Cognitive Trajectories in Schizophrenia
Acting Chair: Philip Harvey
Acting Co-chair: Kiri Granger
Predicated on data available at the time, in 2016 the ISCTM published a consensus paper in which several important issues around patient selection for the treatment of cognitive impairment in schizophrenia (CIAS) trials were addressed:
For example, is everyone with schizophrenia a candidate for treatment of impairments in cognition? If not, how are treatment initiation decisions made? The discussions focused on the questions of whether patient selection should be based on:
• age and duration of illness
• baseline level of cognition
• baseline level of everyday functioning
• severity of positive symptoms
At the time of publication (see Keefe et al., 2016), these were topics for which no studies had been conducted, with few data existing to guide decision making. In the absence of product-specific data to guide answers to these questions, the panel debated these points in general terms to reach a consensus. With pharmacological agents and remediation-based treatments for CIAS yet to be approved by the FDA and/or other regulatory agencies, several stakeholders are increasingly interested in evidence-based answers to these questions, including payers. The following section reviews the current consensus regarding the issue of patient selection based on ‘baseline level of cognition’ before highlighting published findings since 2015 which call for the current consensus around this issue to be discussed and revised.
Patient Selection with Respect to Level of Cognitive Impairment
The discussion on patient selection with respect to baseline level of impairment was focused on whether there should be predilection for selecting patients with a higher level of cognition or functioning.
The assumption is that treatment does not have a preferential effect based on baseline level of cognition. The principal arguments in support of such a position are that:
(1): The less impaired patients have a greater cognitive reserve and may thus have a greater viability of neuroplasticity that is more amenable to change; (2) may be able to demonstrate greater functional gains since they are likely to be in a more active cognitive environment that will provide opportunities to experience the reward associated with a true cognitive benefit. These patients are also assumed, in the absence of evidence to prove it in most cases, to be patients with a higher level of premorbid intellectual functioning with equivalent levels of illness related decline compared to patients with both lower current test scores and lower premorbid functioning. Thus, hypothetically these would be patients for whom cognitive enhancing treatments would have the most potential to be restorative.
(2): The potential functional improvement in a patient who requires a substantial amount of support, unless the effect of the intervention is large, is unlikely to result in a major change in dependency status.
There were also arguments in support of preferentially selecting for treatment with greater levels of impairment:
(1): Depriving an effective medication from a severely ill segment of the population may be unethical.
(2): Cognitive and functional improvements in persons with low levels of cognition may actually be easier to detect than in those with higher cognitive ability. An analogy to this example is the use of clozapine in treatment-resistant patients where improvement is easy to detect. Furthermore, functional milestones are applicable to all patients, regardless of their cognitive ability. The comparison of whether the milestones of a person with low functional ability are as meaningful as those in patients with high functional ability requires a social value judgement.
(3): These lower performing patients may not be adequately represented in treatment trials to date and need to be recruited to provide a representative sample of schizophrenia patients.
Consensus at the time: “Which patients receive treatment should not depend upon their baseline level of cognitive impairment.”
Recent Findings & Publications: Neuropsychological normality in schizophrenia
Recent post hoc analyses from a large phase II CIAS trial suggests that a significant proportion of subjects enrolled performed, at screening, within the neuropsychologically normal range of the population normative data for the MCCB composite score (t score >40) and that these individuals demonstrated different cognitive trajectories across the course of the trial visits than those subjects with greater impairment. The argument, launched at the time of the MATRICS Process, that all patients with schizophrenia had definite cognitive impairment was based on data from observational data, often from patients with chronic symptoms and institutional stays. In some large observational studies relatively few patients met criteria for unimpaired cognitive performance. Leung et al., (2008) studied ambulatory patients only and found an 18% rate of neuropsychological normality, while 47% were in clinical remission. Reichenberg et al., (2009) studied clinically stabilized first episode patients and found that 16% met conservative criteria for neuropsychological normality. In an outpatient study focused on donepezil treatment of CIAS in schizophrenia, Friedman (2002) required that patients score at least 2 SD below normative standards on the CVLT (t-score equivalent: 30). 36/36 otherwise eligible patients met this criteria.
In contrast, in our recent CIAS trial, approximately 25% did not manifest cognitive impairments (exceeding 1-1.5 SD’s below the MCCB normative data range), thus suggesting the seemingly paradoxical result that offering cognitive treatments in a clinical trial may lead to the acquisition of a less cognitively impaired sample. This may be due to the fact that CIAS trials using the MATRICS standards are so complex and demanding that more cognitively impaired patients may not be able to enter or complete them, ultimately resulting in a recruitment bias.
Partially consistent with this point is the report by Georgiades et al. (2017) that included 2616 stable patients entering CIAS clinical trials (all of which had a negative result), found that 2/9 MATRICS neurocognitive tests had mean t scores of 40 or more at baseline. Although the MCCB composite scores were lower on average, partially because of the way that they are calculated, the mean t-score of the individual neurocognitive tests in the MCCB was 37.6, which is only .24 SD less than the cutoff for neuropsychological normality on the assessments. Thus, patient samples selected for these trials must have had a large number of cases whose overall performance was in the neuropsychologically normal range. We have those data and will examine them as part of the working group.
Recent analyses of the correlates of neuropsychological normality have been performed. In a large sample of patients with schizophrenia (n=554; Strassnig et al., in press) those who had MCCB t scores greater than 40 (N =154) did not manifest a significant correlation between everyday functional performance and cognitive performance. In contrast, everyday functioning and work outcomes were both significantly correlated with cognitive test performance. While the less impaired patients had better functional outcomes, in the group with minimal cognitive impairment their cognitive and functional performance did not correlate. This raises the question of whether these are lifelong functional advantages in the high functioning group and question whether cognitive enhancement could improve outcomes if outcomes are not correlated on an individual basis with functioning.
The objective of the working group will be to review the available data pertaining to cognitive trajectories in recent clinical trials, in addition to those mentioned above, to ascertain the potential need to modify the recommendations for inclusion / exclusion criteria (i.e., patient selection) for clinical trials in CIAS.
Specifically, are there individuals who demonstrate cognition in the “normative range” and if so, do they demonstrate trajectories that suggest a decreased likelihood of signal detection?
What are the origins of these higher levels of functioning seen in clinical trials now? Are these selection/recruitment biases? Are they artifacts of the complexity of trial design and the requirements on the parts of patients to successful participate? Do we need to change our trial design so that lower functioning patients can participate and complete the trial? Do we want a more systematic selection process? Do we want to recruit higher performing patients if we can prove that their performance has in fact worsened and if so by how much? Do we want to re-evaluate our go/no-go criteria for drug success by defining statistical criterion that is different / optimal for low versus high performers?
The deliverable for this working group will be a second smaller consensus workshop and depending on results, either a report to ISCTM community through web postings or a 2nd white paper.
Process: Stakeholders with data informing on this question will be asked to perform similar analyses to ascertain the replicability / strength of the findings.