Our meta-analysis of 87 studies summarized the proportion of comorbid personality disorders (PDs) in patients with anorexia (AN) and bulimia nervosa (BN). The mean proportion of PDs among patients with any type of eating disorder (ED) was .52 compared to .09 in healthy controls. Patients with AN and BN had a quite comparable comorbidity profile, and with the borderline and avoidant PDs as the most prevalent [1]. Studies from 1980 to January 2016, published in English or German, were analyzed if the ED (and not the PD) was the primary diagnosis, and if the studies provided comorbidity proportions or data convertible to a proportion. Of particular interest here is that studies of patients below 18 years of age were excluded from the meta-analysis.

We are aware of at least two studies [2, 3] comprising adolescents below this age limit. Using structured [2] or self-report plus structured clinical [3] interviews, these studies reported very similar findings as in our meta-analysis. Similar results in only two additional studies would, however, have had a negligible impact on the “big picture” provided by this meta-analysis.

The general argument in these studies [2, 3] as well as in the commentary article by Gaudio and Dakanalis [4] is that an early identification of a PD may bring about treatment arrangements that may prevent a poor prognosis of the ED, and that such arrangements are possible because a PD among younger ED patients can be reliably assessed. The argument for the reliability of a PD assessment [4] seems to rest on an assumption that the personality of young people below the age of 18 is fully developed.

One should not be negligent about personality problems causing clinical impairment. Hence, the DSM-5 accepts a PD diagnosis (except an antisocial PD) as tentatively valid in adolescents below 18 years of age if personality traits appear as maladaptive, pervasive, and persistent within at least one year, and not limited to a particular developmental stage or another mental disorder. The exclusion criterion in our meta-analysis [1] adheres, however, to the general principle in the literature and in the DSM-5 that a PD diagnosis in adolescents below 18 years is probably invalid because their personality is far from fully developed and because many traits in younger people will not persist unchanged into adulthood. Obviously, individual differences in the development of personality and its disorders make the 18-year limit somewhat arbitrary, and the lowest true age limit for a valid PD diagnosis seems almost impossible to define exactly. Nevertheless, when used consistently, the 18-year limit may serve to prevent false-positive PD diagnoses in clinical work and it leads to few excluded studies in meta-analyses like in our study [1].

What does it mean that the personality in young people is not fully developed? This is a complicated question and attempts to study the personality of children and adolescents has been aptly described as trying to catch a moving target [5] in the crossfire between evidence of trait continuity versus change across the life span.

Personality development can be viewed as the psychological processing of interpersonal experiences in a complex interaction between biological temperaments and the nature and quality of adult–child attachments, and where the processing and accommodation of experiences are changing as a result of neuronal and psychological development. In children and adolescents, the brain continues to develop in terms of myelinisation and formation of synaptic networks in the frontal and striatal regions [6, 7], and there is a reciprocal relationship between a neuronal and psychological development [8, 9]. The road to a fully developed personality also contains a stage-related capacity for formal operations [10] paving the way for meta-cognitive processes and the development of identity [11]. Hence, catching a moving target like personality by using reliable cross-sectional diagnostic procedures [2, 3, 12] may miss clinically relevant interactive and developmental perspectives. Thus, for children and adolescents in a normal developmental process, troublesome behaviors, temporal emotional instability and negative affect may be better or more parsimoniously accounted for by the interaction between neuropsychological development, as well as parental and interpersonal factors than through for instance, a borderline PD diagnosis [13, 14]. Moreover, PD symptoms among adolescents have been found to decrease substantially as a result of improvement in axis I disorders and the increment in social support [15]. AN is the most frequent ED in the early teens. AN can be understood as a more or less deliberate avoidance of biological and psychological maturation, and the AN symptoms generally bring about an arrest in the bio-psycho-social normal development [16]. As a result, personality traits among ED patients may appear as “subclinical PDs”, notably along an affect-dysregulation dimension [17, 18]. Thus, the “psychology of avoidance” in teenagers with AN may be reliably classified as an avoidant PD or at least as a subclinical variant, but we argue that complex developmental conditions for doing so increase the risk of diagnostic invalidity.

Indeed, there is convincing evidence that some personality traits in children continue into adulthood [19, 20] within the structures of the Big Five [21] or the DSM axis II [22, 23]. There is also convincing evidence that personality traits and diagnoses can change considerably, particularly during childhood and early adolescence [24, 25]. From a developmental perspective, both sets of evidence may not be regarded as conflicting but rather as providing a genuine picture of childhood and adolescence as a life phase characterized by a pendulum of continuity and discontinuity across layers and facets of personality and identity development [5, 26, 27].

A PD diagnosis in young people with or without an ED may then appear reductionistic, and it may turn clinicians’ focus away from important interactional and developmental issues that are relevant for clinical investigation and treatment. Moreover, rather than describing adolescents’ personality traits as ‘currently raising concern’ a PD diagnosis may be highly stigmatizing [28], and through stigma and self-stigmatization a possible confirmation bias may falsely support the PD diagnosis and its validity. In sum then, a PD label, indicating maladaptive, pervasive and persistent intrapsychic structures may, despite a good reliability stand the risk of invalidating research as well as being counterproductive to the effort of providing good clinical practice.