How many calories do patients with anorexia nervosa need to eat to gain a kilo (2.2 lbs)? It seems like a simple question and one that we should have figured out a long time ago, given the importance (err, necessity) of refeeding and weight restoration in recovery from anorexia nervosa.
Unfortunately, research in this area has often led to contradictory results (see Salisbury et al., 1995 and de Zwaan et al., 2002 for reviews). Fortunately, a paper by Stephan Zipfel and colleagues (2013, freely available here) sheds light on one potential cause of the discrepancies.
But first, some definitions:
TDEE stands for total daily energy expenditure. TDEE has three components: resting energy expenditure (REE), dietary-induced thermogenesis (DIT), and activity-induced thermogenesis (AIT). The gold standard for measuring TDEE is through something called the doubly labelled water technique. REE is usually measured through indirect calorimetry. (These techniques were used in this study as well.)
Generally, when people lose weight, their REE is also reduced as a result of decreased body weight and decreased lean mass, as well as metabolic adaptations. Since REE contributes significantly to the total energy needs, we’d expect it to also decrease in AN patients. However, previous studies on this are contradictory; some have found decreased TDEE to be decreased in AN patients whereas others have found it to be similar to healthy controls (see de Zwaan et al., 2002).
Unsurprisingly, the amount of excess calories (calories in addition to those calculated to be required for basic survival) needed to gain a kilo (2.2lbs) have also varied widely between studies: between 5340 and 9768 (Forbes et al., 1984; Dempsey et al., 1984, respectively).
Part of the discrepancy might have arisen from the use of different techniques to measure basal calorie requirements, but even in studies using the same technique (indirect calorimetry), the excess calories necessary to gain a kilo differed by more than 2,000 (Russell and Mezey (1962) versus Dempsey et al. (1984)).
Clearly, some else must be going on. As Zipfel et al. point out, previous studies did NOT distinguish between patients with high- and low-activity levels. It is possible that the discrepancies in energy requirements and excess calories required to gain a kilo arise, at least in part, from this.
To assess the effect of exercise on energy metabolism, the researchers recruited 12 AN patients (all females, average age ~22 years, average duration of illness ~4.5 years) and 12 healthy controls. All AN patients had amenorrhea; 8 had AN restrictive subtype and 4 had AN binge-purge subtype.
There were no differences in the amount of daily exercise between the AN and control groups. However, differences emerged when the researchers divided the groups into high- and low-level exercisers. Seven of the AN participants were categorized as “high-level exercisers” compared to only two in the control group. (All the binge/purge AN patients were in the high exercise group.)
The table below illustrates the breakdown for BMI, lean body mass (body mass – fat mass), and duration of exercise per day.
As you can see, the high-level group exercised 3-4x more than the low-level exercisers and the controls. They had slightly higher BMI and lean body mass values but these were not significantly different from the low-level exercisers.
THE MAIN FINDINGS
Here’s where the interesting part comes in. When Zipfel et al. compared REE between the AN group and the controls, there was a significant difference — as you’d expect given that, for one, there are significant differences in BMI between the AN and control groups.
However, when the REE was adjusted for body surface area or lean body mass, those differences disappeared. What that means it that, taken as a group, a kilogram of lean body mass (lean body tissue is much more metabolically active than fat tissue), AN and controls in AN patients was as metabolically active as in the controls. It had the same energy needs.
Then the researchers split up the AN group into high- and low-exercisers and did the comparisons again. I’ve illustrated the main findings in the two figures.
The main finding: Resting energy expenditure in high-level AN exercisers was NOT significantly different from healthy controls. This means that AT REST, high-level AN exercisers require, on average, the same amount of calories as normal-weight healthy controls. This is despite the fact that there were significant differences in lean body mass, leptin levels, and T3 levels.
Unsurprisingly, REE in the high-level AN exercisers and the healthy controls were significantly higher than in the low-level AN exercisers group.
There were also NO differences in the TOTAL energy expenditure between high-level AN exercisers and healthy controls, but both expended nearly twice the amount that low-level AN exercisers did.
When the researchers adjusted the REE for lean body mass, they found that high-level AN exercisers actually had higher resting energy expenditure per kilogram of lean body mass (see Figure 2 below).
This makes sense because the similar overall REE levels have to be driven by something and we know that both high- and low-level AN exercisers have significantly less lean body mass tissue than the healthy controls. (Of course, it is not just lean body mass that could be causing the high REE, but as I said, it is a significant factor, so it is not surprising.)
While the researchers don’t know what’s causing the differences in the REE between high-level and low-level AN exercisers, they do hint at where we may want to start looking for them,
[N]o differences in thyroid hormone levels were observed between the AN subgroups. In fact, two individuals of the same LBM can differ considerably in their REE due to differences in heredity and a range of physiological factors . The elevated REE in high-level AN patients is possibly caused by a secondary effect resulting from the impact of physical activity on a range of physiological processes that in turn influence REE [34,35]. Additionally it is also conceivable that the body core temperatures within the AN subgroups are slightly different 
Zipfel et al. also found that high-level AN exercisers had significantly higher driven for thinness and depression scores compared to the low-level AN exercisers. It will be interesting to see if this holds up in a larger study as there was quite a bit of variability in the scores, especially for the depression questionnaire.
One of the obvious limitations is the small size, although I don’t actually think this is a big problem in this study given the stark differences in the variables of interest. A higher number of participants would likely decrease the variability within the groups but I am not sure it would change any of the relationships (i.e., similarities and differences) as far as the REE and TDEE are concerned. Larger studies are needed, however, to better understand what’s happening on a biological (and psychological) level.
The other limitation that the authors point out is that, outside of a structured interview, they did not assess the types of activities that the participants engaged in. Thus, they were not able to differentiate between, for example, short bouts of intense exercise or near-constant fidgeting, for example. This would be interesting to assess in future studies.
CONCLUSION & CLINICAL IMPLICATIONS
Taken together, our study shows that there are marked differences in energy requirement among AN patients. Over one-half of our AN patients reported high exercise levels directly related to weight control with an overrepresentation of the binge/purge AN subgroup.
I think one of the most clinically relevant findings of this study was that the researchers found that the question “What percentage of your exercise is aimed at controlling your weight?” was able to utilized to distinguish between low-level and high-level exercisers and thus significantly correlated with total energy expenditure when the cutoff was set at <50%.
I think this is pretty neat because, if its predictive value holds up in larger studies, clinicians can use a single question to better assess the energy requirements of their patients and thus better predict the calories required for weight gain. Asked directly, the patients may under-report, so an indirect question, like the one above, would be ideal to “solve this problem” (as the authors put it).
(Please note that I used “calories” throughout the post and in the figures for simplicity’s sake/because of its common usage but, strictly speaking, I’m really referring to kilocalories.)