Research on the interface of cognition, treatment, and outcome in schizophrenia is going through two simultaneous transitions. The first involves our view of cognition and functional outcome, and the second, our view of medications and cognition. Both transitions leave us feeling a little smarter and a lot more humble.
Regarding the connection between cognition and functional outcome, we know from scores of data-based studies that cognition in schizophrenia is related to community functioning and skill acquisition (1). The relationships vary in strength from study to study, but generally they are there. The challenge has shifted from finding an effect to understanding it. To that end, studies have started to examine intervening variables that may be key mediators between basic cognition on the one hand and community functioning on the other. Such proposed mediators include measures of functional capacity and social cognition (2, 3). To successfully map the steps leading to functional outcome, it is necessary to adopt statistical methods that are quite different from more familiar hypothesis-testing approaches. Instead, modeling methods such as path analysis and structural equation modeling can reveal underlying factors and mediating relationships by comparing the degree to which observed data fit predicted theoretical models (4). As we increasingly appreciate the steps between cognition and outcome, we increasingly realize that improvements in cognition would only be the start of a long series of steps toward better functioning. Addition of nonpharmacological interventions ranging from cognitive remediation to vocational rehabilitation would likely be necessary to see results at the community level.
A second key transition is that we are less comfortable with pinning our hopes on antipsychotic medications as a way to achieve cognitive improvement. Optimism that second-generation antipsychotics would yield cognitive improvements has progressively been tempered as treatment effect sizes have progressively dwindled, possibly as a result of dosing factors (as doses of comparators became lower) or patient selection factors (as more patients received second-generation medications). At any rate, the high hopes for beneficial cognitive effects from antipsychotic medications are now hanging by threads. The article in this issue by Keefe et al., along with the cognitive results from the Clinical Antipsychotic Trials of Intervention Effectiveness (CATIE) study (5), will sever some of the remaining threads.
In a three-arm randomized controlled study, Keefe et al. examined the cognitive effects of olanzapine, quetiapine, and risperidone in a large cohort (N=224) of first-episode patients. This industry-supported study complements the CATIE study by including patients in the early course of psychotic illness and involving lower medication doses. Studies of early-psychosis patients have the substantial advantage of allowing the relationships between cognitive performance and community functioning to be examined without the potential confounding effects of chronic disability and social alienation. All three medications yielded modest improvements in cognitive functioning at week 12, and there were no differences among the medications. The study did not include a first-generation antipsychotic medication, so direct comparisons are not possible. However, the improvements seen with the second-generation medications are within the range that one might expect for practice effects, as the authors indicate. For this reason, it is not known whether all three drugs have modest beneficial effects on cognition or whether they are all cognitively neutral. The cognitive effects of the medications at 52 weeks were not significant, because of a slight reduction in effect size and reduced statistical power from attrition. One conclusion from this study is that we should not plan to address our unmet need for cognition enhancement with antipsychotic medications; we need to look elsewhere.
An interesting feature of the study is the examination of the relationships between change in cognition and change in social and vocational outcome. Pooled across treatments, the partial correlations (controlling for baseline cognition and clinical symptoms) were small (0.14 and 0.18, respectively) at 12 weeks and roughly medium (0.22 and 0.36, respectively) at 52 weeks. Depending on one’s enthusiasm or skepticism, it is possible to focus on the smallness of the associations or their statistical significance. It is worth noting both. The fact that the associations are relatively small is to be expected for two reasons. One is that we do not have a clear procognitive treatment effect in this study. Second, and more important, there is no psychosocial treatment component as part of this design that would facilitate translation of any cognitive benefits to the level of community functioning. It is unlikely that cognition enhancement will yield social and occupational gains in the absence of vigorous nonpharmacological interventions that specifically target these outcome areas. All things considered, it is notable that even in the absence of these additional intervention components, change scores for cognition (for reasons that may or may not be related to medications) and community functioning were somewhat related.
Another interesting feature of this study is the inclusion of two cognitive assessment batteries of different lengths; one takes about 90 minutes to complete, and the other, 35 minutes. The conclusion here is largely expected: if one only wants a cognitive summary score, it can be approximated rather well with a short battery. However, if one needs coverage (i.e., assessment of separate cognitive domains) or enhanced reliability, a longer battery can provide those better than a short one. In this study, the individual tests from the longer battery were slightly more sensitive to relatively small treatment differences. This aspect of the article reveals a fundamental trade-off that will pull in different directions, depending on the situation and the question being asked. Consider what will happen when we see our first efficacious cognition enhancer for schizophrenia (it is only a matter of time): many clinicians will want a brief test to monitor changes. This situation will push toward the assembly of a battery with tests that are simple to administer, are brief in duration, and yield summary scores. In contrast, the search for drug effects will push in the opposite direction. There will be a desire to detect smaller effects, to have the possibility of detecting domain-specific effects, and to know whether a drug makes some domains better at the cost of making others worse. Those considerations all lead toward a somewhat longer battery, but not necessarily long by neuropsychological assessment standards. It is easy to forget that not long ago, there were debates about whether patients (and study coordinators) could tolerate a 60- to 90-minute cognitive battery in a clinical trial. A standardized and validated cognitive assessment is now the accepted price of admission for any clinical trial that wants to enter the cognitive theater.
The implication of the Keefe et al. study is consistent with a growing sentiment in the field: meaningful cognitive improvements will probably not come from second-generation antipsychotic medications. Not all second-generation medications have been adequately studied (e.g., aripiprazole), but at this point there is little reason to expect antipsychotics to provide the magnitude of cognitive improvement that we would like to see. It appears that we will need to look elsewhere to find cognition-enhancing drugs for patients with schizophrenia—drugs that work through other neuropharmacological mechanisms and that will be screened with other types of animal models. The mission of National Institute of Mental Health initiatives such as Measurement and Treatment Research to Improve Cognition in Schizophrenia (MATRICS) and Treatment Units for Research on Neurocognition and Schizophrenia (TURNS) is to stimulate the development of new drugs for cognition in schizophrenia by building a pathway for drug approval (6, 7). Now that the pathway has been constructed, there is good reason to expect that the pharmaceutical industry will embark on it and apply their considerable resources to this critically important unmet need.
Thus, with increasing experience and data, the quest to understand the interface of cognition, medication, and community functioning in schizophrenia is becoming more sophisticated. The early hope that second-generation antipsychotics alone would yield sufficient effects on cognitive performance and community functioning is fading. At the same time, there is increasing appreciation of the key role that new pharmacological agents will play. The picture is rounded out by consideration of new nonpharmacological interventions that specifically target cognition, as well as a focus on other factors that influence functional outcome, such as negative symptoms. Gains will not come easily, but the directions are clearer now, and we can start to envision complementary approaches that will yield meaningful improvements in this complex disorder.