Last time, I found that the healthiest BMI range for all-cause mortality is 20–22. But BMI doesn’t tell the whole story. Most obviously, it doesn’t account for body fat vs. lean mass. All else equal, you’d rather have more muscle1 and less fat.

So what’s the healthiest combination of lean mass + fat mass?

I’m not going to answer that question because I can’t. Instead, I will explain why I can’t, and then give a rough guess at the answer.

Scientists have been measuring and collecting data on BMI for decades. You can find plenty of giant BMI studies with three million participants in various countries.

We have much sparser data on body fat. Scientists didn’t start collecting data on body fat until the last few decades. And body fat is harder to measure—we have various methods for estimating body fat, but they’re all more complicated than calculating BMI.

I managed to scrounge together some studies on body fat and mortality. My best guess: the average woman should aim for a BMI of 21 with 20% body fat, and the average man a BMI of 21 with 10% body fat. (Subject to individual variation due to genetics and whatnot.)

Trans men should probably target the same body fat % as cis men, and likewise for trans women and cis women, because hormone therapy alters body fat distribution (Spanos et al. (2020)2).

The evidence weakly suggests that there is no lower bound on healthy fat mass, and no upper bound on healthy lean mass. We have so little mortality data on extremely lean + muscular people that we can’t say how healthy they are.

A more in-depth analysis would look at a variety of health indicators (blood pressure, HDL cholesterol, etc.) and use that to predict mortality. I didn’t do that, I just looked at mortality data.

Contents

I looked at all the relevant research papers with at least 100 citations on Google Scholar. Here’s what I found.

Research on body composition and mortality

The best data comes from the big Danish follow-up study “Diet, Cancer and Health” which followed 50,000 Danish adults aged 50 to 64 from 1993 to 1997. Two different research papers analyzed the data from this study: Bigaard et al. (2004)3 and Bigaard et al. (2005)4. These papers break down BMI into body fat mass index (BFMI) and fat-free mass index (FFMI) (measured using bioelectrical impedance). Just like how BMI equals weight/height2, BFMI equals bodyfat/height2, and FFMI equals fat-free mass (a.k.a. lean mass) / height2 (such that BFMI + FFMI = BMI).

Bigaard et al. (2004) found that mortality monotonically increases with BFMI above 5–6, and gets slightly worse below 5–6. For women, mortality monotonically decreases with FFMI, and for men it monotonically increases up to 19, at which point it starts increasing again. See Figure 1 (dark line represents men, light line represents women).

(BFMI and FFMI were controlled for each other, which fixes the problem that people with more lean mass usually have more fat mass.)

If we combine the ideal BFMI and FFMI from the Danish study, we get:

men:   5 BFMI + 19 FFMI = 24 BMI with 21% body fat
women: 7 BFMI + 17 FFMI = 24 BMI with 29% body fat

But this probably isn’t right. The best meta-analyses on BMI and mortality control for three big confounders: smoking, health conditions, and study follow-up length. All these confounders make low BMIs look unhealthier than they really are. The Bigaard et al. papers controlled for smoking, but not for health conditions (although participants were healthier than average due to selection effects5); plus the Danish study only lasted 5 years, and 5-year studies show a bias toward high BMIs.

In the BMI studies, controlling for these additional confounders reduces the apparent healthiest BMI by 3 or 4 points. And the Danish study found that mortality was minimized at 25 BMI,6 whereas in reality, it’s probably minimized at 20–22 BMI.

Undiagnosed health conditions probably decrease both lean mass and fat mass. That means low fat mass and low lean mass are both healthier than observational studies make them look.

If we naively subtract 4 points from the body-fat mass index,7 we get a new estimate of the healthiest body composition for men and women:

men:   1 BFMI + 19 FFMI = 20 BMI with 5% body fat
women: 3 BFMI + 17 FFMI = 20 BMI with 15% body fat

These numbers are extraordinarily low. But remember that I came up with these numbers by extrapolating what might happen if we controlled for confounders, so the numbers don’t come from actual data and could be pretty far off. My guess is they’re too low—maybe selection bias in the Diet, Cancer and Health Study had a similar effect to controlling for health, and I’ve over-corrected.

I don’t know about the exact numbers, but I expect that low body fat is good.

The data showed a trend of increasing mortality above 19 FFMI for men and 17 FFMI for women, but the trend wasn’t strong, so it’s possible that higher FFMIs are healthier (more on this later).

Other research has shown that ideal BMI increases with age, and the Danish study only included adults aged 50 to 64, so younger people should probably target lower body fat, and older people should aim higher. (It’s possible that younger people actually want to have less lean mass, not less body fat, but that doesn’t sound likely to me.)

Ideal BMI varies by ethnicity. But that’s probably because different ethnicities distribute body fat differently—e.g., South Asians carry more body fat at a given BMI, which makes BMIs on the higher end look unhealthier for South Asians. I don’t know of any cross-ethnic studies on body fat and mortality, but my best guess is that ideal body fat doesn’t vary much by ethnicity.

Lee et al. (2018)8, the second-most useful paper, looked at (predominantly white) males from the long-term Health Professionals Follow-Up Study and found that, controlling for confounders, mortality for men was minimized in the first quintile of fat mass and the third quintile of lean body mass (see Table 2). That is, men should aim for low fat mass and average lean mass.

(Lean mass and mortality had only a weak association except in the first quintile: low lean mass increased risk, but lean mass didn’t clearly affect mortality beyond that.)

Lee et al. (2018) found that male mortality was minimized at 56 kg of lean mass and 5–21 kg of fat mass.9 The median of the 1st quintile had a fat mass of 15 kg which corresponds to 21% body fat.10

Lee et al. found that mortality was minimized at a BMI of 24 or so, which suggests that it hasn’t fully adjusted for confounders. Adjusting fat mass index by 3 points to produce a BMI of 21 (assuming a height of 176 cm, which corresponds to a BMI of 2311 for someone with 56 kg lean mass + 15 kg fat mass) predicts an ideal male body composition of

2 BFMI + 18 FFMI = 20 BMI with 9% body fat

Interestingly, Lee et al. found that if you exclude men with excessively low lean mass, then the mortality-minimizing BMI range is 18.5–20.4, not 20–22 as I previously reported. (The meta-analyses I looked at in my last post didn’t include data on body fat.) And the effect was surprisingly strong—men in the 18.5–20.4 group had a 15% lower mortality rate than the 20.5–22.4 group (see Table 3). This suggests two things:

  1. Having a BMI on the low end (18.5–20.4) has large health benefits and large downsides relative to a mid-range BMI (20.5–22.4).
  2. The downsides pretty much exclusively come from low lean mass, not low fat mass. If you have low-end BMI with sufficiently high lean mass, you get all the upside and ~none of the downside.

(Note: After excluding people with low lean mass, there were zero men left with BMIs below 18.5, out of a sample of 38,000. You might shoot for a BMI even lower than 18.5, but it looks pretty much impossible.12)

Research on waist circumference

A few studies looked at mortality and waist circumference. Waist circumference is often used as a proxy for body fat, and arguably it’s a better metric than body fat % because visceral fat (abdominal fat that’s distributed around the organs) carries greater health risks than subcutaneous fat (distributed under the skin).

Zhang et al. (2008)13 looked at American women in the Nurses’ Health Study. Participants had the lowest mortality in the 1st quintile of waist circumference—less than 28 inches or 71 cm, see Table 3. The study also looked at waist:hip ratio and found that the 1st quintile (<0.73) minimized mortality risk.

Zhang et al. reported a hazard ratio of 1.01 for the 28–29 inch waist group for all participants—that is, 28–29 inches was only slightly less healthy than <28 inches. But among never-smokers, the 28–29 inch group had a hazard ratio of 1.31. Reading between the lines, that implies that the healthiest waist circumference for non-smoking women is considerably less than 28 inches (71 cm).14

Hu et al. (2018)15 found that among middle-aged and elderly Chinese individuals, mortality was minimized at a waist circumference of 83–88 cm for men and 79–83 cm for women. This study used a relatively short follow-up (8.5 years), which means the true healthiest waist circumference is probably lower.

Baik et al. (2000)16 found that American men had the lowest mortality rate for waist circumferences in the 3rd quintile (36.3–37.9 inches or 92.2–96.3 cm), and waist:hip ratios in the 2nd quintile (0.90–0.91), see Table 4. The authors write that the higher mortality for lean men is probably due to confounding—respiratory disease causes men to lose weight and die sooner.

Research on the relationship between BMI and body fat

Some studies provide formulas to convert BMI to body fat. We can use those formulas to estimate ideal body fat from ideal BMI. This method doesn’t really work because the healthiest BMI substantially changes if you know someone’s fat mass / lean mass. But I’m going to use the method anyway and see what happens.

  • Data from the US National Health and Nutrition Examination Survey in Flegal et al. (2009)17 suggest a correspondence between ideal BMI and body fat % (see Table 3 and Table 4):

    demographic body fat
    women under 40 25–30%
    women over 60 30–35%
    men under 40 15–20%
    men over 60 20–25%

  • Meeuwsen et al. (2010)18 measured body fat in UK adults and came up with a linear formula to predict body fat percentage:

    women: BF% =  -1.63 + 1.129 * BMI + 0.140 * age
men:   BF% = -13.51 + 1.129 * BMI + 0.140 * age

    This formula predicts that, assuming you want a BMI of 21 at 30 years old, the healthiest body fat is 26% for women and 14% for men. At age 60, that rises to 30% for women and 19% for men.

  • Mills et al. (2007)19 found a BMI/body-fat correlation that suggests an ideal male body fat of 4% to 9%. But it also says a white male with a BMI of 18 has –1% body fat so I suspect it’s pretty inaccurate on the low end.20

  • Ranasinghe et al. (2013)21 derived formulas to estimate body fat percentage in Sri Lankan adults:

    women: BF% =  3.819 + 0.918 * BMI + 0.153 * age
men:   BF% = -9.662 + 1.114 * BMI + 0.139 * age

    These formulas predict an ideal body fat of 28% for women and 18% for men at age 30.

Good research that nonetheless didn’t answer my question

Some other studies looked at the association between body fat/waist circumference and mortality, but only focused on the effects of overweightness/obesity, which doesn’t tell us the healthiest body composition.

  • Bigaard et al. (2005)4, using the Danish data set discussed previously, found that waist circumference predicted mortality even when controlling for BMI. It also found that, if you control for waist circumference, mortality monotonically decreases with both body fat mass index and fat-free mass index. (I wish I knew what to do with this information—I don’t know how to increase body fat without also increasing waist circumference.22)
  • Britton et al. (2013)23 found that excess body fat was associated with elevated mortality risk, but did not provide granularity on the low end of body fat.
  • Cerhan et al. (2013)24, a pooled analysis of 11 cohort studies with a total of 650,000 white participants, found that mortality monotonically increased with waist circumference 90+ cm for men and 70+ cm for women. But it did not look at waist circumferences smaller than 90 cm for men or 70 cm for women.
  • Chen et al. (2019)25 found greater mortality for waist circumferences of 90+ cm for Chinese men and 80+ cm for Chinese women, but did not provide granularity beyond that.
  • Jacobs et al. (2010)26 found that among 100,000 (mostly white) Americans over a 10-year period, lower waist circumference was associated with reduced mortality, but it did not present data on waist circumferences smaller than 90 cm for men or 75 cm for women.
  • Lee et al. (2018)27 found that body fat predicted mortality more strongly than BMI, and visceral fat more strongly still. It found that people in the bottom third by fat mass had the lowest mortality, as did the people in the bottom third by visceral-fat-to-subcutaneous-fat ratio.
  • Visscher et al. (2001)28 found that male waist circumferences over 94 cm were associated with increased mortality, and failed to find a trend among women.

Is there an upper bound on healthy lean mass?

We know high BMIs are unhealthy, and we know that the main harms to health come from excess body fat. Does excess lean mass increase mortality risk? Or is it better to be as big and lean as possible?

The Danish Diet, Cancer and Health Study found that, after controlling for body fat, mortality risk decreases with lean mass up to a fat-free mass index (FFMI) of 17 for women and 19 for men, and starts increasing again above that point. The increasing trend above 19/17 is not statistically significant: after adjusting for both body-fat mass index (BFMI) and smoking, the upper portion of the FFMI curve has a slope of 1.03 with a 95% confidence interval of [0.93, 1.13] (see Table 4 from Bigaard et al. (2004)3).

This suggests that there’s no harm to having a FFMI above 17 for women or 19 for men, but there’s no benefit, either.

Alternatively, we can look at the relationship between resistance training and mortality because resistance training is closely associated with lean mass. A 2022 meta-analysis by Momma et al.29 found that muscle-strengthening activity was associated with reduced mortality risk up to about 60 minutes per week, but with marginal resistance training above 60 minutes/week showing increasing mortality, and 140 minutes per week showing the same mortality risk as 0 minutes. The meta-analysis did not control for BMI, which I suspected could explain the increase in mortality, but one large cohort study (Patel et al. (2020)30) did control for BMI (among other factors) and also found that resistance training was associated with increased mortality above 60 minutes or so.

I’ve seen some people dismiss this finding on the basis that resistance training is well-known to improve many health indicators like blood pressure and bone density. That’s true, but from what I’ve seen, the research primarily looks at low-dose resistance training, so we can’t say based on that research that increases in training volume monotonically improve health at high doses.

Why might muscle-strengthening activity increase mortality risk? I have seen two proposed mechanisms:

First, Miyachi (2013)31 found that resistance training increases arterial stiffness, which increases heart disease risk.

Second, Nuckols (2022)32 writes:

The studies in [Momma et al. (2022)] mostly used older subjects. It’s entirely possible that the optimal dose of resistance training for older adults is a lot lower than the optimal dose of resistance training for younger adults. For example, oxidative stress and generalized inflammation likely contribute to biological aging, and older adults have higher levels of oxidative stress and generalized inflammation. Resistance training causes oxidative stress and inflammation in a dose-dependent manner, but this is generally a good thing – those stressors are triggers for training-induced adaptations, and they also trigger your body to ramp up endogenous antioxidant production so that you can better handle future stressors (resulting in net reductions in inflammation and oxidative stress at rest). However, excessive training doses can induce too much oxidative stress and inflammation, setting the stage for a variety of deleterious outcomes (which we tend to collectively refer to as “overtraining”. It’s entirely possible – likely, even – that the threshold between productive training-induced stress and unproductive training-induced stress is considerably lower in older adults than younger adults.

Summary of findings

  1. None of the studies fully controlled for confounders, so we have to make a judgment call as to what the results would look like if we did fully control for them.
  2. Two big studies—the Danish Diet, Cancer and Health Study and the American Health Professionals Follow-Up Study—provide data on the association between mortality and lean mass / fat mass. If we extrapolate to what results we might see if we controlled for all confounders, these studies predict that men minimize mortality risk at 5–10% body fat and women minimize mortality risk at 15–20% body fat, as long as they have adequate lean mass.
  3. Some research suggests that people in the lowest quintile of waist circumference have the lowest mortality risk, but we don’t know exactly what waist circumference minimizes mortality risk.
  4. A great deal of research shows that on the higher end, higher body fat / bigger waist circumference is associated with increased mortality risk.
  5. Higher lean body mass decreases mortality risk up to a fat-free mass index of 19 for men and 17 for women. Excess lean body mass above that point might increase mortality risk, but it’s not clear.

Notes

  1. Technically we want to look at lean mass, not muscle mass. Lean mass refers to any body mass that isn’t fat, which includes muscles, bones, organs, etc. But the main way to gain lean mass is by building muscle. Resistance training does also increase bone mass, but much more slowly than muscle mass. I don’t know of any way to increase your organ mass, and even if you could, I don’t know that you would want to. ↩

  2. Spanos, C., Bretherton, I., Zajac, J. D., & Cheung, A. S (2020). Effects of gender-affirming hormone therapy on insulin resistance and body composition in transgender individuals: a systematic review. See Table 1. ↩

  3. Bigaard, J., Frederiksen, K., Tjønneland, A., Thomsen, B. L., Overvad, K., Heitmann, B. L., & Sørensen, T. I. (2004). Body fat and fat-free mass and all-cause mortality. ↩ ↩2

  4. Bigaard, J., Frederiksen, K., Tjønneland, A., Thomsen, B. L., Overvad, K., Heitmann, B. L., & Sørensen, T. I. A. (2005). Waist circumference and body composition in relation to all-cause mortality in middle-aged men and women. ↩ ↩2

  5. Study participants died at only about half the rate of the general population, see Bigaard et al. (2005)3. ↩

  6. Looking purely at BMI, the study found that mortality was minimized at 25. Looking at the combined healthiest BFMI and FFMI, mortality was minimized at 24 for men and 23 for women. I believe the discrepancy comes from the fact that people with low body fat often have insufficient lean mass. ↩

  7. Instead of subtracting 4 points from FFMI, I could subtract 2 points from BFMI and 2 points from FFMI. That seems worse to me because I would expect that few people with a FFMI of 19/17 have health conditions that reduce their lean mass—if they did, their FFMI would be considerably lower. So the observation that an FFMI of 19/17 minimizes mortality shouldn’t be confounded by health conditions. ↩

  8. Lee, D. H., Keum, N., Hu, F. B., Orav, E. J., Rimm, E. B., Willett, W. C., & Giovannucci, E. L. (2018). Predicted lean body mass, fat mass, and all cause and cause specific mortality in men: prospective US cohort study. ↩

  9. According to their “model 2”, which controls for various factors but does not control for health conditions. ↩

  10. Figure 1 shows the lowest mortality at a fat mass of 21 kg, which corresponds to a surprisingly-high 27% body fat. I believe this is an artifact of the model used to generate Figure 1. Figure 1 fits the data to a cubic spline model which smooths out the bumpiness in the quintiles. The 1st quintile had the lowest mortality rate but its 95% CI overlapped with the 2nd and 3rd quintiles, so the cubic spline model smoothed these out and ended up predicting that mortality is minimized in the 3rd quintile. Thanks to co-author Edward Giovannucci for this explanation. ↩

  11. I use 23 instead of 24 because my analysis of Bigaard et al. (2004) found that splitting out lean mass and fat mass reduced apparent healthiest BMI by 1 point. ↩

  12. For me personally, if I somehow dropped by body fat to 0% without losing any lean mass, I’d still have a BMI of 20. ↩

  13. Zhang, C., Rexrode, K. M., Van Dam, R. M., Li, T. Y., & Hu, F. B. (2008). Abdominal obesity and the risk of all-cause, cardiovascular, and cancer mortality: sixteen years of follow-up in US women. ↩

  14. My reasoning: If the 28–29 group looks only slightly less healthy than the <28 group for all participants (including smokers), and if the mortality curve follows an U shape, then the nadir of the curve should be just below 28 inches. If the 28–29 group looks substantially less healthy, then the nadir must be considerably lower than 28 inches because that way most of the function’s mass near the nadir occurs below 28 inches. ↩

  15. Hu, H., Wang, J., Han, X., Li, Y., Wang, F., Yuan, J., Miao, X., Yang, H., & He, M. (2018). BMI, waist circumference and all-cause mortality in a middle-aged and elderly Chinese population. ↩

  16. Baik, I., Ascherio, A., Rimm, E. B., Giovannucci, E., Spiegelman, D., Stampfer, M. J., & Willett, W. C. (2000). Adiposity and mortality in men. ↩

  17. Flegal, K. M., Shepherd, J. A., Looker, A. C., Graubard, B. I., Borrud, L. G., Ogden, C. L., . & Schenker, N (2009). Comparisons of percentage body fat, body mass index, waist circumference, and waist-stature ratio in adults. ↩

  18. Meeuwsen, S., Horgan, G. W., & Elia, M (2010). The relationship between BMI and percent body fat, measured by bioelectrical impedance, in a large adult sample is curvilinear and influenced by age and sex. ↩

  19. Mills, T. C., Gallagher, D., Wang, J., & Heshka, S. (2007). Modelling the relationship between body fat and the BMI. ↩

  20. Then again, I would need to have –9% body fat to get my BMI down to 18 without losing lean mass, so maybe they’re on to something. ↩

  21. Ranasinghe, C., Gamage, P., Katulanda, P., Andraweera, N., Thilakarathne, S., & Tharanga, P (2013). Relationship between body mass index (BMI) and body fat percentage, estimated by bioelectrical impedance, in a group of Sri Lankan adults: a cross sectional study. ↩

  22. I’ve seen claims that low-intensity exercise disproportionately burns visceral fat (Brobakken et al. (2023)33), and cortisol (the “stress hormone”) disproportionately adds visceral fat. But I think even if you exercise a lot and minimize stress, you still don’t want to have too much body fat. ↩

  23. Britton, K. A., Massaro, J. M., Murabito, J. M., Kreger, B. E., Hoffmann, U., & Fox, C. S. (2013). Body fat distribution, incident cardiovascular disease, cancer, and all-cause mortality. ↩

  24. Cerhan, J. R., Moore, S. C., Jacobs, E. J., Kitahara, C. M., Rosenberg, P. S., Adami, H. O., Ebbert, J. O., English, D. R., Gapstur, S. M., Giles, G. G., & Horn-Ross, P. L. (2014). A pooled analysis of waist circumference and mortality in 650,000 adults. ↩

  25. Chen, Y., Yang, Y., Jiang, H., Liang, X., Wang, Y., & Lu, W. (2019). Associations of BMI and waist circumference with all-cause mortality: a 22-Year cohort study. ↩

  26. Jacobs, E. J., Newton, C. C., Wang, Y., Patel, A. V., McCullough, M. L., Campbell, P. Thun, M. J., & Gapstur, S. M. (2010). Waist circumference and all-cause mortality in a large US cohort. ↩

  27. Lee, S. W., Son, J. Y., Kim, J. M., Hwang, S. S., Han, J. S., & Heo, N. J. (2018). Body fat distribution is more predictive of all-cause mortality than overall adiposity. ↩

  28. Visscher, T. L. S., Seidell, J. C., Molarius, A., van der Kuip, D., Hofman, A., & Witteman, J. C. M. (2001). A comparison of body mass index, waist–hip ratio and waist circumference as predictors of all-cause mortality among the elderly: the Rotterdam study. ↩

  29. Momma, H., Kawakami, R., Honda, T., & Sawada, S. S (2022). Muscle-strengthening activities are associated with lower risk and mortality in major non-communicable diseases: a systematic review and meta-analysis of cohort studies. ↩

  30. Patel, A. V., Hodge, J. M., Rees-Punia, E., Teras, L. R., Campbell, P. T., & Gapstur, S. M (2020). Peer Reviewed: Relationship Between Muscle-Strengthening Activity and Cause-Specific Mortality in a Large US Cohort. ↩

  31. Miyachi, M (2013). Effects of resistance training on arterial stiffness: a meta-analysis. ↩

  32. Nuckols, G. (2022). What is the optimal dose of resistance training for longevity? ↩

  33. Brobakken, M. F., Krogsæter, I., Helgerud, J., Wang, E., & Hoff, J (2023). Abdominal aerobic endurance exercise reveals spot reduction exists: A randomized controlled trial. ↩