Long-Term Changes in Cognition Among Patients With Schizophrenia Spectrum Disorders and Different Durations of Illness: A Meta-Analysis

Long-Term Changes in Cognition Among Patients With Schizophrenia Spectrum Disorders and Different Durations of Illness: A Meta-Analysis

Schizophrenia spectrum disorders (SSD) are characterized by distortions in thinking and perception, cognitive impairments, motor abnormalities, avolition and apathy, difficulties in communication, and restricted affective expression.¹ SSD affect people on multiple life domains. Deficits in cognitive functioning are a key feature of SSD.^2–4 Cognitive functioning influences improvement in other recovery domains, such as social functioning, personal recovery, and symptoms.^5–9 As cognition is a crucial part of recovery for people with SSD, and it affects other recovery domains, it is of clinical importance to know in which phase of SSD cognition can improve. Therefore, an overview of actual changes in cognition in distinct phases of SSD is a clinically relevant topic to investigate.

A variety of studies examined changes in cognitive functioning over time in SSD. These changes vary across different domains of cognition.^10–14 In general, small or no improvement of cognition was found, and people with SSD generally have lower cognitive capacity compared to healthy age-matched controls.^12,13,15 Previous research also indicated a cognitive decline in people with SSD,¹⁶ which is larger for aging patients with schizophrenia compared to healthy controls.¹⁷ Furthermore, a variety of factors, such as a younger age, higher quality and quantity of social relationships, a higher education level, a higher level of social functioning, a low severity of negative symptoms or substance abuse, and a short duration of illness (DOI), positively influenced improvement in cognition over time.^{9,11,13,18,19}

Previous literature about changes in cognition is mostly focused on patients with early psychosis. This is related to a paradigm shift to potentially prevent chronic stages of SSD^3,15 by focusing on the first years after the onset of psychosis. In this meta-analysis, we have built on this knowledge by compiling all longitudinal studies investigating changes in cognition over time, also concerning later phases of the disorder, and investigating these changes within different subgroups based on the DOI (the duration after first diagnosis of SSD) of the patient population and follow up length of the study. Furthermore, we also investigated possible moderating effects that may influence both total changes in cognition over time and changes in cognition within each DOI subgroup, as patient characteristics, levels of functioning on different domains, and needs might differ between patients with a short and a long DOI.^20,21 Our aim is to gain insights in whether changes in cognitive functioning are observed in patients with different DOIs, if these changes are achieved after a short or long time period, and which factors contribute to these changes in cognition. Previous meta-analyses already investigated longitudinal changes in cognition for people with SSD^14,15,22,23 or factors that influence changes in cognition,^18,19,24 but a meta-analysis about longitudinal changes in cognition throughout the course of SSD is missing. We aimed to answer the following questions: (1) To what extent does cognition change over the course of SSD? (2) Which moderators at baseline are associated with changes in cognition over time?

METHODS

The meta-analysis followed Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines.²⁵ Our protocol was preregistered in PROSPERO (CRD42022377107).

Data Sources

We identified records through searches in PubMed, PsycINFO, CINAHL, and Cochrane until December 2021. We used terms related to SSD, chronicity, course, and recovery (see Supplementary Table 1). Additional records were traced through references of included studies and systematic reviews.

Eligibility Criteria

Four assessors (L.d.W., K.K., R.M., and A.J.) independently selected the studies. Disagreements regarding inclusion were resolved by consensus. The included studies meet the following criteria:

1. Patient population: Studies including adults (mean age ≥18 years) who are all diagnosed with SSD²⁶ were included. Studies including patients not diagnosed with SSD were excluded.
2. Study design: Longitudinal cohort study or randomized controlled trial, with a follow-up length of at least 1 year, was included. Other study designs were excluded.
3. Outcomes: Studies reporting standardized or uncorrected quantitative and objective assessments of cognition for at least 2 time points were included. Qualitative studies and studies which outcomes could not be calculated into effect sizes were excluded.
4. Publication: Only studies published in English in peer-reviewed journals were included.

Outcome Domains

After study selection, we categorized outcomes of cognition into separate outcome domains. First, we added overall cognition as an outcome, including composite scores of cognitive assessment instruments (eg MATICS Consensus Cognitive Battery²⁷ or Brief Assessment of Cognition in Schizophrenia²⁸) and intelligence tests (eg Wechsler Adult Intelligence Scale²⁹ or Wide Range Achievement Test-Revised³⁰). Furthermore, we chose to categorize our study outcomes in subdomains of cognition following the study of Harvey³¹ and the MATRICS domains.²⁷ The categorization of study outcomes in each domain was executed by 2 authors (L.d.W. and M.T.) and checked by all coauthors. This led to the following outcome domains: sensation and perception, motor skills and construction, attention and vigilance, verbal memory, visual memory, executive functioning, processing speed, language skills, and social cognition. An overview of the categorization and definitions of outcome domains can be requested from the corresponding author.

Assessment of DOI Subgroups

Included studies investigated patients with different DOIs at baseline and assessed outcomes over different follow-up periods. Therefore, in this meta-analysis, we categorized study outcomes in subgroups based on the baseline DOI and follow-up length of our included studies following the categorization process as described in previous publications.^20,21,32 Based on the availability of study data, we categorized outcomes into 4 subgroups based on the baseline DOI: (1) DOI <5 years; (2) DOI 5–10 years; (3) DOI >10 years; (4) DOI unknown. Within each baseline DOI subgroup, we categorized studies into separate subgroups based on their follow-up length: (1) follow-up <2 years; (2) follow-up between 2 and 5 years; (3) follow-up between 5 and 8 years; and (4) follow-up >8 years.

This categorization process shows that the DOI at the follow-up assessment could overlap between the different subgroups. Nevertheless, we still expect substantial differences in changes of outcomes between subgroups and consider these subgroups as the most optimal classification for current study.

Selection and Assessment of Moderators of Outcome

We selected potential moderators at baseline through a three-step approach. First, we established 55 moderators that significantly influenced study outcomes in at least one of our included studies or in comparable reviews.^{6,15,18,22,24,33,34} Second, we extracted baseline data of these moderators from our included studies. If baseline data were available in at least 10 studies, generally indicated as the minimal number for representative outcomes,³⁵ we included this moderator in our analysis (see Statistical Analysis). Third, we added additional moderators that were considered crucial for the interpretation of our findings due to our study design: age at onset, baseline level of cognition, DOI subgroup overlap (ie, if the DOIs of all patients in the study match with the baseline DOI subgroup of the study: yes or no), publication year, the prevalence of schizoaffective disorders, study design (clinical trial or cohort study), and whether treatment was applied that targeted improvement of cognition. Based on this selection process, we selected 19 potential moderators at baseline: age at baseline; age at onset; antipsychotic use; baseline level of cognition; duration of untreated psychosis; DOI subgroup overlap; education level; ethnicity; gender; general functioning; IQ; negative symptoms; overall symptoms; positive symptoms; publication year; schizoaffective disorder diagnosis; schizophrenia diagnosis; study design (clinical trial or cohort study); and delivery of treatment targeting the outcome. For moderators that were evaluated by different assessment instruments (ie, assessment of symptoms, functioning, and baseline level of cognition), we calculated percentile scores based on normative data to ensure that each assessment was assessed in the same scale range. Due to the strict selection criteria, not all potential moderators could be analyzed in each outcome domain. For the outcome domains of sensation and perception (k=9) and social cognition (k=6), the number of included studies was too low to include in this analysis.

Quality Assessment

Quality assessment was conducted using the Quality in Prognostic Studies (QUIPS) tool.³⁶ The first author (L.d.W.) assessed all studies, and a second assessor (A.J.) independently conducted quality assessment of 10% of the studies. The level of agreement was substantial (κ = 0.72). Disagreements were resolved by consensus. We investigated the influence of study quality on outcomes through an analysis of subgroup differences.³⁵

Statistical Analysis

Meta-analytic procedure. Meta-analyses were conducted using RevMan 5.3.³⁷ We calculated effect sizes of change (Cohen d) in study outcomes by comparing outcomes at baseline and follow-up. For clinical trials, we analyzed both treatment and control groups together. Magnitude of effect was considered marginal when d<0.2, small when d≥0.2 and <0.5, medium when d≥0.5 and <0.8, and large when d≥0.8.³⁸ We used random-effects models, weighted by the method of inverse variance.³⁹ Statistical heterogeneity was assessed by calculating the I ² statistic (including 95% CI).³⁹ We controlled for multiple testing effects in all analyses through a Benjamini–Hochberg correction, with the false discovery rate set on 0.3.⁴⁰

Subgroup analyses and calculation of moderators. All study outcomes were categorized into one of the baseline DOI subgroups and subgroups based on the follow-up length. In case 1 study reported multiple outcomes within the same subgroup, we clustered all effect sizes of change within that study into 1 composite effect size of change through the method of inverse variance.³⁹ The influence of moderating effects was analyzed through a metaregression analysis using R.⁴⁰ For significant moderators, we further analyzed moderating effects within each baseline DOI subgroups between studies with high levels or presence vs studies with low levels or absence of any significant moderator, using an analysis of subgroup differences.³⁵

Handling outliers and publication bias. Outliers are defined as effect sizes of individual study outcomes which CI exceeded the upper or lower bound of the CI of the overall effect size. We controlled for the influence of outliers by comparing subgroups of all study outcomes with subgroups in which outliers are excluded through an analysis of subgroup differences.³⁵ Potential publication bias was detected by visual inspection of funnel plots.

RESULTS

Study Selection

We identified 10,477 records through database search and reference tracking. We excluded 9,573 records after title and abstract screening. From the remaining 904 records, we excluded 826 records after full-text screening. Most records were excluded because they did not report on cognition, the articles were no longitudinal studies, or the patient population was not exclusively patients with SSD (see Figure 1). The remaining 78 articles reported results of 57 studies.

Study Characteristics

The 57 studies examined changes in cognition from 6,225 patients with SSD, their mean age was 35.5 years (SD = 8.7 years; range = 21–68.7 years), and 33.7% were female (see Table 1). Twenty-nine studies (50.9%) exclusively included people with schizophrenia. Twenty-three studies (n = 3,214) specifically reported different SSD diagnoses. In these studies, 2.0% of the participants were diagnosed with brief psychotic disorders, 0.4% with delusional disorders, 2.0% with other psychotic disorders, 7.6% with psychotic disorders not otherwise specified, 11.1% with schizoaffective disorders, 9.6% with schizophreniform disorders, and 67.1% with schizophrenia. The 5 remaining studies indicated that all participants were diagnosed with SSD, without further specifications. Seven studies (12.3%) were clinical trials, and 50 studies (81.3%) were cohort studies. Sixteen studies implemented treatment programs; in 3 of those studies (all clinical trials), treatment programs specifically targeted improvement in cognition. In 19 studies (33.3%), all patients used antipsychotics. In 28 studies (49.1%), the baseline DOI was shorter than 5 years; in 7 studies (12.3%), baseline DOI was 5–10 years; in 15 studies (26.3%), baseline DOI was more than 10 years; and in 7 studies (12.3%), baseline DOI was unclear. Finally, in 23 studies (40.3%), the dropout rate was low (ie <20%); in 16 studies (28.0%), the dropout rate was moderate (ie ≥20%–≤40%); and in 17 studies (29.8%), the dropout rate was high (ie >40%). In 1 study, the dropout rate was not reported.

We observed a lower level of motor skills and construction, a higher level of attention and vigilance, and a higher severity of positive symptoms at baseline in studies with a shorter baseline DOI than in studies with a longer baseline DOI (see Supplementary Table 2).

Results of Meta-Analysis of Study Outcomes With Different DOIs

We presented a general overview of the outcomes and differences between DOI subgroups in Figure 2. Detailed information of overall cognition is reported in Table 2 and for all other outcome domains in Supplementary Table 3. In the text below, d stands for the effect size of change, I ² for heterogeneity, and k for number of studies.

Overall changes in cognition. Overall, we found marginal improvement of overall cognition (d=0.13 [0.05 to 0.22]; I ²=78%; k=30). For all other cognitive outcome domains, we found small improvements of verbal memory (d=0.21 [0.13 to 0.28]; I ²=77%; k=31) and processing speed (d=0.32 [0.22 to 0.41]; I ²=76%; k=21) over time. Furthermore, we found marginal improvements of visual memory (d=0.17 [0.07 to 0.26]; I ²=80%; k=24), executive functioning (d=0.19 [0.12 to 0.26]; I ²=75%; k=36), and language skills (d=0.13 [0.05 to 0.22]; I ²=63%; k=15). Finally, we found no significant changes in sensation and perception (d =0.10 [−0.13 to 0.33]; I ²=79%; k=9), motor skills and construction (d=0.05 [−0.07 to 0.16]; I ²=73%; k=20), attention and vigilance (d=−0.02 [−0.07 to 0.02]; I ²=84%; k=16), and social cognition (d =0.11 [−0.07 to 0.28]; I ²=59%; k=6).

Outcomes for subgroups with a baseline DOI of less than 5 years. For overall cognition, we found a small improvement after less than 2 years and 5–8 years of follow-up. For all other cognitive outcome domains, we found for both sensation and perception and motor skills and construction marginal improvements after a follow-up length of less than 2 years, a small improvement after 2–5 years of follow-up, and no significant improvement after more than 8 years of follow-up. For verbal memory, visual memory, and executive functioning, we found small improvements after less than 2 years and 2–5 years follow up, but no significant improvement after more than 8 years follow-up. For language skills and social cognition, we only found a small improvement after a follow-up length of less than 2 years, and for processing speed, after a follow-up length between 2 and 5 years. Finally, we found no significant improvement of attention and vigilance in this subgroup. For social cognition (χ²=8.83; df=1; P <.01) and attention and vigilance (χ²=19.21; df=2; P<.01), we found larger improvements after a shorter follow-up length.

Outcomes for subgroups with a baseline DOI of 5–10 years. We found no significant improvement in overall cognition over time in this subgroup. For verbal memory, we found a small improvement of outcomes after less than 2 years and 2–5 years follow-up. For executive functioning, we found a small improvement of outcomes after a follow-up length of less than 2 years. For processing speed, we found a small improvement after 2–5 years of follow-up. For attention and vigilance, we found a large improvement after a follow-up length of more than 8 years, though this was only based on 1 study. For all other outcome domains, we found no significant improvement over time in this subgroup. For executive functioning, we found a larger improvement after a shorter follow-up length (χ² = 8.79; df = 3; P < .05), and for attention and vigilance, we found a larger improvement after a longer follow-up length (χ² = 8.45; df = 2; P < .05).

Outcomes for subgroups with a baseline DOI of more than 10 years. In the subgroup of studies investigating patients with a baseline DOI of more than 10 years, we found no significant improvements in any outcome domain, except for visual memory, where we found a medium improvement over time after 5–8 years of follow-up. However, this outcome was only based on 1 study. We found no consistent differences between outcomes with a short or long follow-up length in any of the outcome domains.

Differences between subgroups based on their DOI. Analysis of subgroup differences indicated a larger improvement of overall cognition in the subgroup with a baseline DOI of less than 5 years compared with the subgroup with a baseline DOI of more than 10 years. We found similar results in the subdomains sensation and perception, motor skills and construction, verbal memory, visual memory, and processing speed. We found a larger improvement of attention and vigilance but a smaller improvement of sensation and perception and verbal memory in the subgroup with a baseline DOI of less than 5 years compared with the subgroup with a baseline DOI between 5 and 10 years. Finally, we found a larger improvement of overall cognition and verbal memory in the subgroup with a baseline DOI between 5 and 10 years compared with the subgroup with a baseline DOI of more than 10 years.

Outliers and Publication Bias

We found no outliers for motor skills and construction, language skills, and social cognition. Furthermore, we found 1 negative outlier for sensation and perception and 1 positive outlier for processing speed and attention and vigilance. For verbal memory, we found 3 negative outliers and 2 positive outliers, for visual memory, we found 1 negative outlier and 2 positive outliers, and for both executive functioning and overall cognition, we found 3 positive and 3 negative outliers. We found no indications of a significant influence in any direction due to outliers in any of the outcome domains.

We found no skewed funnel plots in any of our outcome domains (see Supplementary Figure 1), which indicates there are no indications of publication bias in the outcomes of this meta-analysis.

Analysis of Potential Moderators of Change in Outcomes at Baseline

A summary of the analysis of subgroup differences is presented in Figure 3. We present the significant moderators from our outcome domains below. More detailed statistics can be requested from the corresponding author.

Metaregression analysis showed that a younger age at baseline (B = −0.01; P < .05) and a high education level (B = 0.29; P < .05) were associated with more favorable changes in overall cognition. The moderating effects of education level were specifically indicated in the subgroup with a baseline DOI of <5 years (χ² = 14.65; df = 1; P < .01).

Furthermore, a lower baseline severity of negative symptoms was associated with more favorable changes in motor skills and construction (B = −0.01; P < .05). We specifically found this moderating effect in the subgroup with a baseline DOI of 5–10 years (χ² = 4.56; df = 1; P < .05). We also found that studies including treatment targeting cognition negatively influenced changes in motor skills and construction. However, these results are based on only 1 study in which treatment was focused on outcomes, and this study was represented in the subgroup with a baseline DOI of >10 years in which generally less favorable outcomes are achieved.

We also found that studies including more females showed better overall improvement of executive functioning (B = 0.01; P < .05) and verbal memory (B = 0.01; P < .05). For verbal memory, this was specifically indicated in the subgroup with a baseline DOI of <5 years (χ² = 5.09; df = 1; P < .05).

Finally, studies in which all participants were diagnosed with schizophrenia showed worse overall improvement of motor skills and construction (B=0.37; P<.05), visual memory (B=0.41; P<.05), and executive functioning (B=0.25; P <.05). For motor skills and construction (χ²=7.00; df=1; P <.01) and executive functioning (χ²=19.97; df=1; P<.01), this was specifically indicated in the subgroup with a baseline DOI of 5–10 years.

Quality Assessment

In general study, the quality was relatively good. However, a relatively larger number of studies reported high risk of bias for the QUIPS items study attrition and study confounding.

Analysis of subgroup differences indicated that a higher quality of study participation positively influenced changes in motor skills and construction, verbal memory, language skills, and overall cognition but negatively influenced changes in attention and vigilance. A lower quality of study attrition positively influenced changes in attention and vigilance. Furthermore, a higher quality of prognostic factor measurement positively influenced changes in motor skills and construction, attention and vigilance, and overall cognition. Finally, a higher quality of outcome measurement positively influenced changes in attention and vigilance. In conclusion, we did not find a consistent line of specific QUIPS domains influencing study outcomes. Results of the quality assessment could be requested from the corresponding author.

DISCUSSION

In this meta-analysis, we examined changes in cognition over time in patients with SSD. We found marginal improvements in overall cognition. For the cognitive subdomains, we found small improvements in verbal memory and processing speed, marginal improvements in visual memory, executive functioning, and language skill, and no significant improvements in sensation and perception, motor skills and construction, attention and vigilance, or social cognition. In all cognitive domains, except for attention and vigilance, improvement was larger for patients with a shorter DOI after a short follow-up, compared to patients with a DOI >10 years. We found no significant improvement in any cognitive domain in the subgroup with a DOI >10 years. Changes in cognition are consistently smaller in patients diagnosed with schizophrenia compared to other diagnoses of SSDs and in males compared to females. On specific domains of cognition, we found moderating effects for age, education level, and negative symptoms.

Reflection on the Influence of DOI on Changes in Cognition

Overall, the findings are in line with results of previous studies: modest cognitive improvement is only observed in people with early psychosis.^10–14,23 Furthermore, we only found improvements in cognition after a short follow-up. This is also in line with previous research, which found stable patterns of cognition with no significant improvement over time after a long follow up.^10,11,13 These short-term effects might partially be explained by a practice effect due to repeated assessments after a relatively short time, especially when parallel versions are not used. Previous research for people with mild cognitive impairment⁴¹ and schizophrenia⁴² already indicated these practice effects. Another explanation of more favorable changes in cognition after a short follow-up might be contributed by a novelty effect. This novelty effect suggests that the novelty of cognitive tasks, after baseline or first follow-up, requires extra cognitive processing and more brain activation at first, resulting in better performances after a short follow-up, especially for participants on a declining neurocognitive trajectory.⁴³ This is often followed by reduced demands of cognitive functioning once a task becomes familiar.⁴⁴ Given those modest improvements in cognition and the substantially lower level of cognitive functioning compared to healthy age matched controls,^12,13,15,17 we presume that these modest changes result in a sustained lower level of cognitive functioning at follow-up compared to healthy controls. However, as we did not include healthy controls in our analysis, we cannot test this hypothesis in our meta analysis. We found small improvements of verbal memory and processing speed, which is in line with previous meta-analyses.^14,23 However, we found less favorable indications of improvement in visual memory and executive functioning compared to previous meta-analyses,^14,23 presumably because we included different studies with a wider selection of assessment instruments, compared to previous meta analyses.

We found no significant improvement of cognition for patients with a DOI >10 years. However, in contrast to previous studies,^16,17,45 we also did not find indications of cognitive decline in this group. This might be explained by the fact that those landmark papers in the field of cognition^17,45 were mostly focused on patients with an older age (ie, patients aged above 50 years), whereas in our analysis we also focused on younger patients with a longer DOI. Taken together, included studies did not indicate cognitive impairment. However, older age negatively influenced changes in overall cognition in our meta-analysis. This suggests that cognitive decline in patients with a long DOI predominantly occurs in older age patients. Furthermore, our findings are in line with previous research that indicated cognitive deterioration throughout the development of psychotic disorders, but stabilization after the first episode of psychosis.^16,46 We found stronger indications of cognitive improvement in patients with a shorter DOI. These differences between DOI subgroups could not be explained by differences in cognitive impairment at baseline. This strengthens our finding that, regardless of cognitive impairment, the highest potential of cognitive improvement could be achieved earlier in the course of illness, especially in patients with a DOI <5 years.

The Influence of Moderators of Changes in Cognition

We found several moderating effects on changes in cognition. First, we found significantly less favorable changes in verbal memory, visual memory, and executive functioning in studies that included more males. Previous studies did not find moderating effects of gender.⁴⁶ However, previous studies did indicate better verbal memory performances in females.^46–48 Nevertheless, our findings might suggest that tailoring interventions targeting cognitive improvement on gender differences might be necessary to achieve optimal results.⁴⁹ For example, we could investigate potential gender-specific adaptations in cognitive tasks to practical learning situations that are more relatable for men or women to facilitate them better in developing strategies to compensate for their cognitive deficits.

Furthermore, we found less favorable changes in motor skills and construction, visual memory, executive functioning, and processing speed for patients who were diagnosed with schizophrenia compared to other diagnoses of SSD. Possibly, schizophrenia patients are more affected by the consequences from neurodevelopmental disruptions compared to patients with other psychotic disorders.^3,50 Our finding that more severe negative symptoms are associated with less favorable cognitive improvement, combined with previous findings indicating a more chronic pattern of negative symptoms throughout the course of illness for patients with schizophrenia,²¹ might also partially explain these moderating effects.

We also found that people with an older age and lower education level show less favorable changes in overall cognition. This may be explained by an overrepresentation of subgroups with an older age in the subgroup with a baseline DOI >10 years. Previous research indicated that cognitive aging may be accelerated for people with SSDs.⁵¹ This might suggest that age and DOI together reinforce less favorable changes in cognition. Furthermore, the association between higher education levels and cognitive improvements is in line with previous research.^11,52 A possible explanation is that patients with a higher education level are more exposed to cognitive activities and therefore better equipped against cognitive decline.⁵³ This explanation is in line with the cognitive reserve theory, which states that longer education and more involvement in cognitive activities in early life might be a protective factor for the development of functional and cognitive limitations in later life.^54,55

Finally, we found that a lower severity of negative symptoms is associated with better improvement of motor skills and construction. This is in line with previous studies that indicated that certain aspects of negative symptoms, especially diminished emotional expressions, are overlapping and interrelated with motor skills and construction.^56,57 Therefore, a focus on improvement of negative symptoms might also contribute to improvements in motor skills and construction.

Limitations

Several limitations should be considered. First, we evaluated changes in cognition on a study level. Therefore, we could not entirely grasp the clinical diversity of our target group and their unique individual process of cognitive changes within each study. Second, the findings concerning the domains sensation and perception and social cognition are based on a limited number of studies, making these outcomes less reliable.³⁵ Furthermore, we included studies conducted in different contexts and using a wide variety of assessment instruments with both uncorrected and standardized scores. This inevitably leads to heterogeneity.⁵⁸ We attempted to explore this clinical heterogeneity through our analyses of moderating effects in this meta-analysis and through an analysis of baseline differences of those moderators between DOI subgroups. Although several of these analyses were based on a limited number of studies, we propose that the combination of these analyses gave reliable insights into subgroup differences which were taken into account in the interpretation of our findings. Another limitation is that DOI in our subgroups was based on the mean DOI of the study sample. Therefore, it is possible that a part of the sample in a given study has a shorter or longer DOI than the upper or lower limit of the DOI subgroup. We controlled for this overlap in DOI and did not find indications that this significantly influenced our study outcomes. Another limitation is that we only focused on objective assessments of cognition and no subjective ratings of cognitive functioning. These subjective ratings could give some valuable additional insights into changes of cognition and are therefore an important topic for future research. Finally, our inclusion criteria are relatively strict by only including longitudinal studies exclusively investigating patients with SSDs with extractable data of cognition on multiple time points and a follow-up length of at least 1 year. As a consequence, long-term outcomes might be based on a selective sample of possibly higher functioning patients who were still able to participate during a long follow-up. Especially, we included a small number of clinical trials targeting cognitive improvement as these generally selected a broader target group, also including patients with no primary diagnosis of SSD, or investigated changes over a shorter follow-up period of less than 1 year. However, we decided to use these strict inclusion criteria to diminish heterogeneity. Additionally, the longer follow-up of our included studies might give a selective, more positive, indication of our outcomes as only patients with a higher level of functioning might be able to complete the cognitive tasks at follow-up as well. This effect of a selective sample might have been applicable for the outcomes of attention and vigilance, as a lower study quality on study attrition positively influenced changes in this outcome domain, and attention and vigilance was the only outcome domain with more favorable outcomes after a long follow up. However, for all other cognitive outcome domains, we did not find any indications for a selective group that remained at follow-up.

Conclusions

Based on the findings of our meta-analysis, we can conclude that people with SSD show modest cognitive improvements in some cognitive domains up until 10 years after their first diagnosis of SSD. This gives a slightly more optimistic view than the conclusion of some previous studies that improvement of cognition is not possible for people with psychosis.^16,45,59 Nevertheless, there is still a long way to go in the recovery of cognition for patients with SSD. Future research could focus on the development of treatment specifically focused on cognitive improvement as early as possible and the interrelationships of cognitive changes with other life domains. Especially, improvement of negative symptoms might lead to more substantial improvements of cognition over time.

Article Information

Published Online: September 25, 2024. https://doi.org/10.4088/JCP.23r15134
© 2024 Physicians Postgraduate Press, Inc.
Submitted: October 10, 2023; accepted July 1, 2024.
To Cite: de Winter L, Jelsma A, Vermeulen JM, et al. Long-term changes in cognition among patients with schizophrenia spectrum disorders and different durations of illness: a meta-analysis. J Clin Psychiatry. 2024;85(4):23r15134.
Author Affiliations: Phrenos Center of Expertise, Utrecht, the Netherlands (de Winter); Department of Psychiatry, Amsterdam UMC Location AMC, Amsterdam, the Netherlands (de Winter, Jelsma, Vermeulen, van Tricht, de Haan); Tranzo, Tilburg University, Tilburg, the Netherlands (van Weeghel); Department of Psychology, Bar-Ilan University, Ramat-Gan, Israel (Hasson-Ohayon); Epidemiological and Social Psychiatric Research Institute, Erasmus MC, Rotterdam, the Netherlands (Mulder); NHL Stenden University of Applied Science, Leeuwarden, the Netherlands (Boonstra); University Medical Center Utrecht, Utrecht, the Netherlands (Boonstra); University of Groningen, University Medical Center Groningen, Groningen, the Netherlands (Veling); Parnassia Psychiatric Institute, The Hague, the Netherlands (Mulder).
Corresponding Author: Lars de Winter, MSc, Department of Psychiatry, Amsterdam UMC Location AMC, Amsterdam UMC, Meibergdreef 9, Amsterdam, AZ 1105, the Netherlands (Lwinter@kcphrenos.nl).
Relevant Financial Relationships: None.
Funding/Support: None.
Acknowledgments: We would like to thank and acknowledge Kete Klaver, MSc (KK), from The Netherlands Cancer Institute and Ravi Meijer, BSc (RM), from Phrenos Center of Expertise for their important contribution to the study selection. We would also like to thank Matthijs Oud, MSc (MO), from Trimbos Institute for his advice and support in the quality assessment procedure. They did not report any conflicts of interest or relevant financial relationships for this study.
Data Availability Statement: The majority of relevant data and materials are presented in the tables and supplementary materials, as well as partially available in the review protocol that is submitted in PROSPERO (CRD42022377107). All other data are not available online. Further questions and requests about availability of the data could be sent to the corresponding author.
Supplementary Material: Available at Psychiatrist.com.

Clinical Points

• Cognitive functioning is crucial for recovery of people with schizophrenia spectrum disorders (SSD). Therefore, it is important to investigate longitudinal changes in cognition in SSD.
• We found only modest improvement of cognition, specifically in people with early psychosis.
• It is important to address improvement of cognitive functioning early in the course of SSD.

The post Long-Term Changes in Cognition Among Patients With Schizophrenia Spectrum Disorders and Different Durations of Illness: A Meta-Analysis appeared first on Psychiatrist.com.