DISCUSSION
Results from five prospective studies (seven different cohorts) suggest that, for the same volume of physical activity, vigorous-intensity and moderate-intensity physical activities reduce mortality to the same extent. There was a 5% lower all-cause mortality among adults reporting regular vigorous-intensity physical activity compared with those reporting moderate-intensity physical activity. When we excluded one study judged as critical risk of bias (overall),33 vigorous-intensity activity was associated with a 2% lower all-cause mortality (online supplemental appendix D). On the other hand, evidence was unclear for CVD mortality cancer mortality due to the lower number of studies.
Our findings are in agreement with the current physical recommendations for health in adults developed in countries such as the USA, Australia and the UK as well as the WHO. Remarkably, all the above-mentioned recommendations do not prioritise vigorous over moderate intensity to maximise the health benefits. Nonetheless, during the last decade programmes of short duration, vigorous-intensity activities have received considerable scientific attention among clinical researchers interested in the cardiometabolic benefits of exercise.8 Given that high-intensity training studies have been conducted in laboratory settings during a short period of time, it is unknown whether vigorous-intensity activities maintained throughout life may offer larger health benefits compared with energy-matched physical activity of lower intensity. We found five prospective studies that aimed to close this gap of knowledge; yet, the existence of biases inherent to epidemiological studies challenges their interpretation. As depicted in figure 2, four of five studies included in this review were deemed as having a serious risk of bias for the confounding domain. To evaluate the risk bias in the confounding domain and following ROBINS-I recommendations, we specified a list of covariates (by expert knowledge) that requires controlling of confounding factors such as age, sex, smoking, adiposity, alcohol consumption, dietary factors and individual-level socioeconomic factors. However, we found that most studies did not control for the whole list of mentioned covariates in their multivariable models.
A well-known limitation of epidemiological studies involving physical activity and mortality is reverse causation. To minimise this problem, we pooled prospective studies that accounted for the diagnosis of any disease in the multivariable model and/or excluded those with diseased participants. Nevertheless, some studies in the literature failed to account for reverse causation. For example, in the main results reported by Gebel et al,23 participants diagnosed with cardiometabolic diseases at baseline were included in the analysis. Therefore, we only included in the meta-analysis estimates obtained from participants without cardiometabolic diseases at baseline.23 Regarding other domains included in ROBINS-I, it is remarkable that most studies did not provide information about missing data. Missing data may occur among other reasons through loss to follow-up, incomplete data collection and exclusion from analysis by investigators. Bias (ie, the effect estimate obtained in the study is different from the one obtained if authors had a complete dataset) may arise if the reason for and/or the proportion of missing data differs according to groups being compared.
Although we observed some asymmetry in the funnel plot, the results obtained in the contour-enhanced funnel plot suggest that causes of the observed asymmetry were likely due to other factors rather than publication bias (ie, systematic differences between studies in the results of large and small prospective studies or differential methodological quality of the prospective studies identified).29 Furthermore, problems with the design of some prospective studies identified in the systematic review forced us to exclude them from the meta-analysis. For example, three studies (included in the narrative review) evaluated moderate-intensity and vigorous-intensity physical activities but employed as counterfactual reference groups: people with no physical activity30 31 or participants with no physical activity and some amount of non-vigorous-intensity physical activity.32 These analytical decisions may overestimate the benefits of vigorous-intensity physical activity on mortality. Evidence of the latter may be found in the analyses of Kikuchi et al,24 where a significant protective effect was reported using the physically inactive group (450 MET min/week) while no association was found using 0% of vigorous intensity to MVPA.
From an evolutionary perspective, it has been suggested that endurance running was instrumental in the survival of Homo sapiens.
34 This opinion was, however, drawn on a narrative review of physiological and anatomical bases of endurance running capabilities of humans versus other primates. An observational study in modern hunter-gatherers (ie, Hadza) showed that adults tend to accumulate (per day) over 135 min of moderate-intensity to vigorous-intensity physical activities (measured by accelerometry), mostly at moderate intensity.1 In a recent study that compared cardiovascular adaptations induced by different types of exercise among humans, gorillas and chimpanzees, authors claimed that humans have evolved multisystem capabilities mainly matched for regular moderate-intensity endurance physical activity.35 Low volume strength and power physical activities (which can be categorised as vigorous-intensity activities) were an occasional form of physical activity in our ancestors. Interestingly, cardiological adaptations induced by strength and power exercise seem less cardioprotective (at least structurally) compared with programmes of moderate-intensity physical activity.35
Although our data indicates that both vigorous-intensity and moderate-intensity physical activities in adulthood may reduce mortality to the same extent, this finding should not downplay the key role of physical activity to improve individual and population health. Some estimates suggest that, worldwide, physical inactivity causes 6% of coronary heart disease, 7% of type 2 diabetes, 10% of breast cancer and 10% of colon cancer.3 The finding that either moderate-intensity activity or vigorous-intensity activity may provide similar reductions on mortality (for the same physical activity energy expenditure) is, if confirmed, good news because physical inactive population groups can find it difficult to attain and maintain intensities of vigorous intensity.
To our knowledge, this is the first systematic review and meta-analysis designed to evaluate whether vigorous-intensity activities (vs moderate-intensity) may provide additional reductions on mortality after controlling for total physical activity. As the results of meta-analysis are only as valid as the quality of the studies included,36 we performed risk of bias assessments of the selected literature using ROBINS-I.
Limitations
We acknowledge several limitations. First, the exposure variable (physical activity) was assessed by questionnaires, which are less accurate than objective methods of physical activity measurement.37 Future epidemiological studies will undoubtedly benefit new technological developments to evaluate objectively physical activity level in large populations. In this sense, a recent prospective study conducted in 16 741 women evaluated the association of number and intensity of steps (by accelerometry) and mortality.38 Authors concluded that stepping higher in the intensity level was not related to lower mortality rates after accounting for total steps for a day.38 Second, only one measurement of physical activity was collected in all epidemiological studies identified (at baseline), which may have led to regression dilution bias. Physical activity levels tend to decrease with age, which in theory could attenuate a protective effect on mortality of vigorous intensity due to transitions toward inactive lifestyles. Further information on physical activity changes during adulthood will increase the quality of the available epidemiological evidence. Third, a low number of studies met our strict eligibility criteria. A reason for additional concern is the lack of epidemiological data about CVD or cancer mortality. Our review, therefore, highlights the urgent need to evaluate cause-specific mortality in future well-designed epidemiological studies. Finally, a small number of cohorts were identified and pooled into the meta-analysis and subgroup or meta-regression analyses were discarded (due to low statistical power) to explore the sources of heterogeneity.