (P interaction or RERI)	Result summary
Andersen 2016 42	1) Cox-proportional hazard models for all-cause mortality for combination of unhealthy LF categories 2) Models stratified by low/high income in sub-group analysis	Enrolment source, education, marital status, neighbourhood deprivation, and BMI (5)	Likelihood ratio tests, comparing main effects models with and without cross-product terms	All-cause mortality: 0.002 (African American men); 0.89 (African American women); 0.04 (White men); 0.49 (White women)	Significant multiplicative interaction for African American and White men only: highest HRs for combination of unhealthy LFs and high income Only stratified (sex, ethnicity) results available
Eguchi 2017 43	1) Cox-proportional hazard models for outcomes for combinations of unhealthy LF categories, stratified by low/high education (analyses for total population and separate analyses further stratified by sex) 2) Cox-proportional hazard models for combinations of unhealthy LF categories and education level using single reference group (all-cause and CVD mortality only) 3) Kaplan-Meier survival curves for combinations of unhealthy LF categories, stratified by low/high education (all-cause and CVD mortality only) 4) Sensitivity analysis examining two modified LF combinations	Age, sex, history of hypertension, history of diabetes, perceived mental stress and regular employment (6)	Cross-product of dichotomous education level and healthy lifestyle score (continuous) in models for total CVD and all-cause mortality outcomes only	All-cause mortality: 0.11 CVD mortality: 0.23 (both for total population only)	1) No evidence of multiplicative interaction, with similar HRs for combinations of unhealthy LFs and both high and low SES 2) Single reference group analysis provides evidence for additive interaction for all-cause and CVD mortality: higher HRs in least healthy combination of LFs and lowest education groups 3) Survival curves suggest additive interaction: steeper curve (highest mortality) for combination of unhealthy LFs in low education group 4) Sensitivity analysis: i) extended definition of healthy sleep and ii) dichotomous diet score (five components) in addition to extended sleep definition - consistent with findings from main analysis
Andersen 2018 44	1) Cox-proportional hazard models for all-cause mortality for combinations of unhealthy LF categories, stratified by NDI quartiles 2) Cox-proportional hazard models for all-cause mortality for combinations of unhealthy LF categories, stratified by NDI quartiles using single reference (also stratified by sex)	Enrolment source, ethnicity, education, income, marital status, and insurance status (6)	Likelihood ratio tests, comparing main effects models with and without cross-product terms	All-cause mortality: 0.28 (men); 0.99 (women)	1) No evidence of multiplicative interaction with similar HRs for combinations of unhealthy LFs in both high and low SES 2) Single reference group analysis provides evidence for additive interaction in men and women for all-cause mortality: highest HRs in the least healthy combination of LFs and lowest SES (highest NDI) group
Foster 2018 19	1) Cox-proportional hazard models for outcomes for combinations of unhealthy LF categories, stratified by SES quintiles (TDI, income, and education examined separately) 2) Cox-proportional hazard models for joint associations of combinations of unhealthy LF categories and SES measures (single reference group)	Age, sex, ethnicity, month of assessment, hypertension, systolic blood pressure, medication for hypercholesterolaemia or hypertension, and BMI (8)	1) Interaction term between combinations of unhealthy LFs and SES variables in models 2) Interaction sensitivity analyses (deprivation index): a) additional models with interaction term and i) dichotomous and ii) continuous combination of unhealthy LF variable b) Estimation of three measures of ‘biological interaction’: RERI, AP, and synergy index	Deprivation index All-cause and CVD- mortality: <0.0001 CVD incidence: 0.11 Income All-cause mortality: 0.001 CVD mortality: <0.0001 CVD incidence: 0.009 Education All-cause mortality: 0.002 CVD mortality: 0.047 CVD incidence: 0.051 (all for total population only)	1) Significant multiplicative interaction between combination of unhealthy LFs and deprivation/education for all-cause and CVD mortality but not for CVD incidence. Significant multiplicative interaction between combinations of unhealthy LFs and income for all outcomes 2) Single reference analysis showed highest HRs for all-cause and CVD mortality in the least healthy combination of LFs and lowest SES groups 3) Interaction sensitivity results consistent with main findings with significant interaction across three measures of additive interaction
Choi 2022 45	1) Cox-proportional hazard models for all-cause mortality for number of unhealthy LFs, stratified by high/low income group	Age, age squared, sex, education, race/ethnicity, acculturation, income assistance, health insurance, and marital status (9)	Unclear, but likely an interaction term between combinations of unhealthy LFs and income in models	Primary outcomes All-cause mortality: <0.05	Significant multiplicative interaction between combinations of unhealthy LFs and income for all-cause mortality. Mortality risk associated with each additional unhealthy LF was higher in higher income group.
Zhang 2021 46	1) Cox-proportional hazard models for outcomes for combinations of unhealthy LF categories, stratified by SES category 2) Cox-proportional hazard models for outcomes joint associations of combinations of unhealthy LF categories and SES measures (single reference group) 3) Sensitivity analyses for models stratified by SES category by examining subgroups: male/female, white/non-white ethnicity, age </≥60 years 4) Sensitivity analyses of joint associations substituting individual-level latent class SES for: a) Each SES component used to generate latent class, separately b) Townsend index (area-level) with adjustment for latent class SES (UKB only) and vice versa	Age, sex, marital status (NHANES only), assessment centre (UKB only), self-reported race/ethnicity, acculturation score, BMI, hypertension, diabetes, CVD, cancer, lung disease (UKB only). (10-14)	1) Interaction term between combinations of unhealthy LFs and SES variables in models 2) Estimation of RERI	All-cause mortality: 0.85; RERI =0 (NHANES), <0.001; RERI >0 (UKB) CVD mortality: 0.002; RERI >0 (UKB) CVD incidence: 0.016; RERI >0 (UKB) Secondary outcomes Heart disease mortality: 0.29; RERI =0 (NHANES) Coronary heart disease mortality: 0.008; RERI >0 (UKB) Stroke mortality: 0.002; RERI >0 (UKB) Myocardial infarction incidence: 0.050; RERI >0 (UKB) Stroke incidence: 0.032; RERI >0 (UKB)	1) NHANES: no significant multiplicative (product term for interaction) or additive interaction (RERI) between combination of unhealthy LFs and SES for all-cause or heart disease mortality UKB: both significant multiplicative and additive interactions between combination of unhealthy LFs and SES for all-cause mortality, CVD mortality, CVD incidence, coronary heart disease mortality, and stroke mortality but not for myocardial infarction incidence or stroke incidence 2) Results for product term for interaction and RERI similar across sensitivity analyses (individual-level SES, individual/area-level SES mutual adjustment) 3) In both cohorts, joint association analysis showed highest HRs in the least healthy combination of LFs and lowest SES groups for all outcomes and across all sensitivity analyses 4) Subgroup analyses showed significant multiplicative and additive interactions between combination of unhealthy LFs and SES for most subgroups (sex/ethnicity/age) and primary outcomes in UKB but not in NHANES 5) Subgroup analyses of the joint associations of combination of unhealthy LFs and SES showed higher HRs in men vs. women and in younger vs. older adults for all-cause mortality in both cohorts, and in younger vs. older adults for CVD mortality in UKB

LFs, lifestyle factors; SES, socioeconomic status; P _interaction, p-value for interaction between combinations of unhealthy LFs and SES; RERI, relative excess risk due to interaction; HR, hazard ratio; NDI, Neighborhood deprivation index; TDI, Townsend deprivation index; BMI, body mass index; CVD, cardiovascular disease; ‘Biological interaction’, the degree of interaction between risk factors in terms of deviation from additivity in adverse health outcome rates ⁵⁰ ; AP, attributable proportion; UKB, UK Biobank; NHANES, US National Health and Nutrition Examination Survey.