Intellectual Decline in Schizophrenic Patients

TO THE EDITOR: Russell and his colleagues reported on the WAIS-R test results of adults with schizophrenia who had been previously tested as children (1). They interpreted their results as suggesting “stable impairment” and provocatively entitled their paper “Schizophrenia and the Myth of Intellectual Decline.” Because this “myth” was based on numerous empirical studies, it is important to examine the methodological details of the study by Russell et al. to reconcile the apparent contradictions. There are five major reasons to doubt their conclusions and to infer that the appearance of stability is, instead, likely evidence of intellectual decline.

1. The WAIS-R short form used on the second testing occasion probably overestimates the full-scale IQ of schizophrenic patients because it does not include several subtests on which poor performance is common among this group. To address this issue empirically, I examined the WAIS-R performance of 103 schizophrenic patients at the National Institute of Mental Health (NIMH). The group's mean full-scale IQ was 88.3 (SD=11.9) with a score of 97.5 (SD=88.3) on the Wide-Range Achievement Test-Revised; those scores suggested a 9.1-point decline (matched pair t test: t=8.15, df=102, p<0.0001). The patients' performance on Russell et al.'s five-subtest short form (mean=8.77, SD=2.2) differed significantly (t=6.53, df=102, p<0.001) from their performance on the remaining six subtests (mean=8.02, SD=1.9) as well as on all 11 subtests (mean=8.40, SD=2.0). This significant difference occurred despite a high correlation (r=0.94) between short-form and full-scale IQs. Thus, subjects retained their relative positions across short-form and full-scale IQs, but the short-form estimates were systematically higher than the actual full-scale IQs. Because all WISC-R subtests were administered at time 1, the comparison of actual WISC-R full-scale IQ versus short-form WAIS-R IQ is biased against detecting differences.

2. Age cohort effects confound the use of WISC-R and WAIS-R scores to determine the longitudinal course of intellectual functioning. As reviewed by Kaufman (2), each restandardization of the major IQ tests has documented substantial “gains” in IQ, estimated at 3 points per decade in the United States, where the IQ tests were normed. This cohort effect influences examinations of individual performance over time: the WISC-R was published in 1974; the WAIS-R, in 1981. At time 1, the group was 13.3 years old; their IQs were calculated relative to those of persons born in 1961. At time 2, the group was 32.9 years old; their IQs were calculated relative to those of individuals born in 1948. This 13-year age difference is likely responsible for a 3–4 point WAIS-R advantage relative to the WISC-R. In addition, the WAIS-R norms for 16- to 19-year-olds have been criticized as producing spuriously high scores (2). At time 2, at least one subject fell in this age range. Again, the psychometric problem decreased the probability of documenting intellectual loss.

3. Several studies reviewed by Kaufman (2) have suggested that WAIS-R scores of intellectually limited subjects are systematically higher than their WISC-R scores, with differences as high as 11 points reported. Thus, equivalent WISC-R IQ and WAIS-R IQ may be evidence of actual IQ decrement among these subjects. Russell et al. included six or seven subjects with WISC-R IQs lower than 75; the inclusion of such intellectually limited subjects would be particularly problematic and would raise possible diagnostic issues.

4. As noted by Russell et al., their group was highly unrepresentative of schizophrenic patients. The subjects had childhood-onset psychiatric symptoms, low overall IQs, and a substantial number of childhood-onset psychoses. What, if any, is the possible justification for generalizing the findings of Russell et al. to schizophrenic patients as a whole, as implied by the title?

5. The inclusion of nine subjects who were psychotic at time 1 undermined any examination of decline related to onset of psychotic illness, confusing this issue with deterioration over illness course. The fact that these subjects did not differ significantly from the rest of the group at time 1 or time 2 (a comparison with remarkably limited power) is not relevant to the main argument. The question of intellectual decline over illness course is distinct from the question of loss of intellectual ability with illness onset. Furthermore, the deletion of these nine subjects would have reduced the group to 25, a small number to use to confidently dispel a myth.

In summary, there are several reasons to suspect that the present comparison of estimated WAIS-R scores and WISC-R scores in schizophrenic patients resulted in the appearance of stability when actual loss, obscured by psychometric confounds, occurred. Considering that a decline of 2.3 points was documented without the contribution of these likely artifacts, a true decline of 8–10 points, the extent of decline estimated by using a variety of methods, may actually have been present.