Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2022 Mar 23.
Published in final edited form as: Neurosci Biobehav Rev. 2019 Sep 23;107:828–845. doi: 10.1016/j.neubiorev.2019.09.031

Multi-Outcome Meta-Analysis (MOMA) of Cognitive Remediation in Schizophrenia: revisiting the relevance of human coaching and elucidating interplay between multiple outcomes

Lana Kambeitz-Ilankovic 1,2,*, Linda T Betz 1,*, Clara Dominke 1, Shalaila S Haas 3, Karuna Subramaniam 4, Melisa Fisher 5, Sophia Vinogradov 5, Nikolaos Koutsouleris 2,*, Joseph Kambeitz 1,*
PMCID: PMC8942567  NIHMSID: NIHMS1762981  PMID: 31557548

Introduction

Impairments of neurocognitive functions are among the core features of schizophrenia (1) and affect multiple domains including working memory, verbal memory or executive functioning (2). Importantly, cognitive deficits are associated with future impairments in psychosocial functioning (3,4) and contribute significantly to the disease burden of affected schizophrenia (SZ) patients as indicated by low occupational status, poor social and community functioning and reduced quality of life (5–7). In current clinical practice, the therapy of schizophrenia is mainly based on antipsychotic and antidepressive drugs, that are effective in the reduction of psychotic and depressive symptoms but often leave cognitive deficits unaltered or even aggravate them (8–10).

Recent results suggest that newly developed intervention programs based on cognitive remediation (CR) are a promising approach to improve cognitive and psychosocial functioning in SZ patients (11–13) and are at the same time cost-effective. Digitalization of the therapy process has increased the usage of computerized CR in the last two decades. Computerized CR offers several advantages compared to traditional CR including structured application of the intervention, training at home, automatic adaption of the exercise difficulty according to patients’ performance and online tracking of patients’ training habits (14–17). CR is often administered in a ‘drill-and-practice’ form with repetitive exercises and in a game-like fashion that stimulates patients’ reward processing systems (18) with potential subsequent effects on negative symptoms (19) and psychosocial functioning (20). In some cases, CR is supplemented by human guidance that refers to therapeutic support in form of strategy provision, employment programs, meta-cognitive training that either coaches or motivates the patient.

The most comprehensive meta-analysis from 2011 (21) on the effects of CR including broader set of studies demonstrated small but robust improvements in cognition and psychosocial functioning, based on data of over 2000 SZ patients. The analysis revealed substantial heterogeneity in the study design of the different publications and potential moderating effects such as stronger improvements in psychosocial functioning in studies with complementary psychiatric rehabilitation. Recently, a recent meta-analysis on solely drill and practice computerized CR (22) found no effects on functional outcome of SZ patients, despite the fact that multiple cognitive domains, including processing speed, have shown improvement. Nonetheless, a systematic attempt to disentangle the contribution that supplementary human guidance (SHG) might have on diverse outcome measures has not taken place yet.

Since the publication of the last comprehensive analysis (21) a large number of new studies investigating the effects of CR in schizophrenia have been published and the more recent meta-analyses are limited to the clinical subdomain of negative symptoms (19) and ‘drill-and-practice’ computerized CR (22). The latest meta-analysis excluded studies with supplementary therapeutic coaching which is useful for investigating effects of the approach in a more homogenous sample of studies. However, this does not entirely resemble the therapeutic reality in which SZ patients are mostly exposed to a combined approach of SHG and computerized CR.

Finally, multiple studies reported that improvement in global and specific cognitive domains co-occurs with the increases in psychosocial functioning (11,23) or decreases in psychotic symptoms (19,24,25), but directionality of these changes cannot be inferred. The causation may be multi-factorial and bi-directional but based on previous studies, two models appear most probable: 1) cognitive improvement including domains of social cognition and verbal memory directly leads to better functional outcome (6,23,26,27) which in turn may lead to remission of symptoms or 2) the relationship between cognition and functional outcome is mediated by symptoms and in particular negative symptoms (28,29). Thus, we conducted an up-to-date meta-analysis of the effects of CR that deliberately combines computerized CT and SHG in SZ patients, investigating the effects on cognitive performance, psychosocial functioning and clinical symptoms separately for both approaches.

Moreover, we conducted an exploratory structural equation modeling (SEM) analysis on a single-study data set to test for directed effects between improvements in cognitive performance, psychosocial functioning and clinical symptoms. This comprehensive investigation of computerized and SHG training modalities along with moderator effects could potentially explain the heterogeneity of current results and pave the way to a more personalized CR.

Methods

Literature search & data extraction:

We conducted a systematic literature search in the PubMed database to identify all relevant studies published until June 2019. The following search term was used: (cognitive”[All Fields] OR “cognition”[All Fields]) AND (“training”[All Fields] OR “remediation”[All Fields] OR “rehabilitation”[All Fields] OR “enhancement”[All Fields] OR “intervention”[All Fields]) AND (“schizophrenia”[All Fields] OR “schizophreniform”[All Fields] OR “non-affective psychosis”[All Fields] OR “schizoaffective”[All Fields]) AND (“computer”[All Fields] OR “computerized”[All Fields] OR “computerised” [All Fields] OR “digital”[All Fields]) AND (“1950/01/01”[PDAT] : “2019/06/01”[PDAT]). Our search identified a total of 690 studies. After removing 437 studies, the remaining 253 studies were screened for inclusion according to our inclusion criteria (Fig.1). Statistical information was independently extracted from the included studies by two researchers (C.D. and L.B.) and was then compared until consensus was reached (between L.K.I, C.D., and L.B.). Additionally, we assured that all the relevant studies from the previous meta-analysis were included (21,22). The overview of the selection procedure and inclusion criteria is given in the flowchart, Figure 1.

Fig.1.

Fig.1.

Open in a new tab

Flowchart of the study selection procedure.

We restricted our analysis to studies with at least 70% of patients with schizophrenia, schizophreniform disorder, non-affective psychosis or schizoaffective disorder as diagnosed by the DSM-IV. Also, to be included in our analysis studies needed to report effects in SZ patients receiving computerized CR according to the standard Cognitive Remediation Experts Workshop definition of cognitive remediation (12,30) and these effects needed to be compared to a control group undergoing active or passive placebo treatment (e.g. video games as opposed to treatment as usual). In case two control groups were presented in the study we used both and used the intervention group twice for comparison whereas if there were two intervention groups, only one control group was used. A comprehensive overview of all the individual studies and their demographic and clninical characteristics is given in the Table 1. General overview with the descriptive measures is shown in the Table 2.

Table 1.

Demographic and clinical characteristics of SZ samples included in the individual studies.

First Author Year Type of CR Type of CG Males TG (%) Males CG (%) Diagnoses TG (%) Diagnoses CG (%) Ethnicity TG (%) Ethnicity CG (%) Illness duration TG (y) Illness duration CG (y) CPZ TG (mg) CPZ CG (mg) Type of supplement to CR Duration CR (h) Duration Supplement (h) Exposure (yes/no) Treatment spread (weeks)
Balzan 2019 Happy Neuron Pro Metacognitive Therapy + 63.0 55.6 74.1 % SCZ; 18.5 % SZA; 7.4 % Psychosis other 66.7 % SCZ; 22.2 % SZA; 11.1 % Psychosis other NA NA 12.37 9.85 425.00 609.11 NA 7 0 yes NA
Bark 2003 Memory or problem solving tasks TAU 66.7 44.4 77.8 % SCZ; 22.2 % SZA 72.2 % SCZ; 27.8 SZA NA NA NA NA NA NA NA 4.17 0 yes 5
Bell 2008 NET Employment Program 60.5 47.1 74 % SCZ; 26 % SZA 65 % SCZ; 35 % SZA 50 % African American; 2.5 % Asian; 45 % Caucasian; 2.5 % Hispanic 44 % African American; 50 % Caucasian; 6 % Hispanic NA NA NA NA Employment Program 113.75 49 yes 52
Bellucci 2003 Captain’s Log software TAU 47.1 (total sample) 47.1 (total sample) 47.1% SCZ; 52.9% SZA (total sample) 47.1% SCZ; 52.9% SZA (total sample) NA NA 16.6 (total sample) 16.6 (total sample) NA NA NA 8 0 yes 8
Benedict 1994 Attention task TAU 50.0 52.9 100 % SCZ 100 % SCZ NA NA NA NA 279.7 346.1 NA 15 0 yes 7.14
Bryce 2018 CogPack CG 65.5 74.1 79% SCZ; 21% SZA 63% SCZ; 37% SZA 90% Caucasian 89% Caucasian NA NA 738.45 666.35 Strategy provision 12.95 NA yes 10
Burda 1994 Captain’s Log software TAU NA NA 68.1% SCZ; 38.1% SZA (total sample) 68.1% SCZ; 38.1% SZA (total sample) 88.4% Caucasian and Hispanic; 11.6% Afroamerican (total sample) 88.4% Caucasian and Hispanic; 11.6% Afroamerican (total sample) NA NA NA NA NA 12 0 no 8
Byrne 2013 Own program TAU 100.0 100.0 100 % SCZ 100 % SCZ NA NA 19.44 24.92 443.37 361.50 NA 9.59 0 no 6
Byrne 2015 Computer-assisted cognitive and facial affect recognition TAU 100.0 100.0 100 % SCZ 100 % SCZ NA NA 12.7 13.2 250 226 NA 11.13 0 no 6
Cassetta 2018 BrainGymmer: working memory tasks TAU 56.5 58.3 65.2% SCZ; 34.7% SZA 45.8% SCZ; 54.2% SZA 65.2% Caucasian; 30.4% Asian; 4.3% African descent 75% Caucasian; 8.3% Asian; 8.3% Indigenous; 4.2% African descent; 4.2% Hispanic 11.04 12.82 NA NA NA 8.29 0 no 10
Cassetta 2018 BrainGymmer: processing speed tasks TAU 54.2 58.3 70.8% SCZ; 29.2% SZA 45.8% SCZ; 54.2% SZA 75 % Caucasian; 16.7 % Asian; 8.3% Indigenous 75% Caucasian; 8.3% Asian; 8.3% Indigenous; 4.2% African descent; 4.2% Hispanic 17.00 12.82 NA NA NA 6.86 0 no 10
Cavallaro 2009 CogPack Computer-aided non-domain-specific activity NA NA 100% SCZ 100% SCZ NA NA 8.28 8.08 NA NA Standard rehabilitation program (skills training, psychoeducation) 36 9 yes 12
Dickinson 2010 Own program Computer Skills training 65.7 75.0 80 % Schizophrenia 75 % Schizophrenia 65.7 % Afroamerican 53.6 % Afroamerican NA NA NA NA Strategy provision 21.46 10.73 yes 15
D’Amato 2010 REHACOM TAU 74.4 76.3 100% SCZ 100% SCZ NA NA 8.7 8.1 337 441 NA 28 0 yes 7
D’Souza 2013 CogRehab Video viewing NA NA NA NA NA NA NA NA NA NA NA 60 0 no 12
Donohoe 2017 Own program (online) Social contact 64.6 54.8 60.4% SCZ; 10.4% SZA; 10.4% BD; 2.1% MDD; 16.7% other psychosis 66.7% SCZ; 11.9% SZA; 7.1% BD; 4.8% MDD; 9.5% other psychosis NA NA NA NA 395.10 599.35 Strategy provision 23.3 6 yes 12
Drake 2014 CIRCuiTS Social contact 67.7 53.3 83.9% SCZ; 16.1 SZA 86.65 SCZ; 13.3% SZA 74.2% White; 3.2 African and African Caribbean; 22.6% Asian 83.3% White; 6.6% African and African Caribbean; 10% Asian NA NA NA NA NA 6.6 0 yes 12
Eack 2009 CET EST NA NA 65.5% SCZ; 35.5% SZA (total sample) 65.5% SCZ; 35.5% SZA (total sample) 69.0% Caucasian; 19.0% African American; 10.3% Asian; 1.7% Other (total sample) 69.0% Caucasian; 19.0% African American; 10.3% Asian; 1.7% Other (total sample) 3.19 (total sample) 3.19 (total sample) NA NA Social-cognitive groups 60 67.5 yes 104
Fernandez-Gonzalo 2015 NeuroPersonal Trainer-Mental Health Computer Skills Training 60.7 68.0 78.6 % SCZ; 21.4 % SZA 84 % SCZ; 26 % SZA NA NA 2.3 3.01 269.73 244.07 NA 30.7 0 yes 17
Fan 2017 Own program TAU 58.3 45.5 100% SCZ 100% SCZ NA NA 16.30 18.55 553.9 412.98 NA 30 0 yes 8
Field 1997 Own program CG 90.0 (total sample) 90.0 (total sample) 100 % SCZ 100 % SCZ NA NA NA NA NA NA NA 6 0 no 3
Fisher 2009 PositScience CG 69.0 76.9 100% SCZ 100% SCZ NA NA NA NA 407.93 469.23 NA 47.9 0 no 10
Fisher 2014 PositScience CG 72.1 76.7 NA NA NA NA 1.57 1.68 235.47 256.36 NA 34.65 0 no 8
Garrido 2013 Own program (based on Gexpert and copy programs) Watching Videos NA NA 100 % SCZ 100 % SCZ NA NA 11.84 10.68 307.64 326.99 Strategy provision 48 NA yes 26
Gomar 2015 FesKits CG 67.4 63.6 100 % SCZ 100 % SCZ NA NA 24.3 22.58 557.21 667.14 NA 27.52 0 no 26
Gomar 2015 FesKits TAU 67.4 74.4 100 % SCZ 100 % SCZ NA NA 24.3 23.38 557.21 675.92 NA 27.52 0 no 26
Greig 2007 NET Employment Program 57.6 48.3 70 % SCZ; 30 % SZA 62 % SCZ; 38 % SZA 45 % Afroamerican; 52 % Caucasian; 3 % Hispanic NA NA NA NA Employment Program 125.93 52.2 yes 52
Hermanutz 1991 Attention task Cognitive Group Sessions NA NA 100 % SCZ 100 % SCZ NA NA NA NA NA NA NA NA 0 no 4
Hermanutz 1991 Attention task TAU NA NA 100 % SCZ 100 % SCZ NA NA NA NA NA NA NA NA 0 no 4
Redoblado-Hodge 2010 NEAR TAU 60.0 (total sample) 60.0 (total sample) 100% SCZ 100% SCZ NA NA NA NA 649.89 (total sample) 649.89 (total sample) Strategy provision 30 NA yes 15
Hogarty 2004 CET EST NA NA NA NA NA NA NA NA NA NA Social-cognitive groups 75 84 yes 104
Hooker 2012 PositScience + Computerized Social Cognition Training (SETT/MindReading) CG 90.9 72.7 54.5 % SCZ; 45.5 % SZA 63.6 % SCZ; 26.4 % SZA NA NA 28.0 20.6 252.5 371.4 NA 47.27 0 no 10
Horan 2011 PositScience Standard illness management skills training 89.5 78.9 70.6% SCZ; 19.1% SZA; 10.3% psychosis NOS (total sample) 70.6% SCZ; 19.1% SZA; 10.3% psychosis NOS (total sample) 21.1% White; 5.3% Hispanic; 10.5% Asian; 57.9% Black; 5.3% Other 36.8% White; 26.3% Hispanic; 36.8% Black NA NA NA NA NA 19.3 0 no 12
Horan 2011 PositScience Standard illness management skills training 92.9 78.9 70.6% SCZ; 19.1% SZA; 10.3% psychosis NOS (total sample) 70.6% SCZ; 19.1% SZA; 10.3% psychosis NOS (total sample) 35.7% White; 21.4% Hispanic; 42.9% Black 36.8% White; 26.3% Hispanic; 36.8% Black NA NA NA NA Social Cognitive Skills Training 12 12 yes 12
Hubacher 2013 BrainStim TAU 60.0 42.9 100 % SCZ 100 % SCZ NA NA 6.27 11.83 NA NA NA 11.80 0 no 4
Iwata 2017 NEAR TAU 24.1 25.8 100 % SCZ 100% SCZ NA NA 11.74 12.04 672.6 674 Strategy provision 24 12 yes 12
Jahshan 2019 PositScience CG 82.1 70.0 86.9% SCZ; 13.1% SZA (total sample) 86.9% SCZ; 13.1% SZA (total sample) NA NA 32.60 29.44 NA NA NA 30.5 0 no 12
Jahshan 2019 CogPack CG 77.5 70.0 86.9% SCZ; 13.1% SZA (total sample) 86.9% SCZ; 13.1% SZA (total sample) NA NA 26.95 29.44 NA NA NA 30.5 0 no 12
Kantrowitz 2016 PositScience CG 60.7 68.8 NA NA NA NA NA NA NA NA Strategy provision 30 7.5 yes 20
Keefe 2012 PositScience CG NA NA 100% SCZ 100% SCZ NA NA NA NA NA NA Strategy provision 40 5 yes 12
Klingberg 2012 CogPack CBT 58.6 53.5 100 % SCZ 100 % SCZ NA NA NA NA 525 561 NA 10.85 0 yes 36
Kukla 2018 PositScience CBT 96.0 88.0 76 % SCZ; 24 % SZA 64% SCZ; 36 % SZA 56 % African American; 40 % White; 4 % Hispanic American 56% African American; 44 % White NA NA NA NA CBT 26 26 yes 26
Kurtz 2007 CogRehab Computer Skills Training 65.7 73.7 NA NA NA NA 11.0 9.8 NA NA NA 67.4 0 yes 52
Kurtz 2015 Cog Rem Computer Skills Training + Social Skills Training 73.1 73.3 NA NA NA NA 12.8 12.4 NA NA Social Skills Training 23.14 38.3 yes 24
Lalova 2013 RECOS REMAu 52.4 55.0 100 % SCZ 100 % SCZ NA NA 6.8 5.9 NA NA NA 12 0 yes 12
Lalova 2013 RECOS MBCT 52.4 63.6 100 % SCZ 100 % SCZ NA NA 6.8 6.9 NA NA NA 12 0 yes 12
Lee 2013 NEAR TAU 53.3 56.7 100 % SCZ 100 % SCZ NA NA 17.75 17.53 316.58 317.08 NA 20 0 yes 12
Lindenmayer 2008 CogPack Computer Session 91.1 87.5 82.2% SCZ or SZA; 17.8% Other 85.0% SCZ or SZA; 15% Other 13.3% White; 57.8% Black; 26.7% Hispanic; 2.2% Asian 12.5% White; 57.5% Black, 27.5% Hispanic; 2.5% Asian NA NA NA NA Strategy provision 24 12 yes 12
Linke 2017 CogPack Jacobson progressive relaxation 60.6 66.7 100% SCZ 100% SCZ NA NA NA NA 412.4 485.8 NA 28.8 0 yes 6
Man 2012 CAEL TAU 55.6 70.0 100% SCZ 100% SCZ NA NA NA NA NA NA NA NA NA no 4
McGurk 2005 CogPack Employment Program 54.5 (total sample) 54.5 (total sample) 72.7 % SCZ; 4.5% SZA; 22.7% Mood Disorder (whole sample) 72.7 % SCZ; 4.5% SZA; 22.7% Mood Disorder (whole sample) 68.2% African American; 15.9 Hispanic; 13.6% Caucasian; 2.3% Asian (whole sample) 68.2% African American; 15.9 Hispanic; 13.6% Caucasian; 2.3% Asian (whole sample) NA NA NA NA Employment Program 24 NA no 13.6
Medalia 2000 ORM (Memory) TAU 66.7 44.4 77.8 % SCZ; 22.2 % SZA 77.8 % SCZ; 22.2 % SZA NA NA NA NA NA NA NA 4.17 0 yes 5
Medalia 2000 ORM (Problem-solving) TAU 66.7 44.4 77.8 % SCZ; 22.2 % SZA 77.8 % SCZ; 22.2 % SZA NA NA NA NA NA NA NA 4.17 0 yes 5
Medalia 1998 Memory or problem solving tasks TAU 77.8 81.5 100% SCZ 100% SCZ NA NA NA NA NA NA NA 6 0 yes 6
Moritz 2011 CogPack Metacognitive Group Training 70.8 58.3 NA NA NA NA NA NA NA NA NA 8 0 no 4
Moritz 2015 Mybraintraining Professional TAU 36.7 33.3 100% SCZ 100% SCZ NA NA NA NA NA NA NA NA 0 no 6
Popova 2014 PositScience TAU 63.2 78.9 100 % SCZ 100 % SCZ NA NA NA NA 544.6 646.2 NA 20 0 no 4
Popova 2014 FAT TAU 57.9 78.9 100 % SCZ 100 % SCZ NA NA NA NA 617.2 646.2 NA 20 0 no 4
Ramsay 2017 Capitan’s Log Software Computer skills training NA NA NA NA NA NA 20.93 18.5 551.8 320.75 Strategy provision 48 8 yes 16
Rass 2012 PositScience TV-watching 58.8 64.7 47.1% SCZ, 52.9% SZA 35.3% SCZ, 64.7% SZA NA NA 17.9 22.6 NA NA NA 40 0 no 10
Rass 2012 PositScience TAU 58.8 90.0 47.1% SCZ, 52.9% SZA 80% SCZ, 20% SZA NA NA 17.9 19.9 NA NA NA 40 0 no 10
Reeder 2017 CIRCuiTS TAU 69.6 59.6 NA NA 28.3% White, 54.3% Black, 4.3% Asian, 13.0% Mixed Race 21.3% White, 61.7% Black, 8.5% Asian, 8.5% Mixed Race NA NA NA NA NA 40 0 yes 12
Royer 2012 REHACOM TAU NA NA 100 % SCZ 100 % SCZ NA NA 10.6 11.8 NA NA Paper-and-pencil exercises with strategy provision 24 120 yes 26
Russell 2008 METT Repeated exposure 65.4 71.4 NA NA NA NA NA NA 362 288.93 NA 1 0 no 1
Sachs 2012 TAR TAU 60.0 44.4 100 % SCZ 100 % SCZ NA NA NA NA NA NA NA 12 0 no 6
Sartory 2005 CogPack TAU 66.7 66.7 100 % SCZ 100 % SCZ NA NA 5.5 6.8 550.92 584.8 NA 11.25 0 no 3
Sato 2014 CogPack Supported Employment NA NA NA NA NA NA 9.71 11.66 647.56 449.51 Supported Employment 24 NA yes 12
Subramaniam 2014 PositScience CG 75.0 71.4 100 % SCZ 100 % SCZ NA NA NA NA 478 410 NA 80 0 no 16
Tan 2013 NET/CogRehab Physical Exercise 58.3 55.9 100 % SCZ 91.2 % SCZ, 8.8 % SZA 77.8% Chinese, 11.1% Malay, 8.3% Indian, 2.8% Other 82.4% Chinese, 14.7% Malay, 2.9% Indian, 0% Other 9.28 11.96 NA NA Strategy provision 60 NA yes 12
Thomas 2018 PositScience TAU 54.17 40.91 NA NA 17% Hispanic speaking, 21% African-American, 4% Asian, 54% Caucasian, More than one Race 12%, Native American 8% 27% Hispanic speaking, 14% African American, 9% Asian, 12% Caucasian, 23% More than one Race, 0% Native American 16.12 15.23 1329.42 982.54 NA 27.94 0 no 12.5
Trapp 2013 X-Cog Occupational Therapy 50.0 50.0 100 % SCZ 100 % SCZ NA NA 7.93 8.97 NA NA NA 12 0 no 3
Vauth 2005 Computer-assisted cognitive strategy training Vocational Rehabilitation 61.7 60.9 100 % SCZ 100 % SCZ NA NA 5.8 7.1 NA NA Vocational Rehabilitation 24 120 yes 8
Vázquez-Campo 2016 e-Motional Training TAU 70.0 55.6 100% SCZ 100% SCZ NA NA NA NA 463.8 380.56 NA 12 0 no 12
Vita 2011 CogPack REHAB 73.1 70.0 100 % SCZ 100 % SCZ NA NA 14.94 14.8 674.08 600.17 NA 31.50 0 no 24
Wölwer 2005 CR TAU 58.3 84.0 100 % SCZ 100 % SCZ NA NA NA NA NA NA Strategy provision 4.5 4.5 yes 6
Wölwer 2005 TAR TAU 89.3 84.0 100 % SCZ 100 % SCZ NA NA NA NA NA NA Strategy provision 4.5 4.5 yes 6
Open in a new tab

Table 2.

Demographic and training characteristics of the samples included in the meta-analysis, and the samples within the subgroups (exposure to a SHG yes or no).

Total sample k = 78 n = 67 Exposure to therapist + k = 43 n = 40 Exposure to therapist − k = 35 n = 28
Demographic variables M SD M SD M SD
N 4067 - 2572 - 1495 -
Sex (% male) 65.0 14.5 64.1 14.4 66.2 14.7
Age (years) 37.4 9.0 36.1 8.3 39.6 10.2
Premorbid IQ 98.3 13.5 97.5 14.1 99.7 12.2
Education (years) 12.1 2.3 12.2 2.4 11.9 2.3
CPZE 496.4 388.7 475.8 403.1 527.0 369.5
Illness duration (years) 13.5 7.5 10.2 6.1 17.1 9.1
Training variables
Computerized training duration (h) 27.2 23.5 30.1 27.0 23.2 17.4
Supplement training duration (h) 9.0 24.6 16.9 31.7 0 0
Total training duration (h) 36.4 40.3 46.5 48.1 21.7 17.4
Total training duration in weeks (n) 15.2 18.0 20.0 22.8 9.6 6.3
Therapist exposure (% yes) 55.1 - 100 - 0 -
Open in a new tab

Abbreviations: CPZE = Chlorpromazine Equivalents, k = number of samples, n = number of studies.

Note: One study (Horan, 2011) reported data on two samples, one with exposure to a therapist and one without.

The rationale to compare the post-intervention effects was to keep the current meta-analysis comparable to previously published meta-analysis as well as to assess the effects of the training in a securely controlled fashion (31).

All studies were required to report at least one of the three outcome measures following the intervention: cognitive performance, clinical symptoms, psychosocial functioning. Cognitive performance was analyzed according to the domains defined in the Measurement and Treatment Research to Improve Cognition in Schizophrenia (MATRICS) guidelines (32): attention/vigilance, reasoning and problem-solving, verbal learning/memory, visual learning/memory, working memory, processing speed or social cognition. For the tests that are not obtained in the MATRICS battery the decisions were based on the factor analysis reported in the MATRICS selection procedures or in the original literature introducing the test instrument. In cases where more than one cognitive test was used to evaluate one cognitive domain, we averaged the results from the tests. In the subdomains of psychosocial functioning we included measures of global functioning, work-related functioning, social functioning, health-related functioning and quality of life. Clinical symptoms included measures of positive symptoms, negative symptoms, total symptoms, general symptoms and measures of depression and anxiety. The detailed categorization of the tests for all the domains is presented in the Table 1 in the supplementary materials.

The main outcome measure for our meta-analysis was the standardized mean difference (Hedges’ g) describing differences after the intervention between SZ patients receiving computerized CR and SZ patients receiving a placebo intervention. In assessing posttreatment scores rather than change scores, we took a conservative approach (31). If the included studies did not report any mean values and standard deviations of outcome measures, Hedges’ g was calculated from Cohen’s d (33,34). In case there was only a group effect given, we calculated Cohen’s d from the available F-value and transformed the effect size to Hedges’ g.

Data Analysis:

For every study included in our analysis, Hedges’ g was calculated and entered into a random-effects meta-analytic model (35,36). Summary effect sizes were computed separately for the three outcome dimensions (cognitive performance, clinical symptoms, psychosocial functioning) using a restricted maximum-likelihood estimator (37). When a study reported data from multiple measures classified in the same outcome domain, the mean of the effect sizes from those measures was used (11). Additionally, random-effects models were fit for the subdomains of each outcome domain. Heterogeneity was assessed via the I2 value which describes the percentage of total variation across studies that is due to heterogeneity rather than chance (38). I2 values of 25%, 50%, and 75% can be interpreted as indicating low, moderate and high heterogeneity respectively (38). A significance level of P < 0.05 (2-tailed) was considered for all analyses. Potential moderating effects were tested using meta-regression for the following clinical and demographic variables: age, gender, antipsychotic medication (chlorpromazine equivalent, CPZ), IQ, illness duration, duration of the computerized CR intervention and duration of supplementary training as well as the time spread of the intervention. To focus on clinically relevant moderator effects, we only report significant effects for ßs with an absolute value larger than 0.01.

Additionally, we conducted a subgroup analysis using a fixed-effects model (39) in which we compared those studies in which the experimental condition featured exposure to the therapist to those where it did not. As recommended, subgroup-analysis was only conducted when more than 10 studies were available for a given outcome (40). Publication bias was assessed by Egger’s test for funnel plot asymmetry (41) for every separate meta-analysis. Egger’s test was not computed when less than 10 studies were available for a given outcome domain. As a sensitivity analysis, trim and fill was used to impute putatively missing studies with the R0-estimator and to compute a corrected effect size including the missing studies. Outlier and influence diagnostics were performed following (42). When influential studies were detected, we re-ran the analysis to recalculate the effect size with that study excluded. All statistical analyses were conducted using the R statistical programming language version 3.6.1 with the packages ‘metafor’ and ‘meta’ (43).

Structural equation modeling (SEM)

In an exploratory analysis, we aimed to determine the directional relationships between improvement in the different outcome domains (cognitive, clinical, and functioning). To this end, we used subject-level, behavioral data from patients with schizophrenia who underwent 50 hours of CR (N = 46; mean age = 40.7 (SD = 11.8), 73.9% male) and a control group who played computer games for the same number of hours (n = 41, mean age = 43.2 years (SD = 10.6), 70.7% male) to build SEMs. Both groups were drawn from the randomized clinical trial of cognitive-training (ClinicalTrials.gov NCT00694889).

In this study data set, we defined three outcomes to map to the cognitive, clinical and functional outcome domains in the main meta-analysis. These included cognitive functioning, measured with the MATRICS global score (32), clinical symptomatology as evaluated with Positive and Negative Symptoms Scale) (44 (PANSS) total scores, and functional outcome assessed with the average of social and work subscales of the quality of life scale (45) (QLS). In keeping with the post-treatment comparison design used in the meta-analysis, we z-standardized the post-treatment scores of each subject in the experimental group to the respective scores of the control group.

We built three different SEMs, all based on the assumption that cognitive training would initially improve cognition: (1) model 1 encoded separate directed paths from cognitive outcomes to clinical and functional outcomes (Fig.7A) (2) model 2 encoded a directed path from cognitive to clinical outcomes via functional outcomes (Fig.7B), and (3) in model 3, a directed path from cognitive to functional outcomes via clinical outcomes was encoded (Fig. 7C). To determine the best model fit to the data set, we evaluated the Bayesian Information Criterion (BIC) scores of the models. We then used the BIC to approximate a Bayesian hypothesis test to provide an estimate of the posterior probability of each model against the other models (46,47). We ran all SEM analyses with the R package ‘lavaan’ (48).

Fig.7.

Fig.7.

Open in a new tab

Structured equation models illustrating posterior probabilities for model 1(a), model 2 (b) and model 3 (c)

Results

Literature Search

Our literature search identified 690 potential studies. After study selection using the defined exclusion criteria, k = 67 studies reporting data on 78 samples, with n = 4067 participants (65.0% male, mean age = 37.4 (SD = 9.0), table 1) were included in the meta-analysis. Nine studies reported data on several samples that were included separately.

Detailed results from the meta-analysis are presented in table 3 and figure 2.

Table 3.

Overview of the results from the meta-analysis.

Outcome k g 95% CI: lb 95% CI: ub p-value I2 (%) Egger’s test: z-value Egger’s test: p-value
Cognitive 70 0.28 0.21 0.34 < .001 3.52 0.77 .440
Attention 34 0.23 0.14 0.33 < .001 0.01 1.29 .198
Verbal Memory 50 0.21 0.09 0.33 < .001 59.87 −0.90 .370
Visual Memory 24 0.12 −0.03 0.26 .108 43.62 −1.18 .236
Social Cognition 23 0.18 0.06 0.3 .002 7.38 0.48 .629
Working Memory 47 0.27 0.18 0.35 < .001 10.07 0.19 .847
Processing Speed 43 0.22 0.12 0.32 < .001 32.13 −0.24 .807
Reasoning 38 0.25 0.14 0.37 < .001 44.84 1.56 .120
Cognitive Global 30 0.26 0.15 0.36 < .001 19.16 −1.91 .055
Clinical 40 0.10 0.01 0.19 .026 13.81 1.21 .228
Positive Symptoms 30 0.06 −0.07 0.19 .343 38.48 −0.60 .550
Negative Symptoms 32 0.13 0.02 0.25 .026 31.92 0.81 .420
General Symptoms 17 0.05 −0.07 0.18 .393 0.53 1.48 .138
Total Symptoms 19 0.10 −0.07 0.28 .241 48.66 −0.25 .802
Depression/Anxiety 14 0.25 0.09 0.4 .002 13.72 2.13 .033
Clinical Global 9 0.22 −0.02 0.45 .067 52.01 0.88 .378
Functioning 49 0.16 0.06 0.25 < .001 33.58 −0.19 .851
Social 21 0.26 0.13 0.38 < .001 12.98 1.79 .074
Health 8 0.10 −0.25 0.45 .580 63.46 - -
Work 14 0.40 0.16 0.64 < .001 67.95 2.53 .011
Quality of Life 9 0.14 −0.04 0.33 .128 4.63 - -
Functioning Global 23 0.04 −0.10 0.18 .577 35.58 −0.51 .609
Open in a new tab

Abbreviations: CI = confidence interval; g = Hedges’ g; k = number of studies; lb = lower bound; ub = upper bound.

Fig. 2.

Fig. 2.

Open in a new tab

Overview of the different effects of computerized CR in SZ patients overall (A), on cognitive performance (B), clinical symptoms (C) and psychosocial functioning (D) including effects of subdomains of these outcomes. Error bars represent 95% confidence intervals.

Cognitive outcomes

There was a moderate effect of CR on cognitive outcomes (g = 0.28, figure 2A). Within the cognitive domains (figure 2B), there were significant effects of CR on attention (g = 0.23), verbal memory (g = 0.21), social cognition (g = 0.18), working memory (g = 0.27), processing speed (g = 0.22), reasoning (g = 0.25), and cognitive global (g = 0.26). No significant effect was present for visual memory.

Clinical outcomes

Overall, we found a small, significant effect of CR on clinical outcomes (g = 0.10, figure 1A). Within the clinical subdomains (figure 2C), CR yielded significant effects on negative (g = 0.13) and depressive/anxious symptoms (g = 0.25). No significant effects of CR were found on positive, general, and total symptoms, nor on clinical global.

Functional outcomes

Globally, there was a small, significant effect of CR on functional outcomes (g = 0.16, figure 2A). Within functional subdomains (figure 2D), CR showed significant effects on social (g = 0.26) and work outcomes (g = 0.40), but not on health, quality of life, and functioning global.

Publication bias

Cognitive outcomes

There was evidence for temporally decreasing effect sizes in the outcome domain processing speed, as indicated by a significant moderation effect of year on effect sizes (k = 43, ß = −0.021, z = −2.63, p = .008). Earlier studies publishing effects of the CR on processing speed reported larger effects.

Clinical outcomes

Egger’s test revealed funnel plot asymmetry for depression and anxiety (supplementary Fig.1. for funnel plot). Trim-fill indicated one missing study on the left side of the funnel plot, with a corrected effect size of g = 0.22.

Functional outcomes

Egger’s test indicated funnel plot asymmetry for the outcome work. Trim-fill analysis indicated seven missing studies on the right side of the funnel plot (supplementary Fig. 2.), with a corrected effect of g = 0.30.

A detailed description of the outlier studies can be found in the supplement.

Moderator analysis

Regarding cognitive outcomes, moderator analyses indicated negative effects of the duration of cognitive training on effectiveness of CR in the domain attention (k = 31, ß = −0.01, z = −2.59, p = .010). Effectiveness of CR on reasoning and problem-solving skills was higher in younger samples (k = 35, ß = −0.02, z = −3.27, p = .004), samples with less males (k = 27, ß = −1.42, z = −3.27, p = .001), and samples with shorter illness duration (k = 19, ß = −0.02, z = −2.51, p = .011). Similarly, effectiveness of CR on scores on cognitive global was increased in samples with shorter illness duration (k = 12, ß = −0.02, z = −2.20, p = .028). Further results on moderator analysis related to non-significant main outcomes can be found in the supplementary materials.

Subgroup analysis: CR vs. CR including SHG

CR was significantly more effective when provided including SHG (Fig.3) in improving cognitive outcomes of verbal memory (Q = 3.38, p = .050), working memory (Q = 4.35, p < .037), and real-world cognitive skills (Q = 6.63, p = 0.01). No further evidence for increased effectiveness of CR provided in conjunction with SHG was found in other cognitive domains (all ps > .772), clinical outcome domains (all ps > .644) or functioning domains (all ps > .551).

Fig.3.

Fig.3.

Open in a new tab

Overview of the different effects of computerized CR in SZ patients overall (A) on cognitive performance (B), clinical symptoms (C) and psychosocial functioning (D) including effects of subdomains of these outcomes, grouped by exposure to SHG (dark shade) and no exposure to SHG (light shade). Subgroup analysis was only conducted for meta-analysis with more than 10 studies. Error bars represent 95% confidence intervals.

The detailed comparison of studies using computerized CR with SHG vs. no SHG for each outcome and sub-outcomes can be found in Table 2 in the supplementary materials.

Forest plots for main outcomes (cognitive, functional and clinical) are presented in the Figures 4–6; additional forest plots for the subdomains of each main outcome domain are presented in the supplementary materials of the manuscript.

Fig.4.

Fig.4.

Open in a new tab

Meta-analysis of the effect of computerized CR on cognitive performance in SZ patients compared to active placebo condition.

Fig. 6:

Fig. 6:

Open in a new tab

Meta-analysis of the effect of computerized CR on psychosocial functioning in SZ patients compared to active placebo condition.

Structural Equation Modeling (SEM)

In the exploratory SEM analysis, BIC scores indicated a preference for model 2 (BIC = 284.4) over model 1 (BIC = 289.0) and model 3 (BIC = 287.9). Approximated posterior probabilities of the models reflected this: compared to the other models, model 2 was most likely (posterior probability (pp) = .786) given the available data. Following the classification by Rafterty (49), this result can be considered positive evidence for model 2. Weak support was found for model 1 (pp = .136) and model 2 (pp = .079) compared to the respective other two (Fig.7).

Hence, given the available data, a directed effect from cognition to psychosocial functioning with a subsequent effect from psychosocial functioning to clinical symptoms is most likely.

Discussion

We conducted a comprehensive MOMA of the effects of computerized CR in SZ patients with and without SHG to provide a synthesis of the currently available evidence and to assess the moderating effects on multiple outcomes. Overall, our results indicate that CR is associated with small to moderate improvements in cognitive performance (g = 0.28), clinical outcomes (g = 0.10) and psychosocial functioning (g = 0.16) on the basis of data from k = 67 studies with a total number of n = 4053 participants. Most of these effects were robust with respect to potential moderator effects, but showed some evidence for a publication bias in subdomains of processing speed as well as general and depressive symptoms.

Effects of CR on different outcomes

Our findings replicate other reviews’ and meta-analyses’ notion that CR - also when administered mostly in a computerized fashion - provides small to moderate benefits (21,22,50) in multiple cognitive domains including attention, working memory, verbal memory, reasoning, processing speed and social cognition (50). Different cognitive domains showed different degrees of susceptibility to moderating effects. For instance, samples that included younger SZ patients with less males and shorter illness duration appeared to respond more effectively to CR by increasing performance in reasoning. Importantly, this claim is limited to the group level and individual odds for improvement in response to CR cannot be inferred.

All previous meta-analyses have delivered more heterogeneous results for clinical as compared to cognitive outcomes. Although the observed effectiveness of CR in reducing clinical symptoms was small, it rendered effective in reducing negative symptoms, depression and anxiety. Previous meta-analyses found a medium effect of CR on total symptoms but no significant effect on negative and positive symptoms (50,51) whereas another meta-analysis found positive effects on positive symptoms and depressive symptoms (22).

Further, we found robust evidence for the effects of computerized CR in improving social and particularly occupational functioning in SZ patients. We found no effect of computerized CR on health, quality of life or global psychosocial functioning. The relevance and comparability of these findings remains questionable as previous meta-analyses did not differentiate between these types of functional outcomes. All previous meta-analyses found a small effect of CR on improving everyday functioning in general (21,52). However, the last meta-analysis on computerized CR (22) included substantially less studies than the present one and included only computerized CR without SHG.

CR with SHG vs. without SHG

Current meta-analysis obtains forty CR studies that included therapeutic coaching in addition to computerized CR and provides—to the best of our knowledge—the first comprehensive study that compared computerized CR versus SHG enriched computerized CR. While the vast majority of cognitive functions benefit from computerized CR two cognitive subdomains including verbal memory, and working memory seem to experience larger gains when patients are exposed to SHG. Abstract reasoning shows to be another promising candidate to benefit from SHG but yielded no significant result in the present analysis.

At the same time attention and processing speed render significant improvement with and without SHG which may be connected to a repetitive ‘drill and practice’ character of computerized CR (22) which is based on a particular learning practice that boosts cognitive function independent of the therapist’s presence. On the other hand, verbal and working memory gain larger benefit when computerized CR is combined with SHG. Supplementary human guidance may provide insight into meta-learning and highlights relevant strategies used, relevant for higher cognitive functions (27,53). Also, memory tasks may have a stronger ecological link to everyday life and translate more easily into daily routine tasks that can be practiced through SHG.

Finally, it has been previously suggested that SHG may work in a synergistic manner with computerized CR (21) by enhancing functional recovery which was not confirmed by our analysis.

Although some improvements in cognitive domains appear overall to be facilitated when human guidance is provided, clinical and functional outcomes show no differential response to this intervention modality as opposed to computerized CR. It is noteworthy, that for social and occupational functioning not enough studies were available to conduct meaningful subgroup analyses.

Nevertheless, this is an important point for practitioners to consider in implementation: while SHG complementing CR is not necessary in improving some cognitive outcomes like processing speed and attention, it will potentially provide better effects in training higher cognitive function like memory and abstract reasoning. Obviously, clinical implications considering financial costs of human support in the medical system in the era of digital therapies is becoming increasingly important. Though these findings need to be challenged in the future they potentially pave the path for optimization of human presence in the relevant outcome domains.

Finally, these finding should not be over-interpreted but rather stimulate ideas to enrich computerized CR programs to creatively tackle improvement in those cognitive domains who respond more effectively to joint intervention.

Structured Equation Modeling (SEM)

Current MOMA estimated the interplay between different aspects of recovery. Our current analysis suggests that a directed effect from cognition to psychosocial functioning with a subsequent effect from psychosocial functioning to clinical symptoms is most likely.

Previous studies have shown that improvement in cognitive ability through cognitive training leads to positive effects in ability to perform critical everyday living skills which leads to improvement in the general functional capacity (12). The observation of multiple individual studies (26,27) that psychosocial functioning gains are mediated by human guidance and clinical improvement does not occur directly through CR, but rather through functional improvement initiated by cognitive intervention seems confirmed in our dataset. Along the same lines previous meta-analysis (6) indicated that different neurocognitive functions are differentially related to different domains of functional outcome. Importantly, community functioning and social behavior were strongly associated with verbal memory and learning, but also visual learning and memory. Those cognitive domains have also been pinpointed in our subanalysis on efficacy of additional SHG. It could be further speculated that particularly those cognitive domains may be responsible for successfully translating cognitive gains into the daily routine. Model 3 suggesting alternative directionality that symptoms, and in particular, negative symptoms or positive subscales of PANSS, could mediate the relationship between neurocognition and functioning has shown to be less probable than Model 2. However, previous studies suggesting this directionality focused on improvements of e.g. theory of mind (29) or meta-cognition (54) which are higher cognitive functions that may be able to affect beliefs and symptoms more easily than the lower cognitive domains, we analyzed in the context of this study. On the same note, several studies that identified relative independence of neurocognition and symptoms (55,56) match these findings. Our SEM analysis is rather limited due to a single-study approach. Future studies using SEM and network approach could potentially explain this view in more detail using bigger patient cohorts and multiple assessment points.

CR in the landscape of current pharmacological and non-pharmacological interventions

The evidence provided by other additional non-pharmacological interventions gathered in the last few years demonstrates that CR administered in addition to pharmacological treatment as usual, delivers highest effects sizes in cognitive recovery. Physical exercise seems to deliver comparable effect sizes in some cognitive domains such as working memory (57), but to be overall less suitable for multiple facets of cognitive function. Other complementary treatments including ‘TMS’ ((58,59)), cognitive enhancers as nicotine and glycine (60) render less effective.

It is noteworthy that there are considerable differences in the effect sizes (Cohen’s d) delivered by atypical antipsychotic medication which range from 0.1 to 0.4 (9,10,52,53). While vigilance and selective attention seem to improve to a larger extent, higher cognitive domains such as cognitive flexibility and abstract reasoning mark less favorable outcomes. If we consider that cognitive impairment means up to two standard deviations in comparison to the general healthy population, it is highly unlikely that the gains from atypical antipsychotics will be sufficient to return SZ patients to the premorbid vocational level. Finally, CR provides patients with specific gains in cognitive domain which are not as clearly achievable with other therapeutic approaches.

Conclusion

The analysis of moderating effects, SHG and causal modeling are providing us with insight into potential mechanism of action of computerized CR. Though these approaches help us to initially break down the heterogeneity of response to this type of intervention, the limitation is that they do not provide us with a prediction at the individual patient level. Future studies could overcome this limitation by implementing machine learning approaches to elucidate the effectiveness of CR, similarly to recent treatment outcome prediction studies (61). Building statistical models by mining existing CR intervention data can enable prospective identification of SZ patients who are likely to respond to a specific form of CR and eventually personalize CR interventions.

Supplementary Material

Supplementary Materials

Fig.5.

Fig.5.

Open in a new tab

Meta-analysis of the effect of computerized CR on clinical symptoms in SZ patients compared to active placebo condition.

Acknowledgments

Lana Kambeitz-Ilankovic was supported by the EU-FP7 project PRONIA (“Personalised Prognostic Tools for Early Psychosis Management”) under the Grant Agreement No° 602152.

Footnotes

Conflict of interest

NK and JK are currently honorary speakers for Otsuka. SV is a site PI on an SBIR grant to Posit Science, a company with a commercial interest in the cognitive training software. SV serves on an advisory board for Forum pharmaceuticals. None of the other authors have any financial interest in Posit Science. All authors declare no other conflicts of interest.

References

  • 1.Reichenberg A, Harvey PD, Bowie CR, Mojtabai R, Rabinowitz J, Heaton RK, et al. Neuropsychological Function and Dysfunction in Schizophrenia and Psychotic Affective Disorders. Schizophr Bull. 2009. Sep 1;35(5):1022–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2.Albus M, Hubmann W, Mohr F, Hecht S, Hinterberger-Weber P, Seitz N-N, et al. Neurocognitive functioning in patients with first-episode schizophrenia : results of a prospective 5-year follow-up study. Eur Arch Psychiatry Clin Neurosci. 2006. Oct;256(7):442–51. [DOI] [PubMed] [Google Scholar]
  • 3.Green MF, Kern RS, Braff DL, Mintz J. Neurocognitive Deficits and Functional Outcome in Schizophrenia: Are We Measuring the “Right Stuff”? Schizophr Bull. 2000. Jan 1;26(1):119–36. [DOI] [PubMed] [Google Scholar]
  • 4.Lin A, Wood SJ, Nelson B, Brewer WJ, Spiliotacopoulos D, Bruxner A, et al. Neurocognitive predictors of functional outcome two to 13 years after identification as ultra-high risk for psychosis. Schizophr Res. 2011. Oct;132(1):1–7. [DOI] [PubMed] [Google Scholar]
  • 5.Velthorst E, Fett A-KJ, Reichenberg A, Perlman G, van Os J, Bromet EJ, et al. The 20-Year Longitudinal Trajectories of Social Functioning in Individuals With Psychotic Disorders. Am J Psychiatry. 2017. 01;174(11):1075–85. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Fett A-KJ, Viechtbauer W, Dominguez M-G, Penn DL, van Os J, Krabbendam L. The relationship between neurocognition and social cognition with functional outcomes in schizophrenia: a meta-analysis. Neurosci Biobehav Rev. 2011. Jan;35(3):573–88. [DOI] [PubMed] [Google Scholar]
  • 7.Tandberg M, Ueland T, Sundet K, Haahr U, Joa I, Johannessen JO, et al. Neurocognition and occupational functioning in patients with first-episode psychosis: a 2-year follow-up study. Psychiatry Res. 2011. Aug 15;188(3):334–42. [DOI] [PubMed] [Google Scholar]
  • 8.Nielsen RE, Levander S, Kjaersdam Telléus G, Jensen SOW, Østergaard Christensen T, Leucht S. Second-generation antipsychotic effect on cognition in patients with schizophrenia--a meta-analysis of randomized clinical trials. Acta Psychiatr Scand. 2015. Mar;131(3):185–96. [DOI] [PubMed] [Google Scholar]
  • 9.Mishara AL, Goldberg TE. A meta-analysis and critical review of the effects of conventional neuroleptic treatment on cognition in schizophrenia: opening a closed book. Biol Psychiatry. 2004. May 15;55(10):1013–22. [DOI] [PubMed] [Google Scholar]
  • 10.Goldberg TE, Goldman RS, Burdick KE, Malhotra AK, Lencz T, Patel RC, et al. Cognitive improvement after treatment with second-generation antipsychotic medications in first-episode schizophrenia: is it a practice effect? Arch Gen Psychiatry. 2007. Oct;64(10):1115–22. [DOI] [PubMed] [Google Scholar]
  • 11.McGurk SR, Mueser KT, DeRosa TJ, Wolfe R. Work, recovery, and comorbidity in schizophrenia: a randomized controlled trial of cognitive remediation. Schizophr Bull. 2009. Mar;35(2):319–35. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Barlati S, Deste G, De Peri L, Ariu C, Vita A. Cognitive Remediation in Schizophrenia: Current Status and Future Perspectives. Schizophr Res Treat. 2013. Dec 17;2013:e156084. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Medalia A, Choi J. Cognitive remediation in schizophrenia. Neuropsychol Rev. 2009. Sep;19(3):353–64. [DOI] [PubMed] [Google Scholar]
  • 14.Vinogradov S, Fisher M, de Villers-Sidani E. Cognitive training for impaired neural systems in neuropsychiatric illness. Neuropsychopharmacol Off Publ Am Coll Neuropsychopharmacol. 2012. Jan;37(1):43–76. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Fisher M, Holland C, Merzenich MM, Vinogradov S. Using neuroplasticity-based auditory training to improve verbal memory in schizophrenia. Am J Psychiatry. 2009. Jul;166(7):805–11. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.Fisher M, Herman A, Stephens DB, Vinogradov S. Neuroscience-informed computer-assisted cognitive training in schizophrenia. Ann N Y Acad Sci. 2016. Feb;1366(1):90–114. [DOI] [PubMed] [Google Scholar]
  • 17.Subramaniam K, Luks TL, Fisher M, Simpson GV, Nagarajan S, Vinogradov S. Computerized Cognitive Training Restores Neural Activity within the Reality Monitoring Network in Schizophrenia. Neuron. 2012. Feb 23;73(4):842–53. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Subramaniam K, Hooker CI, Biagianti B, Fisher M, Nagarajan S, Vinogradov S. Neural signal during immediate reward anticipation in schizophrenia: Relationship to real-world motivation and function. NeuroImage Clin. 2015. Jan 1;9:153–63. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Cella M, Preti A, Edwards C, Dow T, Wykes T. Cognitive remediation for negative symptoms of schizophrenia: A network meta-analysis. Clin Psychol Rev. 2017. Mar;52:43–51. [DOI] [PubMed] [Google Scholar]
  • 20.Reckless GE, Andreassen OA, Server A, Østefjells T, Jensen J. Negative symptoms in schizophrenia are associated with aberrant striato-cortical connectivity in a rewarded perceptual decision-making task. NeuroImage Clin. 2015. May 4;8:290–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.Wykes T, Huddy V, Cellard C, McGurk SR, Czobor P. A Meta-Analysis of Cognitive Remediation for Schizophrenia: Methodology and Effect Sizes. Am J Psychiatry. 2011. May;168(5):472–85. [DOI] [PubMed] [Google Scholar]
  • 22.Prikken M, Konings MJ, Lei WU, Begemann MJH, Sommer IEC. The efficacy of computerized cognitive drill and practice training for patients with a schizophrenia-spectrum disorder: A meta-analysis. Schizophr Res [Internet]. 2018. Aug; Available from: https://linkinghub.elsevier.com/retrieve/pii/S0920996418304791 [DOI] [PubMed] [Google Scholar]
  • 23.Fisher M, Holland C, Subramaniam K, Vinogradov S. Neuroplasticity-based cognitive training in schizophrenia: an interim report on the effects 6 months later. Schizophr Bull. 2010. Jul;36(4):869–79. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Fisher M, Loewy R, Carter C, Lee A, Ragland JD, Niendam T, et al. Neuroplasticity-Based Auditory Training Via Laptop Computer Improves Cognition in Young Individuals With Recent Onset Schizophrenia. Schizophr Bull. 2015. Jan 1;41(1):250–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Hooker CI, Carol EE, Eisenstein TJ, Yin H, Lincoln SH, Tully LM, et al. A pilot study of cognitive training in clinical high risk for psychosis: initial evidence of cognitive benefit. Schizophr Res. 2014. Aug;157(1–3):314–6. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26.Lindenmayer J-P, McGurk SR, Mueser KT, Khan A, Wance D, Hoffman L, et al. A randomized controlled trial of cognitive remediation among inpatients with persistent mental illness. Psychiatr Serv Wash DC. 2008. Mar;59(3):241–7. [DOI] [PubMed] [Google Scholar]
  • 27.Reeder C, Huddy V, Cella M, Taylor R, Greenwood K, Landau S, et al. A new generation computerised metacognitive cognitive remediation programme for schizophrenia (CIRCuiTS): a randomised controlled trial. Psychol Med. 2017. Nov;47(15):2720–30. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28.Meyer EC, Carrión RE, Cornblatt BA, Addington J, Cadenhead KS, Cannon TD, et al. The Relationship of Neurocognition and Negative Symptoms to Social and Role Functioning Over Time in Individuals at Clinical High Risk in the First Phase of the North American Prodrome Longitudinal Study. Schizophr Bull. 2014. Nov;40(6):1452–61. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.Ventura J, Ered A, Gretchen-Doorly D, Subotnik KL, Horan WP, Hellemann GS, et al. Theory of mind in the early course of schizophrenia: stability, symptom and neurocognitive correlates, and relationship with functioning. Psychol Med. 2015. Jul;45(10):2031–43. [DOI] [PubMed] [Google Scholar]
  • 30.Wykes T, Spaulding WD. Thinking About the Future Cognitive Remediation Therapy—What Works and Could We Do Better? Schizophr Bull. 2011. Sep 1;37(suppl_2):S80–90. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.Fu R, Holmer HK. Change score or follow-up score? Choice of mean difference estimates could impact meta-analysis conclusions. J Clin Epidemiol. 2016. Aug 1;76:108–17. [DOI] [PubMed] [Google Scholar]
  • 32.Nuechterlein KH, Barch DM, Gold JM, Goldberg TE, Green MF, Heaton RK. Identification of separable cognitive factors in schizophrenia. Schizophr Res. 2004. Dec 15;72(1):29–39. [DOI] [PubMed] [Google Scholar]
  • 33.Ikezawa S, Mogami T, Hayami Y, Sato I, Kato T, Kimura I, et al. The pilot study of a Neuropsychological Educational Approach to Cognitive Remediation for patients with schizophrenia in Japan. Psychiatry Res. 2012. Feb 28;195(3):107–10. [DOI] [PubMed] [Google Scholar]
  • 34.Kantrowitz JT, Sharif Z, Medalia A, Keefe RSE, Harvey P, Bruder G, et al. A Multicenter, Rater-Blinded, Randomized Controlled Study of Auditory Processing-Focused Cognitive Remediation Combined With Open-Label Lurasidone in Patients With Schizophrenia and Schizoaffective Disorder. J Clin Psychiatry. 2016;77(6):799–806. [DOI] [PubMed] [Google Scholar]
  • 35.idostatistics. Hedges LV, & Olkin I (1985). Statistical methods for meta-analysis [Internet]. IDoStatistics. 2016. [cited 2018 Dec 4]. Available from: https://idostatistics.com/hedges-olkin-1985-statistical-methods-for-meta-analysis/ [Google Scholar]
  • 36.Hedges LV, Vevea JL. Fixed- and random-effects models in meta-analysis. Psychol Methods. 1998;3(4):486–504. [Google Scholar]
  • 37.Raudenbush SW. Analyzing effect sizes: Random-effects models. In: The handbook of research synthesis and meta-analysis, 2nd ed. New York, NY, US: Russell Sage Foundation; 2009. p. 295–315. [Google Scholar]
  • 38.Higgins JPT, Thompson SG, Deeks JJ, Altman DG. Measuring inconsistency in meta-analyses. BMJ. 2003. Sep 6;327(7414):557–60. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39.Borenstein M, Higgins JPT. Meta-Analysis and Subgroups. Prev Sci. 2013. Apr 1;14(2):134–43. [DOI] [PubMed] [Google Scholar]
  • 40.Higgins JPT, Thompson SG. Controlling the risk of spurious findings from meta-regression. Stat Med. 2004. Jun 15;23(11):1663–82. [DOI] [PubMed] [Google Scholar]
  • 41.Egger M, Davey Smith G, Schneider M, Minder C. Bias in meta-analysis detected by a simple, graphical test. BMJ. 1997. Sep 13;315(7109):629–34. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 42.Viechtbauer W, Cheung MW-L. Outlier and influence diagnostics for meta-analysis. Res Synth Methods. 2010;1(2):112–25. [DOI] [PubMed] [Google Scholar]
  • 43.Viechtbauer W Conducting Meta-Analyses in R with the metafor Package. J Stat Softw. 2010. Aug 5;36(1):1–48. [Google Scholar]
  • 44.Kay SR, Flszbein A, Opfer LA. The Positive and Negative Syndrome Scale (PANSS) for Schizophrenia. Schizophr Bull. 1987;13(2):261–76. [DOI] [PubMed] [Google Scholar]
  • 45.Burckhardt CS, Anderson KL. The Quality of Life Scale (QOLS): Reliability, Validity, and Utilization. Health Qual Life Outcomes. 2003. Oct 23;1:60. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 46.Glover S, Dixon P. Likelihood ratios: a simple and flexible statistic for empirical psychologists. Psychon Bull Rev. 2004. Oct;11(5):791–806. [DOI] [PubMed] [Google Scholar]
  • 47.Wagenmakers E-J, Marsman M, Jamil T, Ly A, Verhagen J, Love J, et al. Bayesian inference for psychology. Part I: Theoretical advantages and practical ramifications. Psychon Bull Rev. 2018. Feb 1;25(1):35–57. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 48.Rosseel Y lavaan: An R Package for Structural Equation Modeling. J Stat Softw. 2012. May 24;48(1):1–36. [Google Scholar]
  • 49.Raftery A Bayesian Model Selection in Social Research (with Discussion by Andrew Gelman & Donald B. Rubin, and Robert M. Hauser, and a Rejoinder) 1995.
  • 50.Grynszpan O, Perbal S, Pelissolo A, Fossati P, Jouvent R, Dubal S, et al. Efficacy and specificity of computer-assisted cognitive remediation in schizophrenia: a meta-analytical study. Psychol Med. 2011. Jan;41(1):163–73. [DOI] [PubMed] [Google Scholar]
  • 51.Kurtz MM, Richardson CL. Social cognitive training for schizophrenia: a meta-analytic investigation of controlled research. Schizophr Bull. 2012. Sep;38(5):1092–104. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 52.McGurk SR, Mueser KT, Feldman K, Wolfe R, Pascaris A. Cognitive training for supported employment: 2–3 year outcomes of a randomized controlled trial. Am J Psychiatry. 2007. Mar;164(3):437–41. [DOI] [PubMed] [Google Scholar]
  • 53.Wykes T, Reeder C, Corner J, Williams C, Everitt B. The effects of neurocognitive remediation on executive processing in patients with schizophrenia. Schizophr Bull. 1999;25(2):291–307. [DOI] [PubMed] [Google Scholar]
  • 54.Eichner C, Berna F. Acceptance and Efficacy of Metacognitive Training (MCT) on Positive Symptoms and Delusions in Patients With Schizophrenia: A Meta-analysis Taking Into Account Important Moderators. Schizophr Bull. 2016. Jul;42(4):952–62. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 55.Dominguez M de G, Viechtbauer W, Simons CJP, van Os J, Krabbendam L. Are psychotic psychopathology and neurocognition orthogonal? A systematic review of their associations. Psychol Bull. 2009. Jan;135(1):157–71. [DOI] [PubMed] [Google Scholar]
  • 56.Harvey PD, Koren D, Reichenberg A, Bowie CR. Negative symptoms and cognitive deficits: what is the nature of their relationship? Schizophr Bull. 2006. Apr;32(2):250–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 57.Firth J, Stubbs B, Rosenbaum S, Vancampfort D, Malchow B, Schuch F, et al. Aerobic Exercise Improves Cognitive Functioning in People With Schizophrenia: A Systematic Review and Meta-Analysis. Schizophr Bull. 2017. May 1;43(3):546–56. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 58.Hasan A, Strube W, Palm U, Wobrock T. Repetitive Noninvasive Brain Stimulation to Modulate Cognitive Functions in Schizophrenia: A Systematic Review of Primary and Secondary Outcomes. Schizophr Bull. 2016. Jul;42(Suppl 1):S95–109. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 59.Hasan A, Guse B, Cordes J, Wölwer W, Winterer G, Gaebel W, et al. Cognitive Effects of High-Frequency rTMS in Schizophrenia Patients With Predominant Negative Symptoms: Results From a Multicenter Randomized Sham-Controlled Trial. Schizophr Bull. 2016. May;42(3):608–18. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 60.Sinkeviciute I, Begemann M, Prikken M, Oranje B, Johnsen E, Lei WU, et al. Efficacy of different types of cognitive enhancers for patients with schizophrenia: a meta-analysis. NPJ Schizophr [Internet]. 2018. Oct 25;4. Available from: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6202388/ [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 61.Chekroud AM, Zotti RJ, Shehzad Z, Gueorguieva R, Johnson MK, Trivedi MH, et al. Cross-trial prediction of treatment outcome in depression: a machine learning approach. Lancet Psychiatry. 2016. Mar 1;3(3):243–50. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Materials
NIHMS1762981-supplement-Supplementary_Materials.docx (523.7KB, docx)

RESOURCES