Exercise and Weight Loss: A Response to Herman Pontzer

Herman Pontzer is rekindling debates about the role of exercise in weight loss. He is a researcher of anthropology and energy metabolism, published a trendy book called BURN, and, if you YouTube his name, you’ll find many interviews with catchy titles like ‘Professor Exposes The Big Myth About Weight Loss’ and ‘The Ugly Truth About Exercise and Weight Loss’. His book subtitle is ‘New Research Blows the Lid Off How We Really Burn Calories’, which has, unsurprisingly, caught people’s attention.

Beneath the catchy titles, Pontzer is trying to make people aware of his new model of total daily energy expenditure (TDEE; classically known as “daily calorie burn”) that differs greatly from conventional opinion. Whereas TDEE has long been considered an additive model, where increasing physical activity linearly increases TDEE, Pontzer claims this model is wrong. In its place, Pontzer argues for a constrained TDEE model, where physiological adaptations constrain our TDEE to a narrow range regardless of lifestyle (including exercise levels). He says that TDEE is “more a product of our common genetic inheritance than our diverse lifestyles”, as we naturally compensate for exercise by reducing non-exercise activity expenditure (NEAT) and resting metabolic rate (RMR).

Given this new model, on pages 166–167 of BURN, Pontzer claims that “…the standard line—that exercise helps us burn more calories—is wrong” and that “[exercise] just won’t do much for your weight”. Pontzer also responded to my Twitter poll, in which I asked people whether they thought exercise increases TDEE and promotes weight loss. His firm response was that if you answer yes, you “don’t know the data”. A harsh blow to the 75% of responders who answered yes.

In any case, I think Pontzer is wrong. Not completely wrong, I’ll say, but wrong enough to write about it and enjoy the debates that follow. My position is straightforwardly this: physical activity increases TDEE and promotes weight loss. That’s a generalised claim, of course, I’m not saying it applies to literally everyone who exercises, but in terms of the average effect at a group level, when exercise levels are moderate to high, that is the position I hold with a good deal of confidence. There’s strong evidence to support it, which I’ll discuss. First, though, let’s clarify the interesting beginnings of Pontzer’s model to understand how it all came-to-be.

It All Started in Tanzania

If we go back a decade or so, we arrive at the time when Pontzer started considering this new TDEE model. He was living with Hadza foragers in Tanzania conducting an ecological study, testing whether highly-active hunter-gatherers expend more energy than the average person. He measured the Hadza’s TDEE using gold standard methods, recorded their body mass, and then compared these variables with similar data in less active populations, such as Westerners. What he found (page 158 of BURN) was that “Hadza men and women were burning the same amount of energy each day as men and women in the United States, England, the Netherlands, Japan, Russia”, who by modern standards we consider to be rather lazy on the whole. In turn, Pontzer’s model was born.

But let’s do some basic math to dampen the shock value of these findings a little bit and put them into context. First, the Westerners in this study were 30–31kg heavier than the far more active Hadza. This was not an apples-to-apples comparison with activity levels as the only differentiator. Not even close. Second, it’s quite clear from Table 1 that the far more active Hadza had a ~26% higher TDEE per unit body mass, as we would expect based on their considerably higher activity levels.

A very crude analysis, I admit (we’ll get to the complex stuff), but it certainly doesn’t follow expectations if exercise truly has little to no effect on TDEE. It follows what we would expect: people who exercise a lot expend more energy relative to their body mass than people who do not. And larger people who are not very active will probably have similar TDEEs to smaller people who are very active. Do we need a new TDEE model to explain this? I don’t think so.

I’ve mentioned this finding to Pontzer and he largely dismisses it without much explanation. Instead, he draws attention to another analysis in the study – one that found no significant differences in TDEE between populations when you adjust for (i.e. remove the effect of) fat-free mass. But, in my opinion, this analysis is flawed, as it presents a fundamental misunderstanding between confounders and mediators in data analysis, and when you should and should not adjust for different variables. Pontzer told me “the distinction between confounders and mediators is largely conceptual” and “They would be treated the same in stats analysis such as the one used”, but any statistician would argue the differences are defined and vitally important. Confounders interfere with the causal pathway we’re interested in; mediators are part of the causal pathway we’re interested in. Hence, when statistically adjusting and removing the effect of different variables, we do this for confounders, not mediators – thus removing the interference.

In this case, fat-free mass should have been treated as a mediator between lifestyle and TDEE. This would assume lifestyle has both a direct and indirect effect on TDEE. Direct by expending more energy via activity; indirect by influencing levels of fat-free mass (by far the biggest determinant of TDEE). That sounds reasonable. Instead, however, Pontzer treated fat-free mass as a confounder of the relationship between lifestyle and TDEE. Meaning, he treated fat-free mass as a cause of lifestyle rather than a product of it, which is highly unlikely to be true. That would be like arguing big muscles cause people to exercise, instead of exercising causing big muscles. It doesn’t make any sense. But because he chose to do the analysis this way, he basically removed the biggest determinant of TDEE and the biggest characteristic difference between these populations from the equation, without justification.

In addition, even if we want to zone in some of the faulty analyses in this study, well, they weren’t all in favour of Pontzer’s position anyway. In the supplementary data, there is an analysis that compared differences in the average TDEE between populations (that is, farming populations, hunter-gatherer populations and Market [Western/Bolivian] populations) when adjusted for body mass instead of fat-free mass. It found that lifestyle differences between populations were a significant predictor of TDEE (P=0.02). Why was this not reported in the main text?

Then We Have the Experimental Data

Obviously Pontzer doesn’t rely only on the Hadza study to support his narrative that exercise is of little to no use for weight loss. In BURN, he has a few go-to exercise intervention studies to support his claims too. Most notably, he claims that exercising for one year or more leads to a measly 0–2 kg of weight loss, on average, with reference to a 2014 narrative review by Swift et al. Reader’s will rightly think this sounds pretty poor. But let’s inspect the review in question, which references four exercise studies to support this 0–2 kg figure.

In the first two studies, The Studies of a Targeted Risk Reduction Intervention through Defined Exercise Study (STRRIDE), and The Diabetes Aerobic and Resistance Exercise Study (DARE), all participants were councelled by dietitians to ensure they would not lose weight during the exercise intervention. Consequently, these studies are borderline useless in discussions of exercise and weight loss. Although, side note, there was still significant weight loss (P<0.05) in the exercise groups compared to the sedentary control groups in both studies.

In the next two studies, The Dose Response to Exercise in Women Study (DREW), and The Inflammation and Exercise Study (INFLAME), the exercise intervention was so modest that TDEE and weight loss were predictably small even based on the additive TDEE model. The exercise groups in both studies were expected to increase TDEE by just 48–199 kcal, and lose just 1.1–3.1 kg. These figures convey the main problem: small change, small result. Therefore, zoning in on the small reported weight loss in these studies (0.4–2.4 kg) doesn’t bring anything particularly interesting to the table. Weight loss was expectedly low based on the additive TDEE model.

Also, what’s not mentioned in BURN is that the Swift et al. review has a full section explaining why weight loss is often small in exercise trials. The authors point toward an analysis of 15 aerobic exercise interventions from Thomas et al., saying the researchers “concluded that the major factors limiting the expected weight loss from aerobic [exercise training] were dietary compensation and low aerobic [exercise training] dose.” Two factors that have nothing to do with TDEE constraints. Swift et al. also clarified that weight loss is considerably greater in studies that use higher exercise doses. They cite the Midwest Exercise Trial (MET-1), which observed 5.3 % weight loss after 16 weeks of aerobic exercise with an average TDEE increase of ~286 kcals. They also cite a couple of exercise intervention studies from Ross et al., which observed 6.8–8% weight loss in 12–14 weeks with an average TDEE increase of 500–700 kcal. So, while simply exercising obviously doesn’t necessarily cause massive changes in TDEE and weight loss, I think it’s fair to say the benefits are clinically significant when the exercise dose is sufficient. Not for the first time, the devil is in the dose.

But What About Long-Term Exercise Interventions?

Moving on, what I found most interesting while reading BURN is that Pontzer actually cited some long-term (9–16 months) exercise studies using higher exercise doses (>200 kcal sessions per day) to solidify his claims. However, these studies actually offer pretty clear contradictory evidence. These studies were Westerterp et al., MET-1, and the Midwest Exercise Trial 2 (MET-2).

Here’s a summary of TDEE in these studies:

• For Westerterp et al., in the participants who completed the study, there were significant increases in the TDEE of men and women by 669 kcal (P<0.01) and 430 kcal (P<0.05), respectively. These figures aligned closely with predicted TDEE increases based on the additive TDEE model, 205–625 kcal (calculated by using mean bodyweight in men and women and assuming ~5 kcal expenditure per lb of bodyweight per hour of running).

• For MET-1, male and female exercisers increased their TDEE by 371 kcal and 209 kcal, respectively, again not far off the predicted TDEE increases of ~471 kcal in males and ~314 kcal in females. (actual TDEE increases were 66–78% of predicted values from the additive TDEE model.) No within-group significance testing was done for TDEE; however , the researchers noted that “changes in 24-hour energy expenditure were significantly related to the energy expenditure of exercise”.

• For MET-2, moderate exercisers increased their TDEE by a nonsignificant 191 kcal (P=0.10), and high exercisers increased their TDEE by a significant 289 kcal (P=0.01). As the predicted TDEE increase was ~286 kcal and ~429 kcal in the moderate and high exercise groups at month 10, the actual TDEE increases were roughly ~66% (in both groups) of the predicted values from the additive TDEE model.

As a summary of these studies, then, I cannot reconcile the position that exercise has little to no effect on TDEE. When the exercise dose is moderate to high (200–600 kcal sessions), there are clear increases in TDEE (191–669 kcal) that are no less than 66% of predicted increases from the additive TDEE model. Yes, some energy compensation might exist, but this doesn’t stop exercise from causing meaningful increases in TDEE. And it achieves this while participants are losing weight, might I add (as per next section), which is a process that is typically predictive of reductions in TDEE due to the loss of metabolically active tissue.

Now let’s summarise the weight loss results in these trials…

• For Westerterp et al., exercising men lost a significant 1.8 kg of bodyweight (P<0.01) and exercising women lost a nonsignificant 2.0 kg of bodyweight (P>0.05). Granted, not crazy amounts here by anyone’s standards, but consider that almost all participants were already in the healthy BMI range (19.4-26.4 kg/m²). This wasn’t a weight loss study and there simply wasn’t much weight to lose.

• For MET-1, male exercisers lost a significant 5.2 kg (P<0.05) and female exercisers gained a nonsignificant 0.6 kg (P>0.05; 100% FFM). The latter result may come as a surprise, which it is, but consider that the female control group gained a significant 2.9 kg (P<0.05) by not exercising. In other words, male exercisers had clinically significant reductions in bodyweight, while exercising females were lighter than if they had remained sedentary. Thus, at the end of the exercise trial, there were significant 4.8–5.2 kg differences (P<0.05) in the bodyweight of both male and female exercisers compared to the sedentary control group.

• For MET-2, the average weight loss from exercising was clear across the board. Males and females in the moderate exercise group lost between 3.9–4.1 kg, and males and females in the high exercise group lost between 4.4–5.9 kg. Both exercise groups had significant differences in bodyweight change compared to the sedentary control group (P<0.01), who put on 0.5 kg. The high exercisers also reduced their bodyweight by a nonsignificant 2.1 kg more (P>0.05) than the moderate exercisers by expending an additional ~143 kcal per day.

As a summary of these studies, then, alike to my view on exercise and TDEE, I cannot reconcile the position that exercise has little to no effect on weight loss. When a sedentary control group was included, there were significant differences in bodyweight change favouring the exercise group by 2.3–6.2kg. Compared to baseline, weight loss in the exercise group ranged between 1.8–5.9 kg in 7 of the 8 relevant analyses across the three exercise trials, trending toward the higher end of this range in participants with overweight/obesity. There was only one within-group result that found no weight loss from exercise, which happens to be the finding that Pontzer emphasises the most in his writings. That’s problematic.

Let’s also clarify that these are the primary studies that Pontzer chooses to support his position. So if these aren’t even remotely supportive, then what is? I wouldn’t find it so hard to support my position. I’d cite something like Whatley et al. and just wipe the argument out for good. It showed 19.6kg of weight loss in 12 weeks with 150–400 weekly minutes of exercise alongside a very-low calorie diet, 6.5kg more than the diet alone. That would have made for a really short and slightly boring article, though, wouldn’t it?

Kill the Myth Before it Catches On

Ok, right now, we’re at a sticking point. Most people still think that exercise increases TDEE and promotes weight loss; however, the poorly supported idea this is a “myth” is getting more and more popular. Please don’t let it catch on. While I agree with Pontzer that some energy compensation in response to exercise might occur (look out for some future deep dives), his narratives on TDEE and weight loss are most likely wrong. When exercise doses are respectable and are expected to lead to weight loss, they almost always do. It’s really that simple. Time to go for a run.