Cognitive Remediation and Social Cognition Training in Clinical High-Risk Populations

confidence: high · 5 sources

⚠
Working research — not a recommendation.

This page records what the evidence says a policy does, not what anyone should do about it. It has been synthesized from the cited sources but has not been independently reviewed.

What the evidence says

The evidence base for cognitive remediation (CR) in clinical high-risk (ultra-high-risk, UHR) populations is small. A 2015 trial protocol reports that its search found "only two published studies" of CR in UHR patients. A 2017 systematic review of this literature reports processing speed gains, in a double-blind RCT (Choi et al., 2016) against an active control, with effect sizes of d=0.50–0.84. The gains held at 4-month follow-up. A cohort study without a control group reports d=0.63. A cohort study that compared UHR individuals with patients with schizophrenia reports an attention gain of d=0.69.

Other outcomes show no measured effect. In the 2017 review, zero of five studies that reported clinical symptoms found that CR changed them. A later RCT (Glenthøj 2020) found no between-group difference at treatment cessation on global neurocognition (b=–0.125, 95% CI –0.423 to 0.172, p=0.41). It also found none on functioning, psychiatric symptoms, negative symptoms, or depression. Its one positive result came from an exploratory outcome. The CR group recognised emotions faster (latency b from –152.0 to –226.8, p from 0.01 to 0.002).

Functioning is mixed. Two of four studies in the 2017 review found gains in social functioning or social adjustment, with d=1.04–3.09 in the positive studies. The other two found none.

Two outcomes people expect to be covered have no usable evidence. No study measured the effect of social cognitive training on social cognition in UHR individuals. The only trial that added social cognitive remediation had no social cognitive outcome measure. The effect of CR alone on transition to psychosis cannot be isolated, because the only study of it delivered CR inside an integrated treatment.

Data from related populations give larger and more consistent effects. In schizophrenia, a meta-analysis cited in the 2015 protocol reports global cognition at d=0.45 (95% CI 0.31–0.59) and functional outcomes at d=0.42 (95% CI 0.22–0.62). The cognitive effect persisted at follow-up (d=0.43, 95% CI 0.18–0.67). In inpatients with psychosis, a meta-analysis reports processing speed at g=0.48, memory at g=0.48, and working memory at g=0.56. Average dropout was 7%. Pairing CR with psychiatric rehabilitation is associated with larger functional effects. A meta-analysis of 23 studies and 1819 patients reports SMD=0.41 for employment rate and SMD=0.24 for social skills. These are not UHR estimates.

A narrative review (Gautam 2025) says cognitive impairment is more responsive to remediation in early psychosis than in later stages. It says younger people respond better, and it proposes brain plasticity as the reason. It gives no effect sizes for either claim.

Where the evidence splits

The UHR findings on cognition disagree. The 2017 review reports processing speed and attention gains (d=0.50–0.84, d=0.63, d=0.69). The 2020 RCT found no effect on global neurocognition and no effect on any clinical or functional primary or secondary outcome. The sources on this page do not settle why. The outcomes differ: processing speed and attention are single domains, while the 2020 trial's primary outcome was a global composite. The designs also differ. The positive estimates come partly from uncontrolled or non-UHR-comparator cohorts. The 2020 authors offer a post-hoc explanation: participants attended an average of 10.9 sessions and 11.9 hours of neurocognitive training, "about half the usual dose" for first-episode schizophrenia. No dose-response data test this.

Clinical symptoms disagree across populations. Gautam (2025) states that CR improves clinical symptoms in early-stage psychosis. The UHR sources find no measured effect (0 of 5 studies; the 2020 RCT). The population difference is the only stated contrast. The sources do not say whether it explains the split. The 2017 review also lists as a possibility that symptom measurement was premature.

Functioning is split within UHR studies. Two of four studies found gains (d=1.04–3.09) and two found none. The 2020 RCT also found no effect on its functioning scale (PSP). The sources do not explain which studies differ or why. One secondary analysis finds that better baseline social and role functioning predicted larger emotion recognition gains.

Thresholds and dose-response

No threshold established in the current sources.

Open empirical questions

Findings

Cognitive remediation increases processing speed in UHR individuals

confidence — direction: highconfidence — magnitude: mediumconsensus: thin. The weight leans toward an increase, but it rests on one RCT and one uncontrolled cohort.
mechanism
targeted training of cognitive functions. The sources do not test a mechanism for processing speed specifically.
effect
increase. d=0.50–0.84 in a double-blind RCT against an active control, post-treatment and at 4-month follow-up. d=0.63 in an uncontrolled cohort.
conditions / moderators
not established. The 2020 RCT found no effect on a global neurocognition composite (see below).

Cognitive remediation increases attention in UHR individuals

confidence — direction: mediumconfidence — magnitude: mediumconsensus: thin
mechanism
not established.
effect
increase, d=0.69, in a cohort study that compared UHR individuals with patients with schizophrenia receiving the same program.
conditions / moderators
the comparator was a group with schizophrenia, not an untreated UHR group.

Cognitive remediation plus treatment as usual has no measured effect on global neurocognition in UHR individuals

confidence — direction: highconfidence — magnitude: highconsensus: thin. This is a single RCT, and it conflicts with the domain-level gains above.
mechanism
not established. The authors suggest the dose may have been too low.
effect
no measured effect at treatment cessation on the BACS composite (b=–0.125, 95% CI –0.423 to 0.172, p=0.41).
conditions / moderators
participants averaged 10.9 sessions and 11.9 hours of training, about half the usual first-episode schizophrenia dose. This is a post-hoc interpretation, untested.

Cognitive remediation has no measured effect on clinical symptoms in UHR individuals

confidence — direction: highconfidence — magnitude: highconsensus: settled for UHR populations, leaning toward no effect. Gautam (2025) reports symptom gains in early-stage psychosis, so the result is contested across populations.
mechanism
not established.
effect
no measured effect. Zero of five studies in the 2017 review found a change in symptoms. The 2020 RCT found none on psychiatric, negative, or depressive symptoms (BPRS-E, SANS, MADRS).
conditions / moderators
the 2017 review notes that symptom measurement may have been premature.

Cognitive remediation produces mixed evidence on social and role functioning in UHR individuals

confidence — direction: mediumconfidence — magnitude: mediumconsensus: contested
mechanism
not established.
effect
mixed. Two of four studies found gains in social functioning or social adjustment (d=1.04–3.09). Two found none. The 2020 RCT found no effect on the PSP scale.
conditions / moderators
the sources do not identify which individuals benefit. The 2017 review calls for work on what predicts benefit.

Cognitive remediation plus treatment as usual decreases emotion recognition latency in UHR individuals

confidence — direction: highconfidence — magnitude: mediumconsensus: thin
mechanism
not established. The authors suggest possible cognitive plasticity, if participants engage enough to practise the skills.
effect
faster emotion recognition, latency b from –152.0 to –226.8 (p from 0.01 to 0.002) at treatment cessation.
conditions / moderators
exploratory outcome. In secondary regression, better baseline social and role functioning predicted larger gains.

Baseline status is associated with the size of gains in UHR individuals

confidence — direction: mediumconfidence — magnitude: lowconsensus: thin
mechanism
not established. The pattern for worse baseline performance is consistent with regression to the mean.
effect
worse baseline executive function and visual memory predicted larger gains in those domains at 12 months. Better baseline social and role functioning predicted larger emotion recognition gains.
conditions / moderators
secondary regression analyses of one RCT. Gautam (2025) separately reports better response in adolescents and early adults and greater responsiveness in early-stage illness, with no effect sizes.

Social cognitive training has no measured effect on social cognition in UHR individuals because the outcome was not assessed

confidence — direction: not applicableconfidence — magnitude: not applicableconsensus: thin. This is an absence of evidence.
mechanism
not applicable.
effect
no evidence. The one trial that added social cognitive remediation had no social cognitive outcome measure.
conditions / moderators
not established.

The separate effect of cognitive remediation on transition to psychosis cannot be determined

confidence — direction: not applicableconfidence — magnitude: not applicableconsensus: thin. This is an absence of evidence.
mechanism
not applicable.
effect
no isolable estimate. The only study delivered CR as part of an integrated treatment.
conditions / moderators
not established.

Cognitive remediation increases cognition and functional outcomes in schizophrenia and psychosis populations

confidence — direction: highconfidence — magnitude: mediumconsensus: settled for cognition in these populations, leaning toward an increase. Functional effects are mixed evidence.
mechanism
training of cognitive functions. Gautam (2025) proposes that CR influences brain plasticity and restores neural networks in early psychosis. The claim reports no neuroimaging measurements.
effect
increase. In schizophrenia, global cognition d=0.45 (95% CI 0.31–0.59), functional outcomes d=0.42 (95% CI 0.22–0.62), and cognition at follow-up d=0.43 (95% CI 0.18–0.67). In inpatients with psychosis, processing speed g=0.48, memory g=0.48, working memory g=0.56. Average dropout was 7%.
conditions / moderators
the inpatient meta-analysis found heterogeneity across studies and moderate methodological quality. It found an indication of gains in vocational, social, and global functioning, but "less reliable". Functional effects were greater when psychiatric rehabilitation was also provided. In a meta-analysis of 23 studies and 1819 patients, combined rehabilitation and CR gave SMD=0.41 for employment rate and SMD=0.24 for social skills.