The impact of the SCL‐90 on the validity of the DSM‐IV neurotic or stress‐related disorders

The two studies on the validity of the self-reported symptom checklist (SCL) within the diagnostic spectrum of neurotic, nervous or stress-related disorders released in this issue of ACP (1, 2) have implications for the concept of comorbidity in the DSM-IV corresponding disorders as well as for the monitoring of outcome to treatment of these disorders. The SCL was originally developed to measure the dimension of neurotic distress in people seeking treatment in the outpatient clinic at the Johns Hopkins Hospital in Baltimore 50 years ago (3). This was, thus, before the area of antidepressants. The total score of the 41 symptoms covering both the somatic and the psychic manifestations of neurotic distress was used as the clinical measure of outcome to psychotherapy. In contrast to most other self-rating scales in this field (4), the SCL-41 has been enlarged over the years, first to an SCL-58 version (5) and then step by step, to the most recent version, the SCL-92 (2). In the first decade of the psychopharmacological area, i.e. from 1957 to 67, the SCL became one of the most frequently used self-rating scales for measuring outcome of therapy (6). This position is still valid, both in psychopharmacological and psychological outcome research. However, Pedersen and Karterud (1) have in their primary analysis reduced the number of SCL symptoms to around 60 to cover specifically the six ‘neurotic’ DSM-IV disorders: somatoform disorder; anxiety or panic disorder; social phobia or anxiety disorder; other phobic disorders; and depressive disorder or dysthymia. The six corresponding SCL subscales are conventionally referred to as somatization; anxiety; interpersonal sensitivity; phobia; obsessive–compulsive; and depression. Thereby the authors have used a version of the SCL approximately similar to the SCL-58 version. On the other hand, the most impressive finding that emerged from their study was the validation of the SCL-90R algorithm for the definition of psychiatric caseness. This algorithm says that if a respondent has a mean total score of all 90 symptoms equal to or greater than the national norm score of 63 on a T-scale, then he or she might be a psychiatric case (7). The validity of using the total symptom score of the SCL-90R as an expression of the overall neurotic distress is in agreement with the results obtained when using the Present State Examination (PSE) as an index of external validity (8). Thus, the correlation coefficient between the total symptom score of SCL-90R and the total score of the neurotic symptoms in PSE ranged between 0.72 and 0.82 for the various patient populations investigated (8). As long as we have no official list of ‘neurotic’ symptoms in DSM-IV, it is the concept of comorbidity that corresponds to the total symptom score of neurotic distress. This perspective has only partially been covered by Pedersen and Karterud (1). Recently, Bertelsen (9) has suggested rules for the diagnoses in the ICD-10 chapter on neurotic, stress-related and somatoform disorders when they coexist simultaneously and when not. The psychometric study by Olsen and coworkers (2) on the internal validity of SCL gives further evidence for using the mean total score of the symptoms in the SCL as a meaningful measure of neurotic illness severity. The fundamental criterion for unidimensionality of a rating scale in the item response theory models used by Olsen et al. (2) is that the items of the scale consistently can be rank-ordered according to their relation to the severity of illness. This implies that scorings on lower prevalence symptoms presuppose scorings on higher prevalence symptoms (4). If a rating scale is not unidimensional, a profile score of the different dimensions should be used, not the total score of all symptoms in the scale. Only symptoms measuring the same dimension are additive. Olsen et al. (2) have demonstrated that the six SCL subscales selected by Pedersen and Karterud (1) including 63 SCL symptoms belong to the same dimension of neurotic distress. This SCL-63 version should be considered the most valid measure for neurotic caseness when used conventionally by use of the national norms. Whereas national norms are inherent in the operational definition of caseness in the SCL, the DSM-IV diagnoses are based on the presence of a rather small number of symptoms, e.g. five of nine. In other words, the SCL caseness is based dimensionally with reference to a deviation from the national norms, i.e. culture-bound, while the DSM-IV disorders in this field are arbitrarily defined by algorithms which resist theories of measurement (10). Hence, the arbitrary DSM-IV disorders classify persons as either healthy or ill while the SCL approach is in analogy with dimensional disorders such as hypertension. The validity of the SCL and the corresponding DSM-IV disorders in terms of response to therapy or stability over time has not been analysed by Pedersen and Karterud, neither in their present report (1) nor in their results released elsewhere (11). Tyrer (12, 13) has questioned the utility of a subdivision of the general neurotic syndrome as the type of treatment is similar across these disorders (pharmacological as well as psychological) while their stability over time is rather low, resulting in the sum of symptoms to be the most usable statistic. When comparing the cost-effectiveness of brief interpersonal psychotherapy and the selective serotonin reuptake inhibitor paroxetine in patients with severe irritable bowel syndrome, Creed et al. (14) used the total symptom score of SCL-90R as the most appropriate distress score. They demonstrated that both types of therapy were significantly superior to treatment as usual, indicating that a biological impairment of serotonin as an expression of somatization can be improved not only with corresponding biological therapies, but with psychotherapy as well. In other words, even the total SCL-90R score is a valid measure in neurotic or stress-related disorders, sensitive to discriminating between active and inactive therapy. The results of the two SCL studies in this issue (1, 2) as well as the other two SCL studies referred to (8, 14) have supported the conclusion by the SCL architects themselves based on their own experiences over the first two decades with their scale, that its merits ‘…are its simplicity of administration, its easy comprehension and lack of offensiveness to patients, its sensitivity to change, and its compatibility with a variety of theoretical orientations…’ (5). Obviously, the number of comorbid neurotic or stress-related states before treatment [approximately three in the study by Pedersen and Karterud (1)] and the reduction of this number during therapy is an insensitive measure compared with the total SCL scorings, which is comparable with the measurement of blood pressure in the diagnosis and treatment outcome in hypertension. These SCL results should be taken into account by the diagnostic lumpers when they will meet with the diagnostic splitters for the next revision of the neurotic or stress-related DSM-IV or ICD-10 disorders in the not too distant future.