To our knowledge, the association between SM and anxiety (as a symptom) has never been investigated through a meta-analytic approach, despite its reclassification as an anxiety disorder in the DSM-5. The present study sought to examine the supposed link between SM and anxiety by means of a meta-analysis. Since levels of anxiety in children diagnosed with SM have not always been assessed systematically throughout the literature, data on comorbid anxiety disorders in SM was judged to be the best substitute for examining this relationship. Hence, the main purpose of the present study was to provide an overall estimation of the prevalence rate of comorbid anxiety disorders in children with SM on the basis of meta-analytic data.
A total of 22 studies met the eligibility criteria for inclusion in the meta-analysis, comprising data on 837 children diagnosed with SM. The results indicated that 80% of these children were additionally diagnosed with at least one comorbid anxiety disorder. Social phobia (i.e. social anxiety disorder) was found present in 69% of the children with a diagnosis of SM, making it the most commonly diagnosed comorbid anxiety disorder. This was followed by specific phobia (19%), separation anxiety disorder (18%), generalized anxiety disorder (6%), and obsessive-compulsive disorder (6%). By and large, these figures are in line with the observation made in the DSM-5 that, “In clinical settings, children with selective mutism are almost always given an additional diagnosis of another anxiety disorder—most commonly, social anxiety disorder” (p. 196) . However, they cannot fully support the notion that anxiety is always present in SM, as is currently implied by its classification as an anxiety disorder.
While diagnoses of comorbid anxiety disorder may account for the presence of anxiety in 80% of the children diagnosed with SM, it remains unclear how anxiety is manifested in the remaining 20% of the children that lacked such an additional diagnosis. The first and rather obvious reason for this is that anxiety does not feature in the DSM-5’s list of diagnostic criteria for SM, and that in clinical practice it is hardly ever assessed in a direct way. Moreover, the presence of a comorbid anxiety disorder does not automatically imply that SM originates from the same source (i.e., anxiety)—that is to say, not any more than the presence of a comorbid depressive disorder would imply that SM would be caused by depression. As noted by Tramer as early as 1934, numerous factors are to be taken into consideration if we want to unravel the (probably heterogeneous) etiology of SM . Preferably, one would like to investigate the presence or absence of these factors in populations with SM in a direct way. Regarding anxiety levels, this should ideally be done with the aid of validated questionnaires and psychophysiological measurements of heart rate, blood pressure, and skin conductance, or some other method to assess physiological arousal. Since in clinical practice such assessments are hardly feasible because of the very problem at hand (i.e., mutism and uncooperativeness), the assessment of comorbid anxiety disorders has been embraced as a litmus test for the presence of anxiety. However, our analysis indicates that such comorbid anxiety disorders are undiagnosable in 20% of all children with SM. Moreover, among the remaining 80% there is a group of no less than 19% with a fear for flying, heights, animals, receiving an injection, seeing blood, and so on (i.e., specific phobias). One may ask oneself whether the anxiety underlying these specific phobias is really sufficient to explain a lack of speech in front of specific individuals. In sum, by focusing exclusively on comorbid anxiety disorders, the DSM would appear to restrict its approach to ‘looking where the light shines’. This brings us back to Tramer’s observation that the etiology of SM is likely to be diverse. When Tancer  reviewed the literature to facilitate modifications of the diagnostic criteria of SM in the DSM-IV, she argued that the criteria had been defined very broadly. Like Tramer  before her, she expressed the suspicion that this was likely to result in the inclusion of children with diverse underlying pathology. Consequently, she emphasized that systematic research was needed before more specific criteria could be designed. Now, more than 25 years later, the number of studies investigating samples of children with SM has grown substantially. And yet Tancer’s concern remains topical as the DSM criteria of SM have not been changed, despite its new classification as an anxiety disorder.
Consequences for Classification
Our lack of insight into the full spectrum of etiological mechanisms underlying SM has several consequences for the classification of SM in the DSM and other taxonomies. On the basis of our meta-analysis and subsequent discussion, we conclude that we are currently not in a position to establish with sufficient accuracy whether anxiety plays a key role in all cases of SM, or even in a majority of them. From that vantage point, it is as yet uncertain how SM should be classified (i.e., as a member of the higher-level group of anxiety disorders, or perhaps as a member of the groups of communication disorders, oppositional defiant disorders or neurodevelopmental disorders, to mention some of the various possibilities). The so-called ‘atheoretical’ approach of the DSM has been much debated, but in conformity with this vantage point it might well have been more appropriate to stick to the classification of SM as a disorder of infancy, childhood, or adolescence (i.e., what is currently called a neurodevelopmental disorder in the DSM-5). After all, by relocating SM to the group of anxiety disorders, the implicit theoretical assumption appears to have been made that anxiety constitutes the disorder’s symptomatological and/or etiological cornerstone. Since the DSM-5 does acknowledge the issue of etiological heterogeneity by pointing out the influence of temperamental, environmental, genetic, and physiological factors, albeit under the heading of Risk and Prognostic Factors, it is all the more remarkable that SM ended up under the heading of Anxiety Disorders.
That said, the considerable overlap of SM with social phobia may also suggest that these two disorders, as defined in the DSM-5, are not discrete, separable categories. Psychiatric classification is different from biological classification, but an analogy may nonetheless be illuminating. For example, when defining the housecat as a species, biologists do not have the habit of listing all its characteristic features (i.e., having fur, walking on four legs, purring when petted, being good at ignoring people yet being friendly to others, etc.). Instead, they focus on the features that separate housecats from the other species within the overarching genus Felis. They do not list the features that all cats (including lions, tigers, etc.) have, or all mammals have, etc. In the DSM-5, “Anxiety Disorders” is a family-level category somewhat like “mammal”. In accordance with the principles of biological classification, if SM fits under the category of anxiety disorders, then persons with SM should also meet the more general definition of an anxiety disorder. As a consequence, the definitional task here would not be to simply describe SM, but, instead, to define whatever characteristics separate SM from the other anxiety disorders, especially social phobia . The present study indicates that the current DSM definition of SM does not meet this goal. Sticking with the example of biological classification, the question whether SM is an anxiety disorder, should then become whether most children with SM meet the definition of any of the higher-level groups of disorders.
SM in Other Age Groups
By focusing almost exclusively on children, the DSM-5 also precludes the possibility of a proper comparison of SM across age groups. Clinical practice teaches us that SM is also found in the context of schizophrenia spectrum disorders, catatonia , autism , and numerous other disorders in adulthood, and even outside the domain of psychopathology, where it may serve as a powerful tool to vent one’s misapprehension of—or hostility towards—another person; or, alternatively, be an expression of extreme shyness or insecurity. Although mutism has been studied in adult populations, notably in the context of autism and catatonia, there is a dire need to focus such studies on the presence of SM if we wish to say more about a possible continuum with the population targeted by the diagnostic category of SM in the DSM-5. As pointed out by Kussmaul as early as 1877, the essence of SM appears to be that it is a speech disorder. The diagnostic criteria of SM in the DSM-5 seem to be most in line with this viewpoint. As a consequence, it may perhaps be necessary to revise its current classification as an anxiety disorder—at least until empirical research provides us with data to back up an alternative choice, and the classificatory issues regarding the defining characteristics of higher-level groups of disorders and the distinguishing features of lower-level, individual disorders has been solved.
Our results should be viewed in light of several limitations. The initial search strategy identified a large number of records that had to be filtered out before abstracts were reviewed, which partly included non-peer reviewed material, such as conference abstracts and dissertations. Although we argue that the exclusion of these resources improved the scientific value of our meta-analysis, there is a possibility that important data on groups of children diagnosed with SM went undetected due to this procedure. The same holds for publications not written in English, which comprised nearly one-fifth of the records identified by our initial search strategy.
Another limitation is the observed heterogeneity of prevalence figures, as this impedes the drawing of conclusions from the meta-analysis. Heterogeneity between studies is likely to arise from sampling errors and/or differences in research methodology across studies . In case of the current study, these shortcomings may have been amplified by the broad scope of our search strategy. That is to say, our search query was defined rather broadly in order to capture all publications pertaining to the topic of SM, while the inclusion criteria for meta-analysis made no requirements regarding research design or research objective, since prevalence rates were used as a variable, rather than outcome variables. Consequently, different types of study were deemed eligible for inclusion in the meta-analysis, which likely resulted in varying diagnostic assessment strategies among the studies. We did not exclude papers on the basis of methodological quality due to the small size of the literature on the topic of SM. However, evaluation of methodological quality was rated only moderate on average and, indeed, revealed varying results across the studies. Although this may simply reflect the current state of research in the field of SM, sampling errors and differences in research methodology were likely present and may thus have contributed to the observed heterogeneity.
To deal with this, we explored possible moderating effects through meta-regression analysis. Extracted data enabled us to include age, proportion of female subjects, and year of publication as potential predictors in the equation. Additionally, we included methodological quality as a moderator on the basis of the results of the quality assessments. However, the results of our mixed-effects meta-regression analysis revealed that none of these moderators could explain a significant part of the heterogeneity that was observed in the meta-analysis.
Recommendations for Clinical Practice and Research
Several recommendations can be made following the results of our study. First of all, children with SM should be assessed in greater detail with regard to the anxiety-related symptoms they may experience, including concrete events that are being feared, and the content of associated cognitions. Such an extensive assessment is preferably to be supported by objective measures (e.g., psychophysiological measurements), since subjective assessments made by parents, teachers or clinicians are not sufficient in this respect, especially when children with SM remain silent, and may not even be in a position to confirm or deny whether any of the conclusions tally with their own experiences . Secondly, we need to broaden the scope of these assessments so as to include other factors that might play a role in the etiology of SM, including the temperamental, environmental, genetic, and physiological factors mentioned in the DSM-5. Thirdly, future research could enlarge the target population by also including adolescents, adults, and elderly people, in whom SM has remained largely unexplored. In the fourth place, we advocate the development of standardized questionnaires and psychophysiological measurements so as to promote the homogeneity between studies and increase overall reliability, particularly when clinical anxiety in children diagnosed with SM is being examined. Together, these steps should help us to elucidate the etiological factors underlying SM. Meanwhile, for as long as these remain uncertain, we advocate a revision of the current classification of SM to help us prevent making any premature associations with anxiety or certain age groups, and also to help us to keep an open eye for a possible differentiation of the condition into several subtypes.