Estimated intelligence quotient in anorexia nervosa: a systematic review and meta-analysis of the literature

Eating disorders (EDs) are a group of psychiatric disorders with a lifelong course and considerable morbidity and mortality. In the Diagnostic and Statistical Manual of Mental Disorders, fourth edition (DSM-IV), EDs include anorexia nervosa (AN), bulimia nervosa (BN), and eating disorders not otherwise specified (EDNOS). The aetiology of EDs remains unknown.

There has been increasing interest in the study of the neuropsychological functioning of people with EDs for the last three decades. One of the reasons for this is to gain a better understanding of the aetiology and maintenance of these disorders and to explore ways of improving available treatments [1].

Although there are more than 100 papers on neuropsychology and brain imaging in EDs, in previous systematic reviews conducted by Roberts et al. [2] and Lopez et al. [3] on set shifting and central coherence in EDs, it was demonstrated that there are a limited number of neuropsychological studies on BN and those recovered from the disorder. Most studies have explored neuropsychological functioning in AN. Because there is more available research in neuropsychology in AN, the present systematic review focused on AN only.

AN is the most severe ED affecting mainly, but not only, young women, and has the highest rate of mortality linked to a psychiatric disorder, due to high levels of medical complications and suicide in chronic patients [4].

Currently, psychological therapy is the main therapeutic intervention recommended for the treatment of AN (see, for example, [5]). However, outcomes are far from satisfactory with only 50% of cases reaching recovery in adult populations [4, 6]. It is possible that factors such as cognitive functioning, usually limited during the acute phase of AN [7], influence treatment utilisation and outcome. For instance, most available therapies require the patient to have some level of insight and verbal ability; factors that might be related to optimal intellectual level [8].

The intelligence quotient (IQ) represents a composite score on a variety of tests designed to measure a hypothesised general ability or intelligence [9]. It has been hypothesised that people with AN have a higher IQ level than the general population based on clinical and school performance observations. There is a suggestion, however, that higher perfectionism, but not higher IQ, would explain the better performance at school in this group [10, 11].

As neuropsychological studies have become more popular in AN, there are several studies that have measured intelligence as part of their assessments. However, there are a limited number of studies looking specifically at IQ levels in people with AN. We were not able to find any systematic review summarising research on intelligence in AN. Therefore, the question about IQ in AN remains unanswered. It is expected that examining the available literature in this area would be helpful in providing information about intellectual functioning in AN, investigating how comparable neuropsychological studies from different academic groups are in the context of IQ estimation, helping to clarify what the most appropriate IQ measure would be for future studies, and exploring any association between severity of illness (measured by Body Mass Index (BMI)) and IQ levels.

With these questions in mind, a systematic review of the literature and meta-analysis with the available data on IQ in AN were conducted to address the hypothesis that people with AN show superior scores on well validated IQ tests in comparison with the average IQ of the normative population (norm). This hypothesis is based on the common but conflicting assumption that high intelligence is a trait among people with AN [12, 13].

We have also predicted that people with a past history of AN, now recovered, would demonstrate higher IQ scores than those who are in the acute phase of AN and the normative population. We based this hypothesis on the fact that it was shown that IQ predicts termination from treatment; that is, patients with a higher IQ are more likely to remain in psychological treatment [14].

After careful assessment on a case-by-case basis, 30 studies (including 849 AN patients in total) were found using the above-mentioned criteria. All of the identified articles used either the NART (N = 14) or Wechsler's derived tests (WAIS, WISC, short forms of WAIS and HAWIK tests) (N = 16).

The aim of this study was to provide a comprehensive systematic review of the literature including a meta-analysis for estimated IQ data in AN compared with the average IQ obtained from normative data. Studies included in this review were selected on the basis of quality of the data and validity of instruments used to estimate intellectual ability. Thus, NART and Weschler's scales were selected as the most common and reliable measures.

It was established that most people with AN have higher average IQ scores compared with the average of normative data. Specifically, studies using the NART consistently showed a higher IQ in AN patients in comparison with established norms. However, those using Wechsler's scales obtained more heterogeneous results, with half of the studies showing moderate to high effect sizes (average ES = 0.75) and half of them with low or negligible effect sizes (average ES around 0). Examination of the characteristics of the various studies (for example, age at testing, BMI, reported comorbid conditions, and so on) was not able to explain this heterogeneity. These results overall show that people with AN have at least as high IQ as norms, which indicates a difference, compared to other psychiatric conditions.

There are two points that are worth highlighting in this discussion about the heterogeneity of results. Although the WAIS and NART are highly correlated, the NART estimates premorbid IQ and Wechsler's scales measure current ability. Predictions could be made from NART scores about performance, verbal and total IQ based predominantly on verbal abilities, which are thought to be generally preserved in AN [1, 7], whereas Weschler's scales assess mixed verbal, performance and visual spatial abilities. The latter have been reported to be more impaired in people with acute AN (see, for example, [7]). It might be possible that those studies using Wechsler's scales with lower effect sizes are demonstrating differences in the performance of the samples that are not evident in full scale IQ data (for example, differences in performance versus verbal IQ or lower scores in scales involving visual spatial abilities). We are not able to clarify this point with this review, as most studies provided full scale IQ only, as previously mentioned. Also, Wechsler's scales used in the studies included in this review are composed of different scales (for example, WAIS, WISC, short version, and so on) and there was no consistent use of one single instrument.

Both premorbid and current IQ yield valid and interesting information for future studies and the measure selected will depend on the nature and objectives of future studies. Therefore, it seems that simple and reliable measures such as the NART provide more consistent data on IQ, without the bias that anomalies in cognitive performance (typically present in the acute phase of AN) may introduce in test performance. Also, from the available literature, it is still hard to draw firm conclusions regarding performance and verbal IQ. Clear reporting in future studies (for example, separating verbal from performance IQ) will help to address this point.

A coordinated approach and consensus of IQ measures in the field will make data more comparable and will provide better insight into the relationship between illness severity and the neuropsychological profile of AN. For example, meta-analyses and systematic reviews conducted in schizophrenia allowed researchers to identify that, before the onset of psychosis, IQ scores are approximately 0.5 standard deviations below that of healthy comparison subjects, and low IQ could be considered as one of the risk factors for schizophrenia [33, 34].

In order to explore the question about the potential contribution of IQ as a predictor of recovery, we made an attempt to review studies on recovered AN populations. Only four studies [35‐38] reported IQ data on women with a past history of AN. We did not have efficient power to draw strong conclusions from the available studies. However, a meta-analysis showed that people who had recovered from the illness had higher IQ in comparison to norms and studies on acute AN groups, included in this paper. There is a growing literature suggesting that IQ level can predict treatment outcome in psychiatric conditions such as schizophrenia and autism (see, for example, [39]). We predict that this line of research will also be highly informative for the ED field. For example, it was reported that higher IQ predicts completion of psychological treatment in early studies [14]. To our knowledge, none of the treatment studies conducted in EDs looked at drop-out data in the context of IQ. In general, it is clear that all treatment studies in AN have high drop-out rates [40]. From the results of this study, it seems likely that successfully treated patients with AN have higher premorbid IQ, which would support the hypothesis of the higher the IQ, the better the treatment prognosis. However, more studies will be needed to confirm this prediction and which other factors may be involved (for example, environmental, physiological, and so on).

We believe that this systematic appraisal of the literature was helpful in highlighting a trend that suggests that people with AN have average or higher than average IQ (in both NART and WAIS studies). The benefit of this appraisal of current knowledge will help researchers in planning future studies and formulate important questions, such as: do patients with higher IQ have better prognosis? How could high IQ be effectively used in psychological treatment? Is IQ decline evident in AN? Whereas the most obvious reason will be malnutrition, none of the studies so far have used premorbid and current IQ measures simultaneously.

Finally, this study helped us to reflect on the fact that comparison clinical groups should be carefully selected in future studies on AN, because IQ will be an important contributing factor in social cognition, cognitive tasks, either using self-report or experimental instruments.

This review has some limitations. Firstly, one of the limitations is the retrospective nature of the data. Secondly, results in studies using Wechsler's scales showed high heterogeneity, which makes it difficult to draw strong conclusions from these scales. We examined the influence of the use of different versions of the test as well as different age groups and other clinical characteristics of the samples. Despite efforts to clarify the reasons behind such different results, we were not able to identify a consistent factor across studies. We have also confirmed a lack of literature related to other ED diagnostic categories, such as BN, EDNOS or recovered ED groups in the context of IQ. This fact precludes the possibility to generalise these results to other ED groups and more importantly, the examination of potential relationships between IQ, recovery and prognosis. Thirdly, it is important to mention that most of reviewed studies involved clinical participants (inpatients or outpatients) or registered volunteers. One of the possibilities is that these samples are highly selected because they are not population-based samples, rather people who seek treatment in the clinics or are willing to participate in research. This may relate to higher education and IQ performance but these questions are beyond the scope of this study. Finally, as the main outcome of most of studies included in this review was not IQ performance, it is likely that a publication bias exists, however it is not possible to address it in this study.

From this review some recommendations arise: the majority of accessed studies used the NART or Wechsler's scales. It will be useful if future studies continue to use these measures to make future data comparable. It will also be desirable for all studies looking at neuropsychological factor to include an IQ measure due to the effect that it may have on neuropsychological task performance. Finally, since neuropsychological research is increasing, it will be helpful to report covariate analysis in relation to IQ; for example, with treatment outcomes, symptom severity and recovery.

This research highlights an important and underexamined factor in AN. The main conclusion is that IQ in AN is at least as high as the average IQ found in the normative population and most studies show that this group have a high average IQ. There is a preliminary but important observation about IQ in the recovered population, which is that this group may represent a group with higher IQ than norms and current AN groups, opening the question about the influence of this factor on treatment and recovery. We think that exploring IQ in the context of treatment and recovery may provide useful information for clinicians and researchers.