Mediating effect of cardiorespiratory fitness in the association between physical activity and myocardial infarction

Abstract

Background: Cardiovascular disease (CVD) is one the global leading cause of mortality and myocardial infarctions (MI) make up a significant proportion of this. It has been previously found that the relationship between physical activity (PA) and all-cause mortality is largely mediated trough cardiorespiratory fitness (CRF), and in this context I hypothesised that CRF also mediates the association between PA and MI as well. Method: This thesis is a longitudinal cohort study, using data form the fifth wave of Tromsø Study in 2001. PA was measured using the Saltin-Grimby Physical Activity Level Scale (SGPALS) grouped in three levels of PA: inactive, moderately active, and highly active. CRF was calculated using a non-exercise formula (estimated CRF). MI was derived from hospital records through 2014. Cox regression was used to assess the association between inactive (reference), moderate and high levels of PA and MI, adjusted for age, sex, education, smoking and waist circumference. The same analysis was performed with eCRF included from the abovementioned formula. Thereafter, Vanderwheele´s four-way decomposition analysis was applied in the Cox regression where I assessed the mediation effect of CRF on the association between PA and MI. Results: In total, of the 5175 participants, 296 suffered an MI during the median 13.26 (interquartile range: 0.49) years of follow up. Comparing inactive as reference, moderately and highly active displayed 35% (HR: 0.65, 95%CI: 0.49-0.85) and 33% (HR: 0.66, 95%CI: 0.45-0.97) lower risk of MI, respectively. When including eCRF in the analysis, the associations were attenuated, where moderately active showed 28% (HR: 0.72, 95%CI: 0.55-0.94) lower risk of MI compared to inactive, while highly active versus inactive showed CIs crossing unity (HR: 0.83, 95%CI: 0.57-1.22). In the 4-way decomposition in both estimates (inactive versus moderately-, and versus highly active), the pure direct effects were insignificant (both p>0.50). The proportion explained by reference interaction were 60% (95%CI: 19, 102) and 76% (95%CI: -14, 166), and mediated interaction were -34% (95%CI: -55, -14) and -97% (95%CI: -211,18), and pure indirect effects explained 52% (19, 85) and 84% (95%CI: -5, 177%) of proportion in the association of lower MI risk. Conclusion: Higher PA was associated with lower risk of MI. When assessing the mediating effect of eCRF in the association between PA and MI, eCRF appeared to fully mediate the association between PA and lower risk of MI.