Too Much Fitness is Barely Enough

This graph, from Mandsager et al., shows the rate at which people die, depending on how fit they are. The top 2.3% of the people studied (Elite) had the slowest rate of death, and then each group of less and less fit participants died at a faster and faster rates.

“Too much of anything is bad, but too much good whiskey is barely enough.”

The above quote has been widely attributed to Mark Twain, and although I have not been able to find any source that debunks this attribution, neither have I been able to confirm it to my satisfaction. Anyway, it really doesn't matter who said it, the quote is amusing, especially to those of us who love our whiskey. But it also raises the serious question; is fitness like "anything", or is it like "whiskey"? Over the last few years, there have been a series of scientific journal articles reporting that some extreme athletes have health problems, suggesting that exercise might be a Goldilocks activity, where too little or too much is bad for you, that to stay healthy you want a level of fitness which is "just right". A recent paper [1] (hereafter referred to as Mandsager et al.) on the impact of fitness on longevity weighed in on this topic, and the authors seems to say that, when it comes to fitness, more is better. How convincing is this paper?

The most striking aspect of Mandsager et al. is the number of participants, a stunning 122,007! How did they get so many? They didn't, their participants were actually patients of the Cleveland Clinic who had their cardiopulmonary fitness measured by exercise treadmill testing between 1991 and 2014 as part of their regular medical care. Mandsager et al. reviewed their medical records and asked if there was any relationship between the fitness measured in that test and the probability that the patient had died for each year after the test. What they found was that the the higher the level of fitness, the less likely a patient was to die. But what is perhaps as striking as the number of participants is the magnitude of this effect. To put this into context, they had information about other characteristics of the patients as well, if they smoked for example. Smoking is generally believed to be highly harmful to health, and it is. Smokers in this group of patients were 1.4 times as likely to die in any given year as non-smokers. In comparison, those in the group with the lowest level of fitness were 5 times as likely to die in any given year as those with the highest level of fitness! But is this the right way to answer the question? What is the question? What do they mean by fitness?

Fitness was determined by the maximum speed and steepness of a treadmill the patient could manage before they said they were too tired to continue or until they developed symptoms which required the test to be discontinued. Fitness in this test was measured in units of relative maximum metabolic equivalents (METs.) (Maximum MET is closely related to the VO2max we in the exercise community use so often.) To make the MET scores relative, the patients were divided into groups of similar people. First, men and women were separated into two different groups and then these two groups were subdivided by age into 10 year subgroups. Each of these age and sex groups were ordered by MET score, and divided into the the 25% with the Lowest scores, the Below Average 25% (25%-50%), the Above Average 25% (50%-75%), the 25% with the Highest score (75%-100%) and the Elite 2.3% (97.7%-100%) [2]. In supplementary data, Mandsager et al. also looked at the top 5% and 1%. Groups with different levels of fitness, relative to their age and sex group, were compared one to another with regards to the probability that each group would die during each year following the test.

Could it be that the longevity of the top 2.3% of patients could be due to something other than exercise, their genes for example? Absolutely, this is a possibility that Mandsager et al. explicitly acknowledge in their paper. What can be done to get around this problem? Realistically, nothing. Mandsager et al. is a retrospective study. A retrospective study cannot prove if one thing (e.g. fitness) causes another (e.g. longevity.) The only way to definitively show that fitness leads to longevity would be to do a randomized, prospective study. Such a study would have to use a flip of the coin (or some other random process) to divide its participants into two groups. One group would be required to follow a specific exercise plan as a way to improve their fitness. The other group would be forbidden to exercise, even if they wanted to. Both groups would have to keep up these behaviors for years, or even decades, to determine if those who were randomly selected to exercise died less frequently than those who were randomly prevented from exercising. Even if one could get volunteers for such a study, no Institutional Review Board would approve it, the evidence that exercise (or fitness) improves health is so convincing it would be considered unethical to prevent the no exercise group from exercising. In the face of this impossibility, how can we determine if exercise or fitness is good for health? We do so by combining the results of a wide range different kinds of studies, each with their own limitations, and if all of these different kinds of studies point to the same conclusion, we become more and more convinced that this conclusion is correct. Given the massive amount of data on exercise, fitness, and health we have accumulated over the years, the evidence is strong that exercise and fitness improves health, even in the absence of a definitive, prospective study.

The four things that Mandsager et al. add to our understanding are first, by taking yet one more different approach to looking at this question, they increase the likelihood that our belief that exercise and fitness leads to long life is correct. Second, because of the huge amount of data their study contains, they provide the most striking estimate to date as to the size of this benefit. Third, again because of the amount of data, they can provide a convincing evidence that, within reason, for most people, the fitter you are, the longer you will live. Finally, their study suggests this is true if you are male or female, if you have high blood pressure, high cholesterol or not, if you are less than 50 years old or more than 70 years old, this is true; it appears to be true for almost everyone.

Do Mandsager et al. really argue that no amount of fitness is too much? Not quite, the title of my post is an oversimplification. What Mandsager et al. do argue is that within the 122,007 patients examined in their study, very few if any of them were too fit. In fact, even when Mandsager et al. look at the fittest 1% of their participants, it was still the case that they lived longer than the top 2.3%, so within their study group, it seems like too much fitness is barely enough. The problem is that even the fittest 1% of their participants do not have MET scores anywhere near as high as fittest athletes in the world. So, if you are an ultra-marathon runner and are looking for validation for the healthfulness of your hobby, Mandsager et al. provide a weak validation at best. However, for most of us, Mandsager et al. encourage us to get as fit as possible. Finally, it is important to note that although the fittest 2.3% had a 5 times lower mortality rate than the least fit 25%, the next to the worst group, the Below Average 25% to 50%, had a 1.95 times lower mortality rate than the Lowest fitness 0% to 25% group, an effect larger than that of not smoking, so while it is true that the more fitness the better, it is also true that some fitness is much better than none.

It is important to understand the difference between fitness and exercise. For most people, the right amount and the right kind of exercise does increase their fitness, but genetics also has a huge impact, as do many other factors. Two people who exercise the same amount will almost never achieve the same level of fitness. Some people who almost never exercise will have a higher level of fitness than others who exercise diligently. Mandsager et al. in and of themselves do not deal with exercise at all, only fitness. However, there are other studies that survey people, asking them how much they exercise, and then follow those people over time. One such study is Lear et al. [3] (I selected this study in particular, because it also had an extremely large sample size, 130,000 participants. I hope to cover this landmark paper in a future post.) Lear et al. estimated a 1.5-fold increase in longevity due to exercise, still large, but much less than the 5-fold effect due to fitness estimated by Mandsager et al. What could account for the difference? It is impossible to know based just on these studies, and in fact I am unaware of any studies that answer this question convincingly, but there are some obvious guesses. Some or all of that difference might represent the genetic component to fitness, which might also impact longevity. Fitness can also be impacted by general health issues, people who are unwell both cannot exercise as much and also probably don't benefit as much from the exercise they do, so will measure low on the fitness scale. (Mandsager et al. did analyses to rule out this particular concern, but only in part. It is possible that there some illnesses whose impact these analyses could not eliminate.) One point that Mandsanger et al made in support of their approach of looking at fitness rather than exercise is that people are not always honest when surveyed about how much they exercise. They may remember incorrectly and/or flatter themselves by overestimating the amount of exercise they do. Such people would dilute the benefits of exercise found in Lear et al.; they would count as high exercisers but have the life expectancy of low exercisers, meaning that the the 1.5-fold benefit of exercise reported by Lear et al. might be an underestimate.

All the above limitations notwithstanding, taking these two papers at face value [4] leads to some interesting conclusions. Note that the benefit of maximum fitness (5-fold) is much greater than the benefit of maximum exercise (1.5-fold). There are a very large number of theoretical explanations for that observation, but an obvious one is that the benefit of exercise comes not from the act of exercising per se, but from the increased fitness that exercise produces. Thus, when comparing the health benefits of different exercise plans, measuring how much fitness they produce, something that can be done relatively easily in a short term study, may be a good way of determining which exercise plans are better. For example, the results of Gillen et al. advocating for High Intensity Interval Training become more convincing. At first glance, it may seem counterintuitive that 1 minute of high intensity intervals can produce as much health benefit as 45 minutes steady cycling, but because we can easily demonstrate that the impact on fitness is the same, and because we believe that fitness is what determines health, we can accept this counterintuitive result with increased confidence. For an individual, this is even more valuable. There is no way for one individual to measure the effect of any given exercise protocol on their longevity, but it is easy to measure its effect on their fitness. If fitness leads to longevity, then the exercise protocol that leads to the highest level of fitness will also lead to the greatest longevity. Once again, none of this is proven by these studies, it is just one plausible explanation, but one that I find persuasive. Thus, I expect to refer back to this post frequently as it will make interpretation of the exercise literature much easier and more direct. Stay tuned.

---

[1] Association of Cardiorespiratory Fitness With Long-term Mortality Among Adults Undergoing Exercise Treadmill Testing by Kyle Mandsager, MD; Serge Harb, MD; Paul Cremer, MD; Dermot Phelan, MD, PhD; Steven E. Nissen, MD; Wael Jaber, MD in JAMA Network Open (2018) Volume 1 e183605, doi:10.1001/jamanetworkopen.2018.3605.
https://jamanetwork.com/journals/jamanetworkopen/fullarticle/2707428

[2] This seemingly odd choice of the top 2.3% as the elite group was made because this value corresponds to those who are statistically two standard deviations or more above the average level of fitness. For the purposes of this post, it is not important you understand the significance of this, it is here for the curious.

[3] The effect of physical activity on mortality and cardiovascular disease in 130,000 people from 17 high-income, middle-income, and low-income countries: the PURE study by Scott A Lear, Weihong Hu, Sumathy Rangarajan, Danijela Gasevic, Darryl Leong, Romaina Iqbal, Amparo Casanova, Sumathi Swaminathan, R M Anjana, Rajesh Kumar, Annika Rosengren, Li Wei, Wang Yang, Wang Chuangshi, Liu Huaxing, Sanjeev Nair, Rafael Diaz, Hany Swidon, Rajeev Gupta, Noushin Mohammadifard, Patricio Lopez-Jaramillo, Aytekin Oguz, Katarzyna Zatonska, Pamela Seron, Alvaro Avezum, Paul Poirier, Koon Teo, Salim Yusuf in Lancet (2017) Volume 390 Page 2643.
https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(17)31634-3/fulltext

[4] ...and noting but not belaboring the "Correlation Does Not Prove Causation" and other stringencies of the scientific method. I have tried to phrase this section cautiously to give such issues their due without obscuring the obvious (if not proven) working conclusions that a reasonable person might draw from these papers.