Combined associations of physical activity, diet quality and their changes over time with mortality: findings from the EPIC-Norfolk study, United Kingdom

Abstract

Background: Physical activity (PA) and diet quality have each been shown to be inversely associated with mortality but their combined impact on longevity has been less explored, particularly when considering their changes over time. This study aimed to examine the separate and combined associations of PA, diet quality and their changes over time with mortality outcomes.

Methods: A prospective cohort study was performed on 9349 adults aged 40 to 79 years from the population-based European Prospective Investigation into Cancer in Norfolk Study, with repeated measurements of PA and diet (from 1993 till 2004) and subsequent follow-up till 2022 (median follow-up 18.8 years). Validated questionnaires were used to derive physical activity energy expenditure (PAEE) as a proxy of total PA and adherence to the Mediterranean diet score (MDS, range 0-15 points) as an indicator of overall diet quality, and their changes over time (∆PAEE and ∆MDS). Cox regression models adjusted for potential confounders and mediators were used to estimate hazard ratios (HRs) and 95% CIs.

Results: Over 149,681 person-years of follow-up, there were 3534 deaths. In adjusted models, for each 1-SD difference in baseline PAEE (4.64 kJ/kg/day), ∆PAEE (0.65 kJ/kg/day per year), baseline MDS (1.30 points) and ∆MDS (0.32 points per year), HRs (95% CI) for all-cause mortality were 0.90 (0.86 to 0.94), 0.89 (0.85 to 0.93), 0.95 (0.91 to 0.99) and 0.93 (0.90 to 0.97), respectively. Compared with participants with sustained low PAEE (< 5 kJ/kg/day) and low MDS (< 8.5 points), those with sustained high PAEE and high MDS had lower all-cause mortality (HR 0.78; 95% CI: 0.68-0.91), as did those who improved both PAEE and MDS (0.60; 0.44-0.82). There was no evidence of interaction between PA and diet quality exposures on mortality risk. Population impact estimates suggested that if all participants had maintained high levels of PA and diet quality consistently, cumulative adjusted mortality rate would have been 8.8% (95% CI: 2.4 to 15.3%) lower.

Conclusions: These findings suggest that adopting and maintaining higher levels of PA and diet quality are associated with lower mortality. Significant public health benefits could be realised by enabling active living and healthy eating through adulthood.

Keywords: Cancer mortality; Cardiovascular diseases mortality; Cohort studies; Diet quality; Mediterranean diet; Mortality; Physical activity; Prospective studies; Trajectories.

Fig. 1

EPIC-Norfolk Study design and timeline

Fig. 2

Joint associations of trajectory patterns of PA and diet with all-cause mortality in the EPIC-Norfolk Study. A Comparison across all trajectories together (G1 as reference); B Stratified analysis by baseline exposure level (stable behaviours as reference groups). Analysis is based on model 4 (see ‘Methods’). Definition for high diet quality: MDS ≥ 8.5 points; high PA: PAEE ≥ 5 kJ/kg/day. MDS, Mediterranean diet score; PAEE, physical activity energy expenditure; PA, physical activity; DQ, diet quality. Groups were formed according to reaching a high (H) or low (L) level of baseline PA, baseline DQ, repeated PA and repeated DQ as follows: G1 = LLLL, G2 = LLHL, G3 = LLLH, G4 = LLHH, G5 = HLLL, G6 = HLHL, G7 = HLLH, G8 = HLHH, G9 = LHLL, G10 = LHHL, G11 = LHLH, G12 = LHHH, G13 = HHLL, G14 = HHHL, G15 = HHLH and G16 = HHHH

Fig. 3

Association of mutually adjusted baseline and within-person changes in PAEE and MDS with all-cause mortality in different strata of baseline age, sex, BMI, smoking status and pre-existing comorbidities per 1-SD difference in each exposure, in the EPIC-Norfolk Study. Analysis is based on model 4 (see ‘Methods’). MDS, Mediterranean diet score; PAEE, physical activity energy expenditure. 1-SD increment in baseline PAEE equals to 4.64 kJ/kg/day, in ΔPAEE equals to 0.65 kJ/kg/day per year, in baseline MDS equals to 1.30 points and in ΔMDS equals to 0.33 points per year. Pre-existing comorbidities were defined as having any of the following: cancer, cardiovascular disease, diabetes or hypertension. †P values for interaction in subgroups

Fig. 4

Population impact by estimating the cumulative adjusted mortality rate of the population during two decades of follow-up, under observed physical activity and diet exposures, compared with two counterfactual scenarios. Counterfactual scenario 1; applying the adjusted mortality rate of trajectory group 1 (G1, low PA and low diet quality at both assessments) to the whole population. Counterfactual scenario 2; applying the adjusted mortality rate of trajectory group 16 (G16, high PA and high diet quality at both assessments) to the whole population. See Fig. 2 and ‘Methods’ for the definition of each trajectory group. Error bars indicate 95% confidence intervals (95% CI) for percentage of changes in adjusted mortality rate of the population under counterfactual scenarios, compared with adjusted mortality rate under observed physical activity and diet exposures