Psychopharmacology

, Volume 187, Issue 3, pp 356–363

The dimensions of clinical and cognitive change in schizophrenia: evidence for independence of improvements

Authors

    • Department of PsychiatryMt. Sinai School of Medicine
  • Michael F. Green
    • Neuropsychiatric Institute, UCLAVA Greater Los Angeles Healthcare System
  • Christopher Bowie
    • Department of PsychiatryMt. Sinai School of Medicine
  • Antony Loebel
    • Pfizer Inc.
Original Investigation

DOI: 10.1007/s00213-006-0432-1

Cite this article as:
Harvey, P.D., Green, M.F., Bowie, C. et al. Psychopharmacology (2006) 187: 356. doi:10.1007/s00213-006-0432-1

Abstract

Background

As cognitive impairments are related to deficits in everyday functioning in schizophrenia, treatment of these impairments may have the potential to reduce these functional deficits. To determine if treatments truly reduce cognitive impairment, it is important to discriminate direct cognitive effects of treatment from generalized treatment benefits on the multiple clinical dimensions of schizophrenia. Thus, this study used a database from an existing clinical trial and examined the relationships between changes in clinical symptoms and cognitive deficits with several different strategies.

Materials and methods

Two hundred and seventy stable but symptomatic outpatients with schizophrenia entered a study where they were switched from previous treatment to open-label ziprasidone. The present data are from the 6-month endpoint (n=184). Patients were examined at baseline and the 6-month endpoint with ratings of clinical symptoms based on the Positive and Negative Syndrome Scale (PANSS) and a neuropsychological (NP) assessment battery including aspects of cognitive functioning known to be related to functional outcome in schizophrenia.

Results

Changes on the individual PANSS items and NP test scores were examined with two separate principal components analyses, revealing four dimensions of clinical change (psychosis, negative symptoms, affective symptoms, and agitation) and two dimensions of NP improvement. Pearson correlations between changes in the (1) factors derived from the analyses, (2) individual NP items based the four clinical dimensions of change, and (3) the 30 PANSS items and the two NP dimensions of change suggested minimal relationships (largest r=0.15).

Implications

This sample was selected because previous findings suggested that clinical and NP symptoms of schizophrenia significantly improved from baseline after a switch to ziprasidone treatment. While four dimensions of change in clinical symptoms and two dimensions of cognitive improvements were identified, clinical changes, regardless of how they were defined, were not related to NP improvements.

Keywords

SchizophreniaClinical effectivenessCognitive functionAntipsychoticTreatment

Cognitive impairment in schizophrenia is associated, both cross-sectionally (Green 1996) and longitudinally (Green et al. 2004), with impairments in everyday functioning. Because functional deficits in domains of employment (McGurk and Mueser 2004), independent living (Hegarty et al. 1994), and social functions (Harvey et al. 2004a) are a major cause of disability in schizophrenia, reduction of cognitive impairments was identified as a priority for new treatment development in schizophrenia (Hyman and Fenton 2003). Multiple studies have indicated that both pharmacological interventions, including atypical antipsychotic medications (Harvey et al. 2003, 2004b; Keefe et al. 2004; Velligan et al. 2002) and other medications (Friedman et al. 2001; Turner et al. 2004), as well as behavioral interventions such as cognitive remediation (Hogarty et al. 2004; McGurk et al. 2005; Wexler and Bell 2005) can reduce cognitive impairments in schizophrenia to a variable extent (Kurt et al. 2001). Therefore, it is important to determine if treatments aimed at reducing cognitive impairments have a direct impact on cognition or if they are active through alternative means such as reduction of other symptoms of the illness, reductions in anxiety, or by affecting general improvements in motivation or cooperation. Since schizophrenia is a heterogeneous and multidimensional illness, this is a complicated question.

For a medication to be approved to treat cognitive impairments, regulators must believe that a treatment effect is direct and not explained by other aspects of treatment response. Given the consistent findings regarding the lack of any substantial cross-sectional (Keefe et al. 2006) or longitudinal (Velligan et al. 1997) correlation between the severity of cognitive impairment and the severity of psychotic symptoms, it seems unlikely that reductions in psychosis, regardless of mechanism, would improve cognition. However, considerable data suggests that cognitive impairments display a moderate cross-sectional (Keefe et al. 2006; Addington et al. 1991), if not longitudinal (Harvey et al. 1996a), correlation with the severity of negative symptoms. Thus, determining whether cognitive change is a by-product of change in other symptoms is an important research goal.

One issue that is relevant to the question of the directness of cognitive change that has not been addressed in detail is that of the potential structure of change in symptoms and cognitive deficits. While literally hundreds of studies have examined the symptomatic structure of schizophrenia with factor analytic techniques (White et al. 1997) and many studies have examined treatment-related changes in these a priori clinical symptom factors (Conley and Mahmoud 2001; Beaseley et al. 1997; Simpson et al. 2004), there is essentially no information regarding the factorial structure of symptom change. The fact that symptom dimensions can be defined with factor analysis does not rule out the possibility that treatment-related changes in the symptom severity might influence the structure of symptoms. The only study that examined this issue (Harvey et al. 1996b) found that the structure of symptoms, as defined by confirmatory factor analysis, remained constant despite improvements in symptom severity after treatment with conventional antipsychotic medication. That study did not examine the factors or dimensions of clinical change; rather, it measured the impact of changes in severity of symptoms on predefined symptom dimensions. That study also used haloperidol at high doses as the treatment of choice; the results may be different with treatment with atypical medications.

The present study had two main goals. The first goal was descriptive: to identify the dimensions of clinical and cognitive treatment response in patients with schizophrenia who were switched from previous treatment to a new antipsychotic using factor analysis. Using a database from a study (Weiden et al. 2003; Harvey et al. 2004c) where significant changes in symptomatology and cognitive functioning in schizophrenia patients was already reported, we used statistical techniques to empirically identify the factor structure of clinical and cognitive changes. The second goal was to evaluate the extent to which clinical changes, in both the derived factor structure of symptom and the individual Positive and Negative Syndrome Scale (PANSS) items, were correlated with treatment-related changes in performance on neuropsychological (NP) tests. These correlational analyses also related the dimensions derived by factor analysis of cognitive change to the dimensions of clinical change derived likewise and to changes on the individual PANSS items.

We hypothesized, based on our previous study with conventional medications, that the dimensions of clinical change would resemble the classical dimensions of symptomatology in schizophrenia derived with factor analysis in nontreatment studies. We also hypothesized cognitive change in schizophrenia would be multidimensional, as well, although we had no basis to predict how many dimensions of change would be detected. Finally, we hypothesized that cognitive and clinical changes would be minimally related to each other, consistent with previous studies reporting that changes in a priori dimensions of symptoms were essentially uncorrelated with cognitive improvements (Harvey et al. 2003, 2004b; Keefe et al. 2004). To evaluate this final issue, we also investigated the correlation between changes in the classical three dimensions of the PANSS (positive, negative, and general psychopathology) and cognitive changes.

Materials and methods

Subjects

These data come from a pooled extension of three identical 6-week open-label switch studies of schizophrenia outpatients who were switched to ziprasidone from conventional medications, olanzapine or risperidone. Data on the 6-week efficacy portion of these studies was previous published (Weiden et al. 2003; Harvey et al. 2004c). Eligible patients were diagnosed with schizophrenia or schizoaffective disorder, as defined in the Diagnostic and Statistical Manual for Mental Disorders—Fourth Edition (DSM-IV), and were between the ages of 18 and 55 with at least an 8th grade reading level. In addition, patients were taking oral doses of antipsychotic medication within 25% of the recommended daily dose for a minimum of 3 months with no history of treatment resistance. Major exclusion criteria were the presence of other current DSM-IV Axis-I disorders not in remission; a history of treatment resistance; urine toxicology positive for illicit drugs or a history of psychoactive substance abuse or dependence not in current remission; more than moderate depression; or treatment with psychoactive medications other than antipsychotic medication. Patients provided written informed consent and all procedures were approved by institutional review boards at each site.

Patients were recruited from multiple treatment sites within the US. Patients who were receiving treatment with conventional antipsychotic medications, risperidone, or olanzapine were eligible. At the completion of the 6-week efficacy trial, patients who had completed the study were offered the opportunity to participate in a 1-year extension to assess the efficacy and tolerability of ziprasidone. Of the 270 patients who were switched at baseline, 185 entered the extension study, with 184 of these providing cognitive data at 6 months or earlier termination. Cognitive assessments were performed at baseline, while patients were receiving their previous medication at the close of the 6-week efficacy study, and again at 6 months during the continuation phase. There were no statistically significant differences between the patients who agreed to continue in the extension study and those who did not on demographic or baseline cognitive or symptomatic variables.

Clinical assessments The main clinical assessments were carried out with the Positive and Negative Syndrome Scale (Kay 1991), which was administered weekly from baseline through 6 weeks and then again at month 6. For the current report, all 30 items in the PANSS were used. Difference scores from the baseline assessment to the 6-month assessment corresponding to the final cognitive assessment were calculated and used to derive dimensions of clinical change.

Cognitive assessments

These assessments covered domains of verbal learning and memory, attention and vigilance, executive functions, and language skills. As we were planning to use factor analysis to identify the dimensions of change and did not want to weigh tests differentially, we selected a single dependent measure from each test.

Verbal learning and memory was assessed with the Rey Auditory Verbal Learning Test (Spreen and Strauss 1998). The critical dependent measure was total learning for trials 1 to 5. Attention and vigilance were assessed with three tasks: the Digit Span Distraction Test (Oltmanns and Neale 1975), the Continuous Performance Test, Identical Pairs version (Cornblatt et al. 1989), and the Trail Making test, part A (Reitan and Wolfson 1993). The dependent variables selected from these three tests were proportioned correctly during distraction performance: the total d′ score over 450 trials and total time to completion. Verbal fluency was assessed with two measures. Phonological fluency was measured with the Controlled Oral Word Association Test (Spreen and Strauss 1998), and semantic fluency was measured with the animal naming test (Spreen and Strauss 1998). Executive functioning was measured with part B of the Trail Making Test and the Wisconsin Card Sorting Test (WCST; Heaton et al. 1993), with the dependent variables being total time to completion and total WCST errors, respectively.

Data analyses

Dimensions of clinical and cognitive change were identified with two separate, unrotated principal components analyses. An a priori decision was made to require eigenvalues of 1.0 or greater in order to interpret a factor as meaningful. An exploratory joint analysis was also performed. Changes from baseline in the clinical and cognitive variables were examined with 95% confidence intervals, and relationships between the clinical and cognitive change scores and the extracted principal components were examined with Pearson correlations. We also correlated the change scores for the classical PANSS dimensions (positive, negative, general psychopathology) as well as changes in the PANSS total scores with changes in cognitive performance.

Results

At baseline, patients had a mean age of 38.2 (SD=10.1; range 18–61). Sixty-two (35.5%) patients were female. Education was coded by level rather than number of years of schooling. This yielded 8 (4%) patients with less than 8 years of education, 28 (15%) with at least some high school, and 148 (80%) with at least a high school degree.

Extent of clinical and cognitive change

Of the 30 PANSS item change scores, the means for 26 exceeded (p<0.05) the 95% confidence interval for improvements from baseline to endpoint (somatic concern, mannerisms and posturing, disorientation, and inattentiveness did not change) and the classical PANSS total, positive, negative, and general scores all improved significantly in this time frame as well [all t (183)>6.27; all p<0.001]. Change scores for the cognitive variables in effective size units are presented in Fig. 1. Statistically significant improvements from baseline were found for seven out of eight cognitive variables, with only Trail making test part B not showing a statistically significant change from baseline to the 6-month endpoint.
https://static-content.springer.com/image/art%3A10.1007%2Fs00213-006-0432-1/MediaObjects/213_2006_432_Fig1_HTML.gif
Fig. 1

Cognitive changes over the 6-month extension study

Dimensions of clinical and cognitive change

To identify the dimensions of clinical change, separate principal components analyses were applied to the 30 PANSS item change scores and the eight cognitive change scores. The PANSS analysis resulted in four factors that had eigenvalues of greater than 1.0 and the cognitive analysis resulted in two factors. The factor loadings for the two analyses are presented in Tables 1 and 2. As can be seen in Table 1, these clinical dimensions of change have loadings that appear to describe changes in psychotic symptoms, negative symptoms, agitation, and affective variables, accounting for a total of 42% of the variance across the four factors. The factor loadings of these clinical change scores were quite consistent with the results of previous naturalistic confirmatory factor analyses of the PANSS, with one exception. There is no factor that resembles the “cognitive” or “disorganization” factor that is often found in these studies.
Table 1

Factor structure of clinical change in PANSS scores over 6 months

PANSS item

Loadings

Factor 1 “Psychosis”

Factor 2 “Negative”

Factor 3 “Agitation”

Factor 4 “Affective”

Delusions

0.78

0.07

0.12

0.20

Hallucinations

0.60

0.14

0.40

0.00

Conceptual disorganization

0.41

0.22

0.10

0.22

Excitement

0.12

0.02

0.77

0.13

Grandiosity

0.33

−0.21

0.38

0.15

Suspiciousness

0.72

0.16

0.23

0.14

Hostility

0.07

0.11

0.50

0.41

Blunted affect

0.10

0.73

−0.03

0.06

Emotional withdrawal

0.32

0.62

0.07

0.09

Poor rapport

−0.15

0.68

0.07

0.13

Passive social withdrawal

0.33

0.62

−0.03

0.20

Abstract thinking

0.12

0.36

0.36

−0.10

Lack of spontaneity

−0.05

0.73

0.02

−0.07

Stereotyped thinking

0.18

0.17

0.08

0.03

Somatic concern

0.04

−0.08

0.10

0.39

Anxiety

0.21

0.20

0.01

0.72

Guilt

0.38

−0.24

0.27

0.43

Tension

0.15

0.12

0.25

0.70

Mannerisms

0.06

0.14

0.04

0.25

Depression

0.30

−0.14

0.03

0.52

Motor retardation

−0.14

0.60

0.18

0.12

Uncooperativeness

−0.18

0.20

0.35

0.24

Unusual thought content

0.77

−0.14

0.02

0.05

Disorientation

0.04

−0.20

0.03

0.24

Attention

0.07

0.27

0.49

0.17

Judgment and insight

0.23

0.20

0.32

−0.04

Volition

0.14

0.50

0.48

−0.02

Impulse control

0.18

−0.01

0.58

0.04

Preoccupation

0.29

0.10

0.33

0.16

Active social avoidance

0.61

0.21

0.03

0.00

Eigenvalue

6.43

2.96

1.92

1.38

Variance accounted for (%)

21

10

6

5

Loadings over 0.40 are highlighted in bold

Table 2

Factor structure of cognitive changes scores over 6 months

Cognitive Variable

Loadings

Factor 1 “Verbal/Learning”

Factor 2“Attention”

RAVLT learning

0.61

0.12

Trail making part A

0.32

0.63

CPT d′

0.22

0.57

Digit span distraction

0.09

0.31

Trail making part B

0.23

0.11

WCST errors

0.06

0.30

Animal fluency

0.55

0.36

FAS fluency

0.75

0.13

Eigenvalue

2.60

1.36

Variance accounted for

0.20

0.17

The cognitive change factors shown in Table 2 have several features. As can be seen from inspection of the factor loadings, the two factors describe changes in verbal/memory and attentional functions. These two factors account for 37% of the variance in the cognitive changes. It is interesting to note that digit-span distraction performance and WCST errors, while both improving to a statistically significant extent, did not manifest differential loadings on the two factors.

Correlations of clinical and cognitive changes

See Table 3 for correlation between changes in the cognitive items and changes in the PANSS factors and a priori dimensions of the PANSS. When the two cognitive and four clinical change factors were correlated with each other, there was no statistically significant correlations between them (all r<0.09, p>0.50). Furthermore, 1 of 30 PANSS item change scores correlated with cognitive change factor 1 (passive-apathetic social withdrawal, r=0.23, p<0.05). None of the eight individual cognitive item change scores was significantly correlated with any of the four clinical change factors and, similarly, all correlations between cognitive change scores and changes in the classical PANSS dimensions, as well as changes in the PANSS total score, were nonsignificant.
Table 3

Correlations between dimensions of clinical change and changes in cognitive performance

 

Cognitive variables

RAVLT learning

Trail-making part A

CPT d′

Digit span distraction

Trail-making part B

WCST errors

Animal fluency

FAS fluency

Factor analysis derived dimensions

        

Positive

−0.05

−0.12

−0.14

0.13

0.10

0.01

0.13

0.00

Negative

0.01

−0.01

0.07

−0.02

0.02

0.01

0.02

−0.14

Agitation

0.15

−0.12

0.05

0.08

0.08

−0.12

0.01

0.11

Affective

−0.10

−0.08

−0.11

−0.09

0.01

0.01

0.03

−0.01

Classical PANNS scores

        

Total PANSS

0.05

0.12

0.04

−0.14

0.10

0.00

−0.13

0.09

Positive

0.07

0.09

0.08

−0.10

0.07

0.01

−0.07

0.01

Negative

0.04

0.07

0.12

−0.13

0.09

0.02

−0.14

−0.14

General

0.03

0.12

0.05

−0.09

0.08

−0.01

−0.10

0.07

Trail making test part B did not improve significantly from baseline. All other cognitive variables improved

Joint factor analysis

In a final exploratory analysis aimed at looking at the convergence of clinical and cognitive changes, all 30 PANSS items and all 8 cognitive scores were entered into a final unrotated principal components analysis. A six-factor solution emerged, all with eigenvalues greater than 1.0, accounting for a total of 44% of the variance. These factors were very similar to the four clinical and two cognitive factors that were extracted from the previous independent analyses. Most striking about these results was that no single cognitive variable had a loading of 0.40 or greater on any of the first four factors, and no single PANSS item had a loading of 0.40 or greater on either of the final two factors. While this is not a confirmatory analysis, the results are quite consistent with the correlational analyses presented immediately above, suggesting separate structures for cognitive and clinical changes with treatment.

Medication and extrapyramidal symptoms effects

To determine if the medication from which patients were switched influenced the results, a baseline treatment × time repeated measures analysis of variance was performed on each of the 30 PANSS variables and the 8 cognitive outcome variables, with the original treatment medication (conventional, N=71; risperidone, N=42; olanzapine, N=71) entered as a between-subjects factor. No significant between-group differences or interactions of group × time were found for any variables (all F<2.6; all p>0.05), reflecting no differences in cognitive or clinical treatment response as a function of the original treatment. There was no significant correlation between total changes in extrapyramidal symptoms, as measured by the Extrapyramidal Symptoms Scale (Chouinard and Ross-Chouinard 1980), and changes in any of the eight cognitive variables (all r<0.11, all p>0.50).

Discussion

In this study, treatment-related clinical and cognitive changes after a change to a different antipsychotic were of a multidimensional character. Four different dimensions of clinical change were detected, with these dimensions generally consistent with the dimensions of psychopathology detected in multiple previous studies of the factor structure of schizophrenia symptoms. Two dimensions of cognitive change were also found. Despite statistically significant improvements in clinical and cognitive domains, which were interrelated as indexed by the substantial eigenvalues and reasonable amounts of variance extracted with principal components analysis, there was essentially no correlation between clinical and cognitive changes. Regardless of whether the clinical and cognitive changes were evaluated in terms of items or factors, there was no correlation between these change scores. As a result, this study provides evidence arguing against the “pseudospecificity” (Laughren 2001) of cognitive changes.

It is of interest that there was no “cognitive” dimension of clinical change detected. This may be due to the fact that several of the critical items for this factor were among the few PANSS items that did not change. Questions about the validity of the PANSS cognitive factor have been raised by two previous studies that demonstrated in previously treated (Harvey et al. 2001) and first-episode (Good et al. 2004) patients before their first treatment that this factor was not strongly correlated with NP performance. While previous studies of atypical antipsychotic treatments showed that there was no significant correlation between treatment-related changes in a priori defined symptom dimensions and NP improvements (Harvey et al. 2003, 2004b; Keefe et al. 2004), this study is unique in that statistically coherent dimensions of clinical and cognitive changes with treatment were defined empirically. In this study, the chances of identifying correlations between treatment-related changes in clinical and cognitive features of schizophrenia would appear to be likely to be maximized, because the database selected had previous evidence of statistically significant changes in both clinical and cognitive domains. Thus, our finding that NP and clinical improvements with treatment are uncorrelated could not be due to low levels of change or low potential for correlation in general.

Among the implications of these findings is that treatment-related improvements with novel antipsychotics are multidimensional. Further, these results support the proposal that it may be possible to target cognition as a unique treatment target in schizophrenia (Gold 2004). While improvements in depression, negative symptoms, and cognition all detected in this study might be expected given the pharmacological profile of ziprasidone (Harvey and Bowie 2005), the current study cannot determine which of the pharmacological mechanisms of ziprasidone treatment are responsible for these effects or even whether they are specific to ziprasidone treatment.

The possibility that these cognitive changes are at least partially due to retesting effects requires consideration. It is possible that the use of identical forms may lead to practice effects. However, practice effects are not linear and unlimited in nature. Hawkins and Wexler (1999) found that practice effects on the California Verbal Learning Test were twice as large from the baseline to the first reassessment as they were from the first to second reassessments, with those later changes not statistically significant. Furthermore, they found that not all tests had practice effects; no practice effects were detected on the Trail making test, which was also one of the tests on which performance improved in this study. Second, recent data have suggested that patients with schizophrenia may have more modest practice effects than those expected in the healthy population (Blyler and Gold 2000). In fact, we recently demonstrated that older patients with schizophrenia who are treated with conventional antipsychotic treatments did not improve to a statistically significant extent over an 8-week retesting period on any test in a 21-test cognitive evaluation (Harvey et al. 2005). Focusing on memory performance, the change from the baseline assessment to retest were not only not statistically significant, it was an effect size of d=0.08 (about one third of the change in this study). Finally, a systematic study of practice effects on attentional performance indicated that patients with schizophrenia treated with low doses of conventional medications failed to manifest any appreciable improvement in performance with over 8,000 practice trials (Harvey et al. 2000).

There are several limitations of the study, some of which are shared with the previously published data from the 6-week trial. The open-label nature of this trial is a limitation and may contribute to the extent of improvements seen (Woodward et al. 2005). Since this was a switch study and not a parallel design, the comparative efficacy of ziprasidone and the other two atypical antipsychotics was not evaluated. Also, there was no objective monitoring of treatment compliance in this trial. However, it is hard to understand how poor adherence could be argued to be responsible for consistent clinical and cognitive improvements. As with the majority of clinical studies of cognitive enhancement in schizophrenia, no data regarding real-world functional change were collected. Our factor analyses were exploratory, so replication of these findings is clearly required. Not all patients entered the extension, so the representativeness of the results may be affected as a result. Finally, the overall magnitude of both clinical and cognitive improvements, though statistically significant, was small to moderate. Larger improvements might have the potential to detect more substantial correlations between these changes.

In summary, treatment change associated with switch to a different atypical antipsychotic was associated with multidimensional changes in symptoms and cognitive functioning. While clinical changes had a factor structure similar to previous multivariate studies of clinical symptoms of schizophrenia, cognitive changes were unrelated to clinical change. Thus, concerns about whether cognitive change is due to global improvements in psychopathology are somewhat allayed by these data, in which both clinical and cognitive change was found to occur independently of each other.

Acknowledgement

This research was supported by Pfizer Pharmaceuticals.

Disclosures Philip D. Harvey has worked as a consultant for Pfizer Pharmaceuticals; Michael F. Green has also worked as a consultant for Pfizer Pharmaceuticals. Christopher R. Bowie has no current relationship with Pfizer Pharmaceuticals; Antony Loebel is a full-time employee of Pfizer, Inc.

Copyright information

© Springer-Verlag 2006