Why Does Recovery Not Seem to Help With Mental Functioning?

We’ve seen (in part three of this series) strong evidence that cognitive capacities improve with recovery from anorexia and also (in part four) some counterevidence that they don’t. Here we’ll dig into what’s happening with results suggesting that recovery makes no difference.

Let’s start by looking at how these studies and reviews define the improved state that’s being compared against the starting state. Here are the criteria for the six I cited in part four:

• Talbot et al. (2015): “weight recovered” = participants “had a lifetime history of AN and had maintained a BMI of >18.5 for a minimum of eight weeks”; “fully recovered” = participants “met the criteria for weight recovery, no longer met DSM-IV-TR criteria for AN, had no eating disordered behaviors (binging, purging, restricting, and driven or compulsive exercise) over the past three months (at minimum) and had scores on all sub-scales of the EDE-Q within one standard deviation of population norms”
• Fürtjes et al. (2018): “weight restoration” = “a minimum of 10 percent increase in BMI.”
• Miles et al. (2020) [review]: “fully recovered” = “participants were required to: (a) have a past diagnosis of AN, (b) have a current BMI over 18.5, and (c) engage in no bingeing, purging or restrictive behaviours for at least three months” (most required it for 12 months).
• Seidel et al. (2021): “after weight increase” = “A subsample of patients with AN (n = 13) were also assessed before discharge (between baseline and follow-up measurement, which was on average after 100 days; BMI increased, on average, by 2.01).”
• Tenconi et al. (2021): “clinical improvement” = “(1) an increase in BMI by at least three percentile points; (2) a general improvement in eating patterns in terms of pace, amount, and variety of foods accepted and regularly taken; (3) a significant modification in global functioning (i.e., restoration of school/work attendance, if previously interrupted, resumption of social relationships and exchanges); (4) a reduction in depressive and anxious symptoms as detected by self-reported assessment, but also as reported by both the patient him/herself and significant others.”
• Hemmingsen et al. (2022) [review]: “following weight gain” = “Follow‐up mean BMI varied from 14.9 to 20.6.”

This sample illustrates the two main approaches to before/after measures: focus on weight restoration alone (4/6) or go for something broader (3/6) (Talbot et al. do both). The first route sounds like it ought to be simple, but there’s no obvious single generalized criterion for what should count as enough weight gained. In practice, the outcomes are the result of a competition between 1) clinical and commonsensical ideas about what is ideal versus 2) participants’ ambivalence about weight gain and the pragmatics of either helping individuals to achieve that ideal or recruiting individuals who already have. (It’s a lot quicker and easier to support people to reach—or locate people who have already reached—a BMI of 18.6 than the BMI at which they naturally stabilize without reintroducing restriction.)

It seems that pragmatics won. In all but one case, the thresholds are laughably low, and it therefore isn’t surprising that little is found to have changed cognitively. This would be reasonable if the claims were merely that “if you do a tiny bit of weight restoration, nothing much will change yet.” But that’s unlikely to get your paper published. And by using phrases like “after weight restoration,” quite a different impression is conveyed. No one explicitly says “after weight restoration adequate to each individual”, but that’s the conveniently unspoken implication.

Expedient equivocation is everywhere. Seidel et al., for example, don’t actually *say* that the weight gain that’s happened is the amount that needed to happen, but phrases like “at the follow-up time point after weight gain” certainly make it easy, for someone who’s not reading very carefully, to come away with the impression that this was pre- and post-weight restoration as a complete process.

Their section on study limitations made no mention of the meagre difference between pre and post, instead drawing the conclusion that would have been justified if they’d done things properly: “we could not find a correlation between cognitive performance and BMI, which may speak against an impact of BMI (or stage of weight restoration) on the domains of cognitive performance measured in the study.”

Others do admit limitations. Hemmingsen et al.’s review, for instance, includes the acknowledgement that “Follow‐up mean BMI varied from 14.9 to 20.6, indicating that some patients were still severely underweight at follow‐up whereas other patients reached weight recovery."

When using “after” thresholds with broader bases than bodyweight alone, there’s a lot more latitude. You can take the easy route, as Miles et al.’s review does, and allow a minimal definition of “fully recovered” that will be very easy for many not-at-all recovered individuals to meet. The authors flag the problem of inconsistent and overly narrow definitions of recovery, but their own definition only contributes to it.

Or you can do things properly and go for something a lot more ambitious, as Talbot et al. do, adopting the criteria proposed by Bardone-Cone et al. (2010), which were intended to ensure that “recovered” individuals would “appear indistinguishable from healthy controls (defined as having no history of an eating disorder) on indices reflecting behavioral and psychological aspects of eating disorders.” Their fully recovered participants had an average BMI of 21.6 (compared to 21.8 for the healthy controls), and their EDE-Q scores actually showed a (nonsignificant) trend towards being even better (i.e. less indicative of any ED-like symptoms) than the controls—a lovely example of the very real possibility of achieving “110% recovery”.

Sadly, this kind of stringent design is still the exception. In Tomba et al.’s (2019) meta-analysis of 31 studies (about half those included in the review as a whole), large effect sizes indicated that remitted/recovered patients reported significantly lower BMI and significantly greater symptomatology than healthy controls, independently of remission and recovery criteria strictness, age, and study quality. That is, they weren’t remotely recovered on any definition anyone is likely to aspire to for their own future.

Eating Disorders Essential Reads

How Children with Eating Disorders Are Groomed Online

Reevaluating Military Fitness Standards

And the major difference doesn’t deter the authors from drawing the patently unjustified conclusion: that there is “presence of comorbid psychiatric symptoms, general psychopathology, dysfunctional personality traits, neurophysiological alterations, cognitive deficits, as well as compromised quality of life and positive functioning in remitted and recovered ED patients” (my emphasis).

People who write papers are human; they cut corners, don’t read things carefully enough, and massage inconvenient details. Untruths get turned into supposed truths by lazy referencing. One striking example is Dulawa’s (2021) conclusion that “Reduced cognitive flexibility is likely an endophenotype of AN since this phenotype is also observed in healthy siblings of patients with AN, is independent of current body weight, and persists after recovery from AN”.

She is citing the Holliday et al. (2005) study I mentioned in part 3, quietly letting their “normal weight = BMI over 17.5” definition mean that individuals who achieve this are “after recovery”. This kind of thing happens all over the place, and it makes it extremely hard to grasp the true state of the results without spending a lot of time and effort reading with a fine tooth comb.

It may seem too neat a story that every study that’s inconclusive or finds little/no improvement is working with people who aren’t very recovered yet, but sometimes simple is right. And BMI is always a good place to start if you want to do some basic fact-checking: not because it means everything, but because 1) for researchers and clinicians, it’s such an easy “before” versus “after” demarcation and 2) for someone with a restrictive eating disorder, it means far more than it should, which makes it the standard brake on the whole recovery endeavour. As I argued here, if you didn’t have to get heavier and fatter to recover from anorexia, everyone would do it.

As for the sporadic findings that people who are ill do better than those who aren’t—well, Occam’s razor suggests one obvious hypothesis: If you measure enough things (and this field has a lot of people measuring a lot of things), you’ll get some weird results.

In the next (penultimate) part of the series I explore the compensation effects and other factors that might be in play when we find that anorexia seems not to make things worse—as well as touching on the question of whether it’s really legitimate to use value-laden terms like “deficit” and “impairment”. Read on here.