Much research has been done on personality traits associated with eating disorders, and, as I’ve blogged about here and here, on personality subtypes among patients with EDs. For example, researchers have found that individuals with AN tend to have higher levels of neuroticism and perfectionism than healthy controls (Bulik et al., 2006; Strober, 1981). Moreover, some traits, such as anxiety, have been associated with a lower likelihood of recovery, whereas others, such as impulsivity, with a higher likelihood of recovery from AN (see my post here).
Personality refers to “a set of psychological qualities that contribute to an individual’s enduring and distinctive patterns of feeling, thinking and behaviour” (Pervin & Cervone, 2010, as cited in Atiye et al., 2014). Temperament is considered to be a component of personality and refers to, according to one definition,”the automatic emotional responses to experience and is moderately heritable (i.e. genetic, biological) and stable throughout life.”
One popular model for classifying temperamental traits was developed by Cloninger (1987) and consisted of three dimensions (novelty seeking, harm avoidance, and reward dependence). The model has been updated with the addition of persistence as another dimension of temperament (Cloninger, Przybeck, & Svrakic, 1991). These four traits (described below, with the exception of persistence) are measured by the Temperament and Character Inventory (TCI):
Novelty seeking is a tendency for intense excitement or exhilaration in response to novel stimuli or cues for potential rewards or potential relief of punishment, harm avoidance is a tendency to respond intensely to signals of unpleasant stimuli, which leads to inhibition behaviour in order to avoid punishment, novelty or frustrating non-reward and reward dependence is a tendency to respond intensely to signals of reward, which leads the subject to maintain or resist the extinction of behaviour that has previously been related to rewards or relief from punishment.
While a lot of research has been done on temperament in eating disorder patients, there has not been a comprehensive summary of the findings. Consequently, the authors of this paper, Minna Atiye and colleagues, sought to utilize a meta-analytical approach to summarize the studies published to date studying temperamental traits (using Cloninger’s model) in ED patients.
The authors searched for studies on EDs using Cloninger’s temperament dimensions. There were 14 case-control studies that fulfilled the criteria, with 8 studies including more than one patient group. There was a total of 3315 cases (i.e., individuals with an ED) and 3395 controls:
- 11 studies had AN patients (446 cases)
- 7 studies had AN-R patients (345 cases)
- 7 studies had AN-BP patients (471 cases)
- 12 studies had BN patients (1485 cases)
- 2 studies each had EDNOS (185 cases) and BED patients (383 cases)
Men were included in 4/10 studies, comprising 2.2% of the cases and 5.3% of the controls (72 and 179 individuals, respectively). Two studies included individuals who had recovered from an ED (295 individuals).
Comparisons between individuals with and without EDs
The table below summarizes the pooled effect sizes for differences in the four temperament dimensions between individuals with and without EDs. For simplicity, I excluded statistically non-significant effect sizes greater than 0.2, which were included in the original table. (More on interpreting effect sizes here).
As you can see, novelty seeking is significantly higher than in controls on in individuals with BN (d=0.41). On the other hand, compared to controls, harm avoidance was elevated among all diagnostic groups. Harm avoidance was highest in the AN-R and AN-BP groups (d=0.76 for both).
Reward dependence did not differ between individuals with EDs and controls for any diagnostic groups. Finally, persistence was elevated in individuals with AN, with the highest effect sizes in those with AN-R (d=0.52).
Comparisons between ill and recovered individuals
Only two studies compared ill and recovered individuals (total number of cases: 159). Differences were observed only in the AN group, with harm avoidance being significantly lower and reward dependence significantly higher in those who had recovered compared to those who were ill.
WHAT DO THESE FINDINGS MEAN?
Before I get into what the findings mean, I think it is important to point out that the majority of the studies included individuals with “current or lifetime” eating disorders. I suspect the vast majority of the studies were done when patients were ill. This, of course, raises the question: Are these traits premorbid, and do they predict ED onset, or are they somehow altered because of the ED itself?
As mentioned above, individuals with AN had the highest scores among the diagnostic groups on harm avoidance and (specifically for AN-R) on persistence. The authors suggest that this may mean individuals with AN have “an even stronger tendency towards fearfulness and worry” (harm avoidance) as compared to the other diagnostic groups and may explain why individuals with AN have “a strong tendency to maintain behaviour despite frustration and intermittent reinforcement” (persistence).
With regard to the high novelty seeking scores among the BN group, the authors suggest that this is “in line with established observations of impulsive and borderline traits” among individuals with BN. Although I did not include it in the graphic, the authors found a nonsignificant reduction of novelty seeking (relative to controls) among those with AN.
Regarding differences between ill and recovered individuals, Atiye et al. write that these findings “suggest an improvement in social interactions and an alleviation of anxiety along with recovery from AN.” Maybe, maybe not (see limitations below).
There are important limitations to consider when interpreting these findings. First, the results for BED and EDNOS come from only two studie. The results for EDNOS are especially tricky to interpret because the diagnostic group is very heterogenous, and the results for BED come from only 185 cases, compared to, for example, 1485 for BN.
Second, there were only two studies that compared temperamental traits between ill and recovered individuals. Importantly, these comparisons did not look at the same group of individuals over time, assessing them when they were ill and then again when they were deemed to be recovered. Instead, they took individuals who were ill and compared them to a different group of individuals who were recovered. Thus, we don’t know whether the process of recovery leads to lower harm avoidance and higher reward dependence OR that individuals who have lower harm avoidance and higher reward dependence are more likely to recover. This is an important question, and one that requires longitudinal research.
Third, there’s evidence of a publication bias (also called the file-drawer effect) in three comparisons: harm avoidance in AN and BN, and persistence in AN. Publication bias refers to the fact that studies that do not find a positive effect are less likely to be published. So when researchers decide to do a meta-analysis and combine all of the published research on a topic to see where the consensus lies, they may be misled and conclude that the effect they are seeing is greater than it really is in reality. What does this mean here? It means that the differences between individuals with EDs and healthy controls on harm reduction in AN and BN and persistence in AN might not be as big as we think. It is hard to say.
We need longitudinal studies. Longitudinal studies will not only help us understand what, if anything, happens to temperamental traits during and after recovery, but also what, if any, traits can predict recovery. In the future, we might be able to use this information to also predict what treatment approaches might work best for individuals depending on their temperament.
More interesting to me, however, is that longitudinal studies might be able to tell us what happens regarding temperament when individuals cross over from one diagnosis to another. This happens fairly frequently (more for some diagnostic crossovers than others, though), as I’ve blogged about before. And I have often wondered what this high rate of diagnostic crossover means for studies that evaluate individuals with EDs at a single time point and then compare different diagnostic groups, whether they compare prevalence of comorbid psychiatric disorders, temperamental or personality traits, or anything else.
So, how different are individuals with AN compared to those with BN really? And, how much of the differences that we see are due to premorbid differences and how much are they byproducts of the current (ill) state? I am not claiming there are not differences — there are, of course there are. Certainly, many people do not crossover, and crossover rates differ depending on initial diagnosis. Still, is it possible that we are overestimating the differences between these diagnostic groups? And how elastic, or state-dependent, are temperamental and personality traits (or are the differences/changes we see due to imperfect measuring instruments)?
On this, the authors write,
The general assumption posits that temperament is relatively stable over time (Goldsmith et al., 2000), but significant changes in psychiatric disorders in response to state effects have also been reported in longitudinal designs (Abrams et al., 2004; Kampman et al., 2012).
I’m looking forward to potentially seeing some longitudinal studies addressing these questions soon. Readers, what are you thoughts? In particular, do you feel your temperamental traits changed throughout the eating disorder? If so, do you feel the temperamental changes led to changes in the ED “status” (i.e., from illness to recovery, or a diagnostic crossover) or vice versa?
Atiye, M., Miettunen, J,, & Raevuori-Helkamaa, A. (2014). A Meta-Analysis of Temperament in Eating Disorders. European Eating Disorders Review PMID: 25546554