How Much Cardio For Longevity

Meta-analyses of RCTs

I found three meta-analyses on the effect of exercise on all-cause mortality:
Year # of Trials Risk Ratio Source
2016 63 0.96 (CI = 0.88 - 1.04) Anderson
2011 47 0.87 (CI = 0.75 - 0.99) Heran
2004 48 0.80 (CI = 0.68 - 0.93) Exercise-based rehabilitation for patients with coronary heart disease: systematic review and meta-analysis of randomized controlled trials

and on cardiovascular mortality:

Year # of Trials Risk Ratio Source
2016 63 0.74 (CI = 0.64 - 0.86) Anderson
2011 47 0.74 (CI = 0.63 - 0.87) Heran
2006 19 0.72 (CI = 0.55 - 0.95) Unal
2004 48 0.64 (CI = 0.61 - 0.96) Exercise-based rehabilitation for patients with coronary heart disease: systematic review and meta-analysis of randomized controlled trials

The main downside of these meta-analyses is the same as the downside of the trials: they are overwhelming performed on older people with significant medical issues (usually cardiovascular) since they are easier to recruit to exercise for years and also yield higher death sample sizes. However, there use makes it unclear whether the conclusions generalize to the general population.

Longitudinal Trials

Longitudinal studies offer a way to overcome the problem of extremely skewed samples, but in looking at them we give up the ability to readily infer causality.

Inferences of causality are always troublesome in longitudinal studies, but in this case the issues are even more pronounced: as people approach death they frequently are less able (or inclined) to exercise, which means even very good controlling for confounders is basically guaranteed too fail.

Nevertheless, for completeness, here are some brief summaries of a few large-scale studies:

Harvard Alumni

The authors of one study Associations of light, moderate, and vigorous intensity physical activity with longevity: the Harvard Alumni Health Study followed 13,485 Harvard Alumni men. After controlling for age, BMI, smoking, alcohol, and how many of their parents died before age 65, they found that exercising more than 1000 calories per week predicted 23% lower mortality relative to less than 1000 calories per week. There was essentially no gain from exercising more than that.

The authors of another study Hyde followed 16,936 Harvard Alumni men. After adjusting for age, blood-pressure status, cigarette smoking, age of parental death, and whether they had gained weight since college, they found burning 500-2000 calories per week corresponded with 1.29 years of additional life while burning 2000+ calories added an additional 0.86 years (see Table 4).

The First National Health and Nutrition Examination Survey

The First National Health and Nutrition Examination Survey followed 9,790 people Fang. After controlling for age, gender, race, income, education, diabetes history, smoking, blood pressure, cholesterol, eating disorders, caloric intake, and BMI, they found mortality decreased by between 13% and 53% between the people who described themselves with "much exercise" and "little or no exercise" (where the only other choice was "moderately active"). The difference between "much" and "moderate" exercise was between -8% and +26% (see Table 2).

Oslo

The Oslo study Holme followed 14,846 men between the ages of 40 and 50 for 12 years. After adjusting for age, education, smoking, diabetes, history of heart attack, and history of stroke, they found both light-intensity exercise, hard-intensity exercise, and degree of leisure all correlated with reduced mortality. You can see Table 2 for details, but the tl;dr is that hard intensity exercise only correlated with reduced mortality for the first ~hour per week.

Some Synthesis

Since no one has studied how quantity of exercise affects mortality in RCTs, we're left groping in the dark with only longitudinal studies for company.

This chart summarizes studies Associations of light, moderate, and vigorous intensity physical activity with longevity: the Harvard Alumni Health Study, Hyde, and Holme (assuming 500 calories is equivalent to an hour of hard exercise):

These points represent the estimates that controlled for the most variables from each study.

If you treat time spent exercising as equivalent to death, we can use the social security actuarial table Actuarial Life Table and these longitudinal studies to estimate the optimal amount of exercise per week. For a 25-year-old, the numbers come out to 0.5 hours, 3 hours, and 3 hours (for each study respectively).

However, we know that the longitudinal studies overstate the benefits from exercise because RCTs generally only find a ~10% reduction in overall mortality rather than the ~25% found in the longitudinal studies. This suggests longitudinal studies overstate the effect of exercise on mortality by roughly 2x.

We can naively adjust for this by cutting the benefits suggested by longitudinal studies in half. Curiously, this doesn't actually change the optimality estimates.

Conclusions?

At the end of the day, the latest meta-analysis of RCTs finds near-zero all-cause mortality benefits, and even if we accept the 2011 analysis and attempt adjustments to the longitudinal studies, the suggested mortality benefits are ambiguous. Given this, I'd say the balance of the evidence suggests that cardiovascular exercise probably extends life, the effects are small or moderate, and that starting earlier is generally better.

Additional evidence to consider is health metrics: cardiovascular exercise improve metrics associated with health, lowering blood pressure, reducing body fat, and decreasing your resting heart rate. Most experts believe this represents good evidence that cardiovascular exercise has significant health benefits. The fact that this wasn't born out by the 2016 meta-analysis is puzzling, so I remain pretty uncertain here. The hypothesis that seems to fit the data best is that running reduces cardio deaths by ~25% while having pretty much 0 effect on other causes of mortality.

Finally, of course the value of your life consists not merely in how many years it is, but also in how well each year is lived, and cardio has significant mental health benefits.

An Alternative Model

The conclusion above suffers from two major problems. The first is rather obvious: trying to infer causation from correlation. The second is less obvious: how does exercise's effect on mortality change as we age?

The naive model is that exercise simply scales mortality. For instance, a 25-year-old has about a 0.08% chance of dying this year. 3 hours of weekly exercise might reduce that to 0.07%. We can compute the expected years of life gained by multiplying their life expectancy (54.7 years) by the change in mortality risk (0.01%) to get ~2 days. 3 hours of weekly exercise takes 10 days of conscious time per year (assuming 8 hours of sleep each night). This makes exercise look like a bad deal for the young.

Fortunately, we don't have to fully rely on the naive model since this question was briefly examined by Hyde in Table 4. They generally find that the gains in years of life decrease with age, which is to be expected if the benefit is just a scaling in mortality. The interesting bit are the numbers themselves: they find the 35-39 group has 3.5x the benefit of the 70-74 group. If we assume exercise reduces mortality risk by a constant multiple, then we'd expect this number to be around ~1.5x. Alternatively, if we look at life expectancies at those ages, we find a ratio of 3.1-to-1. The fact this is close to 3.5-to-1 is consistent with the theory that running doesn't reduce mortality so much as it prevents aging since the benefit (in years) is proportional the years of life you have left. In particular, suggesting that running slows down aging by about 5%. All of this is correlational, scaling as we've done before suggests running prevents aging by ~2%. Since we have ~112 conscious hours per week, this suggests running extends conscious hours by ~2 hours each week, which would make it worth doing for the longevity benefits alone.

Anderson, L., Oldridge, N., Thompson, D. R., Zwisler, A. D., Rees, K., Martin, N., & Taylor, R. S. (2016). Exercise-based cardiac rehabilitation for coronary heart disease: Cochrane systematic review and meta-analysis. Journal of the American College of Cardiology, 67(1), 1-12. https://doi.org/10.1016/j.jacc.2015.10.044. Heran, B. S., Chen, J. M., Ebrahim, S., Moxham, T., Oldridge, N., Rees, K., ... & Taylor, R. S. (2011). Exercise‚Äźbased cardiac rehabilitation for coronary heart disease. Cochrane database of systematic reviews, (7). https://doi.org/10.1002/14651858.CD001800.pub2. Taylor, R. S., Unal, B., Critchley, J. A., & Capewell, S. (2006). Mortality reductions in patients receiving exercise-based cardiac rehabilitation: how much can be attributed to cardiovascular risk factor improvements?. European Journal of Cardiovascular Prevention & Rehabilitation, 13(3), 369-374. https://doi.org/10.1097/01.hjr.0000199492.00967.11. Lee, I. M., & Paffenbarger Jr, R. S. (2000). Associations of light, moderate, and vigorous intensity physical activity with longevity: the Harvard Alumni Health Study. American journal of epidemiology, 151(3), 293-299. https://doi.org/10.1093/oxfordjournals.aje.a010205. Paffenbarger Jr, R. S., Hyde, R., Wing, A. L., & Hsieh, C. C. (1986). Physical activity, all-cause mortality, and longevity of college alumni. New England journal of medicine, 314(10), 605-613. https://doi.org/10.1056/NEJM198603063141003. Fang, J., Wylie-Rosett, J., Cohen, H. W., Kaplan, R. C., & Alderman, M. H. (2003). Exercise, body mass index, caloric intake, and cardiovascular mortality. American journal of preventive medicine, 25(4), 283-289. https://doi.org/10.1016/s0749-3797(03)00207-1. Holme, I., & Anderssen, S. A. (2015). Increases in physical activity is as important as smoking cessation for reduction in total mortality in elderly men: 12 years of follow-up of the Oslo II study. Br J Sports Med, 49(11), 743-748. https://dx.doi.org/10.1136/bjsports-2014-094522. Taylor, R. S., Brown, A., Ebrahim, S., Jolliffe, J., Noorani, H., Rees, K., ... & Oldridge, N. (2004). Exercise-based rehabilitation for patients with coronary heart disease: systematic review and meta-analysis of randomized controlled trials. The American journal of medicine, 116(10), 682-692. https://doi.org/10.1016/j.amjmed.2004.01.009 Social Security: Actuarial Life Table. (n.d.). Retrieved May 14, 2020, from https://www.ssa.gov/oact/STATS/table4c6.html