University of Groningen Neurophysiological signature(s) of visual hallucinations across neurological and perceptual Dauwan, Meenakshi

Neurophysiological signature(s) of visual hallucinations across neurological and perceptual

Dauwan, Meenakshi

IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite from it. Please check the document version below.

Document Version

Publisher's PDF, also known as Version of record

Publication date: 2019

Link to publication in University of Groningen/UMCG research database

Citation for published version (APA):

Dauwan, M. (2019). Neurophysiological signature(s) of visual hallucinations across neurological and perceptual: and non-invasive treatment with physical exercise. Rijksuniversiteit Groningen.

Copyright

Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).

Take-down policy

If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.

Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons the number of authors shown on this cover page is limited to 10 maximum.

CHAPTER

Exercise improves clinical

symptoms, quality of

life, global functioning

and depression in

schizophrenia: a systematic

review and meta-analysis

Schizophrenia Bulletin, 2016; 42(3): 588-599

1 _{Brain Center Rudolf Magnus, University Medical}

Center Utrecht, The Netherlands

2 _{Department of Clinical Neurophysiology and MEG}

Center, VU University Medical Center, Amsterdam, The Netherlands Meenakshi Dauwan1,2 Marieke J.H. Begemann1 Sophie M. Heringa1 Iris E. Sommer1

9

VSHFWUXPGLVRUGHUVDVLWPD\KDYHEHQHÀFLDOHIIHFWRQFOLQLFDOV\PSWRPVTXDOLW\RI life and cognition.

Methods: A systematic search was performed using PubMed (Medline), Embase,

PsychInfo, and Cochrane Database of Systematic Reviews. Controlled and uncontrolled studies investigating the effect of any type of physical exercise interventions in schizophrenia spectrum disorders were included. Outcome measures were clinical symptoms, quality of life, global functioning, depression or cognition. Meta-analyses were performed using Comprehensive Meta-Analysis software. A random effects model was used to compute overall weighted effect sizes in Hedges’ g.

Results: 29 studies were included, examining 1109 patients. Exercise was superior to

control conditions in improving total symptom severity (k=14, n=719: Hedges’ g=.39, SSRVLWLYHN Q +HGJHV·J SQHJDWLYHN Q +HGJHV· J SDQGJHQHUDON Q +HGJHV·J SV\PSWRPVTXDOLW\ RIOLIHN Q +HGJHV·J SJOREDOIXQFWLRQLQJN Q +HGJHV· J SDQGGHSUHVVLYHV\PSWRPVN Q +HGJHV·J S<RJD VSHFLÀFDOO\LPSURYHGWKHFRJQLWLYHVXEGRPDLQORQJWHUPPHPRU\N Q +HGJHV· J SZKLOHH[HUFLVHLQJHQHUDORULQDQ\RWKHUIRUPKDGQRHIIHFWRQFRJQLWLRQ

Conclusion: Physical exercise is a robust add-on treatment for improving clinical

symptoms, quality of life, global functioning, and depressive symptoms in patients with schizophrenia. The effect on cognition is not demonstrated, but may be present for yoga.

Schizophrenia, a severe psychiatric disorder, affects approximately 24 million people worldwide (Abi-Dargham, 2014). This disorder is characterized by 1) positive symptoms VXFKDVKDOOXFLQDWLRQVDQGGHOXVLRQVQHJDWLYHV\PSWRPVLQFOXGLQJDIIHFWLYHÁDWWHQLQJ DORJLDDQGDYROLWLRQDQGQHXURFRJQLWLYHGHÀFLWVLQFOXGLQJSHUFHSWLRQPHPRU\DQG attention, among others (DSM, 1994). Negative and cognitive symptoms, emerging in the pre-psychotic stage, appear to be related (Strassnig et al., 2015a). Higher negative DQGFRJQLWLYHV\PSWRPVDUHVLJQLÀFDQWO\DVVRFLDWHGZLWKSRRUHUIXQFWLRQDORXWFRPH (Lepage et al., 2014; Ventura et al., 2009). Treatment with antipsychotic drugs, applied DVÀUVWOLQHWKHUDS\W\SLFDOO\UHVXOWLQUHGXFWLRQLQSRVLWLYHV\PSWRPVZLWKPLQLPDOWR no effects on negative and cognitive symptoms (Gold, 2004). In addition, antipsychotics result in the side effects weight gain and metabolic syndrome (Friedman et al., 2010; Goughari et al., 2015). Furthermore, reduced physical capacity in patients with schizophrenia is strongly related to negative and cognitive symptoms (D. Vancampfort et al., 2012a, 2012b). These risk factors are major contributors of cardiovascular diseases in schizophrenia which in turn is associated with 2-3 fold higher mortality rate compared to the general population (Laursen et al., 2012). Saha et al found an all-cause standardized mortality ratio of 2.58 showing an increase in mortality in these patients over the last decades (Saha et al., 2007). Therefore, it is time to implement a therapy for patients with schizophrenia that decreases the negative symptoms and FRJQLWLYHGHÀFLWVDQGDOVRLPSURYHVWKHIXQFWLRQDODQGFOLQLFDORXWFRPH

Physical inactivity has been described as the leading risk factor for global mortality (Blair, 2009). The World Health Organization (WHO) 2009 (World Health Organization, 2009) reported that physical inactivity accounts for 27% of diabetes and 30% of ischemic heart diseases, whereas an active lifestyle reduces these risks, largely improving general health and wellness, and life expectancy (World Health Organization, 2009). Furthermore, physical activity in healthy aging populations is associated with improvement in cognitive functioning and depressive symptoms, delay in age-related cognitive decline and neurodegeneration (Cotman et al., 2007; van Praag, 2009). On the brain level, exercise induces neurogenesis, modulates synaptic plasticity

neurogenesis in the hippocampus and that this is associated with enhanced synaptic plasticity (Cotman et al., 2007; van Praag, 2009; Voss et al., 2013c). Basically, these effects are mediated via the exercise-related increase of several growth factors in the brain such as brain-derived neurotrophic factor (BDNF) (Ahmed et al., 2015; Cotman et al., 2007; Voss et al., 2013c). In schizophrenia, lower peripheral BDNF has been associated with poorer neurocognitive functioning and smaller hippocampal volumes (Ahmed et al., 2015; Song et al., 2014). Exercise has also been associated with changes in regional brain volume and integrity (Voss et al., 2013c). These exercise-induced effects on the brain have been replicated in healthy humans and clinical populations, such as dementia patients, showing increased brain volume in grey and white matter regions, increased white matter integrity in frontal and temporal lobes, increases in BDNF, reduction in depressive symptoms, and improvement in cognitive functioning in young and old (Barnes, 2015; Cerrillo-Urbina et al., 2015; Colcombe et al., 2006; Erickson et al., 2011; Knaepen et al., 2010; Knapen et al., 2015; Knöchel et al., 2012; Smith and Blumenthal, 2010; Stroth et al., 2009; Voss et al., 2013a, 2013b).

Thus, negative symptoms in schizophrenia are an important predictor of unfavorable disease course and outcome and are related to physical inactivity. Therefore, stimulating SK\VLFDOH[HUFLVHLQSDWLHQWVZLWKVFKL]RSKUHQLDPLJKWKDYHEHQHÀFLDOHIIHFWRQFOLQLFDO symptoms. The aim of the present study was to quantitatively review the effects of physical exercise on clinical symptoms in patients with schizophrenia spectrum disorder. Secondarily, the impact of exercise on quality of life (QoL), global functioning, depression, and cognition was investigated.

2.1 Literature search

This meta-analysis was performed according the Preferred Reporting for Systematic Reviews and Meta-analysis (PRISMA) Statement.(Moher et al., 2009) A systematic search was performed in the databases Pubmed (Medline), Embase, PsychInfo and Cochrane Database of Systematic Reviews (independently by M.Dauwan and I.E. Sommer). Combinations of the following search terms were used: ‘schizophrenia’, ‘schizoaffective’, ‘schizophreniform’, ‘psychosis’, ‘psychotic’, ‘exercise’, ‘soccer’, ‘physical’, training’, ‘endurance’, ‘aerobic’, ‘sport’, and ‘yoga’ (Table S1). The search cut-off date was 31st of July 2015. Additionally, the reference lists of the retrieved articles were examined for cross-references. When necessary, corresponding authors were contacted to provide full text details of the study outcome measures. In case the full text of articles was not available and the corresponding information of the authors was not traceable, the abstract was used to retrieve the necessary information to avoid publication bias. There were no year or language limits.

2.2 Inclusion criteria

By consensus (between M. Dauwan and I.E. Sommer), studies meeting the following inclusion criteria were included:

1. Studies investigated the effect of any type of exercise as an intervention on psychiatric symptoms, QoL or global functioning, depressive symptoms or cognition. Both controlled and uncontrolled studies were included.

2. Studies including patients with a diagnosis of schizophrenia or schizophrenia spectrum disorder (schizophreniform, schizoaffective, delusional or psychotic disorder QRWRWKHUZLVHVSHFLÀHGDFFRUGLQJWRWKHGLDJQRVWLFFULWHULDRIWKHDiagnostic and

Statistical Manual of mental Disorders2 _{(DSM-III, DSM-III-R, DSM-IV, DSM-IV-TR), or}

WKH,QWHUQDWLRQDO&ODVVLÀFDWLRQRI'LVHDVHVRU

data upon request.

Studies using combined programs of weight reduction and exercise or using additional cognitive training besides exercise were included in order to obtain as much information as possible.

2.3 Exclusion criteria

1. Studies investigating the effect of exercise on the outcome measures in an uncontrolled study or in a mixed population of patients with schizophrenia and any other psychiatric disorder were excluded from the meta-analyses.

2. Studies that only measured an outcome in the intervention group and not in a FRQWUROJURXSZHUHH[FOXGHGIURPWKHDQDO\VLVRIWKHVSHFLÀFRXWFRPHPHDVXUH Studies that were excluded from the meta-analyses based on these criteria were, however, included in the descriptive (Table S2) and methodological assessment of quality table of the study (Table S3). The outcomes of these studies are reported separately (Table S4).

2.4 Outcome measures

The primary outcome measures were the standardized mean difference in clinical symptoms, measured by total scores on the Positive and Negative Syndrome Scale (PANSS) (Kay et al., 1987) or the Brief Psychiatric Rating Scale (BPRS) (Overall, J, E and Gorham, D, 1962), standardized mean difference in general, positive and negative symptom subscores of the PANSS or scores on the Scale for the Assessment of Positive Symptoms (SAPS) (Andreasen, 1984), or the Scale for the Assessment of Negative Symptoms (SANS) (Andreasen, 1981). Secondary outcome measures included QoL and depressive symptoms as assessed by various QoL and depression questionnaires, global, social and occupational functioning assessed by Global Assessment of Functioning scale (GAF) (Endicott et al., 1976) and Social and occupational Functioning (Assessment) Scale (SOFAS/SOFS) (Saraswat et al., 2006), and mean change in cognitive domains.

'DWDZHUHSRROHGIRUDOOVWXGLHVH[DPLQLQJWKHHIIHFWRIH[HUFLVHRQDVSHFLÀFRXWFRPH measure. Since the follow-up period of the included studies differed and some studies had more than one follow-up moment, the last follow-up moment of the study was used. For studies that added cognitive training to their program a few weeks after the start of the study, the last follow-up moment before the addition was used to investigate only the effect of exercise.

In order to study the effect of the moderating variables type of exercise and type of FRQWUROFRQGLWLRQVWXGLHVZHUHUHFODVVLÀHGEDVHGRQWKHW\SHRIH[HUFLVHDQGW\SH RIFRQWUROFRQGLWLRQ7\SHVRIH[HUFLVHVZHUHUHFODVVLÀHGLQWRWKUHHPDMRUGRPDLQV aerobic (e.g. endurance training, cardiovascular exercises, treadmill walking), anaerobic (muscle strength training), and yoga. Studies using an aerobic exercise program but LQFOXGLQJDQDHURELFH[HUFLVHVWRSURYLGHYDULDWLRQZHUHFODVVLÀHGDVDHURELFZLWKDQ DQQRWDWLRQRIDQDHURELFWUDLQLQJ7\SHVRIFRQWUROFRQGLWLRQVZHUHUHFODVVLÀHGLQWR active (e.g. schizophrenia patients playing table football, computer games or following occupational therapy) or passive (e.g. healthy controls, schizophrenia patients included as waiting list or treatment as usual). See descriptive table for further details (Table S2). When possible, moderator analyses for type of exercise and type of control condition were performed. For exercise, only aerobic and yoga exercise were compared because DQLQVXIÀFLHQWQXPEHURIVWXGLHVH[DPLQHGWKHHIIHFWRIRQO\DQDHURELFH[HUFLVH Effect sizes were computed using Comprehensive Meta Analysis Version 2.0 (Borenstein et al., 2005). Per outcome measure, the magnitude and direction of effect was calculated for each individual study. Since most studies reported pre-exercise and post-exercise means and standard deviations, these values were used to compute the effect sizes. When possible, change scores were used instead of pre-exercise and post-exercise scores in order to avoid overestimation of the true effect size because of pre-post intervention correlation. When means and standard deviations or change scores were not available, effect sizes were computed using exact p- or t-values. Hedges’ g was used to quantify effect sizes of combined studies using a random effects model. A random effects model was used because of the variances in type and duration of applied

of the results on population level (Hedges and Vevea, 1998). First, analyses were performed including all suitable studies per outcome measure. Subsequently, analyses ZHUHUHSHDWHGE\H[FOXGLQJRXWOLHUVWXGLHVGHÀQHGDVVWXGLHVZLWKDQHIIHFWVL]HWKDW deviated more than 2SDs from the mean weighted effect size (Hedges and Olkin, 2014). (IIHFWVL]HVZLWKDSYDOXHRIZHUHFRQVLGHUHGVLJQLÀFDQW

Furthermore, heterogeneity of results across studies was assessed by calculating the homogeneity statistic, I2_{. I}2 _{describes the percentage of total variation across studies}

due to heterogeneity rather than chance. High heterogeneity (i.e. I2_{•PDNHV}

interpretation of results unreliable (Higgins et al., 2003). Potential publication bias ZDVÀUVWLQYHVWLJDWHGE\YLVXDOLQVSHFWLRQRIWKHIXQQHOSORWV$QDV\PPHWULFDOIXQQHO plot indicates publication bias. Afterwards, the funnel plot asymmetry was tested with (JJHU·VWHVWZLWKDOSKDRIVHWDVVLJQLÀFDQFHOHYHOWZRWDLOHG(JJHUHWDO

3. RESULTS

The literature search yielded 61 quantitative studies (Figure S1) that investigated the effect of any type of exercise on the outcome measures (Table S2) (Abdel-Baki et al., 2013; Auchus and Kaslow, 1994; Ball et al., 2001; Battaglia et al., 2013; Dogan, 2008) (Beebe et al., 2005; Behere et al., 2011; Bhatia et al., 2012)(Centorrino et al., 2006; Daumit et al., 2011; I. et al., 2013)(Karen J Dodd et al., 2011; Duraiswamy et al., 2007; Falkai et al., 2013; Faulkner and Sparkes, 1999; Gholipour et al., 2012)(Gomes et al., 2014; Heggelund et al., 2014, 2012, 2011; Q.-L. et al., 2004)(Ikai et al., 2013; Kaltsatou et al., 2014; Kim et al., 2013, 2014, Kimhy et al., 2015, 2014)(B. et al., 2011; J. et al., 2011; Kuo et al., 2013)(Malchow et al., 2015b, 2015a)(Manjunath et al., 2013; Margariti et al., 2012; Marzolini et al., 2009; Oertel-Knöchel et al., 2014a; Paikkatt et al., 2012; Pajonk et al., 2010)(Poulin et al., 2007; Rosenbaum et al., 2015; Scheewe et al., 2013, 2012)(R.C. et al., 2013; Skrinar et al., 2005; Strassnig et al., 2012; Svatkova et al., 2015) (F.L.W. et al., 2010; R.E. et al., 2013; Takahashi et al., 2012; Vancampfort et al., 2011a; Varambally et al., 2012)(J. et al., 2013; Visceglia and Lewis, 2011; Warren et al., 2011; Zwick et al., 2010)(Chan, 2014; Kim and Jin, 2015; Leone et al., 2015; Nath et al., 2001;

investigating 1109 patients, were suitable to be included in the meta-analyses. The assessed methodological quality of all retrieved studies is shown in Table S3.

3.1 Primary outcome measure: clinical symptoms

Fourteen studies, including 659 patients, examined the effect of exercise on total symptom severity (Duraiswamy et al., 2007; Heggelund et al., 2012, 2011; Ikai et al., 2013; Kaltsatou et al., 2014; Malchow et al., 2015a; Manjunath et al., 2013; Q.-L. et al., 2004; Scheewe et al., 2013; Takahashi et al., 2012; Varambally et al., 2012; Visceglia and Lewis, 2011)(Chan, 2014; Kim and Jin, 2015). (Table 1) Exercise showed DVXSHULRUHIIHFWRYHUDFWLYHFRQWUROVVWXGLHV1 +HGJHV·J S SDVVLYHFRQWUROVVWXGLHV1 +HGJHV·J SDQGDQRYHUDOOVLJQLÀFDQW HIIHFW1 +HGJHV·J S)LJXUHLQUHGXFLQJWRWDOV\PSWRPVHYHULW\ Heterogeneity was high (I2_{=61%). The study by Visceglia et al (Visceglia and Lewis,}

2011) was considered an outlier (Hedges’ g = 2.08) and excluded from the analysis. $IWHUH[FOXVLRQWKHRYHUDOO+HGJHV·J S7DEOHDQGPRGHUDWRUEDVHG PHDQZHLJKWHGHIIHFWVL]HVUHPDLQHGVLJQLÀFDQWDFWLYH+HGJHV·J SSDVVLYH +HGJHV·J S7KHHIIHFWVL]HIRUFRPSDULVRQZLWKSDVVLYHDQGRYHUDOOFRQWURO group decreased slightly. The degree of heterogeneity among studies decreased, but remained high (I2_{=54%; Table 1). Egger’s test before and after excluding the outlier}

ZDVQRQVLJQLÀFDQWLQGLFDWLQJQRSXEOLFDWLRQELDV

T a b le 1 : 2 YH UY LHZU HV XO WVI RUD OOR XW FR PHPH DV XU HVV KR Z LQ JH II HF WV L] H   & RQ ÀG HQ FH, QW HU YD OS Y DO XH D QG, 2 Outc om e m e a sure S tudies (N) S ub je ct s /pa ti en ts (N) 1 H e d g e s’ g 9 5 % C I p -va lue I 2 (% ) P A NSS T o ta l 14 6 59 /6 59 .3 9 .1 9 - . 5 8 < .0 0 1 6 1 T o ta l w it h ou t ou tl ie r 13 6 4 1 /6 4 1 .3 7 .1 8 - . 5 7 < .0 0 1 54 P o si ti ve 1 5 64 1 /64 1 .3 2 .1 4 - . 5 0 < .0 1 5 0 Neg ativ e 1 8 7 6 5 /7 6 5 .4 9 .3 1 - . 6 7 < .0 0 1 6 0 Neg ativ e with o u t o u tl ie r 1 8 7 5 0 /7 5 0 .4 9 .33 - . 6 6 < .0 0 1 47 G e n e ra l 1 0 4 36 /4 36 .2 7 .0 4 - . 5 0 < .0 5 5 8 Q u a li ty o f L if e T o ta l 11 2 77 /2 77 .5 5 .35 - .7 6 < .0 0 1 4 9 T o ta l w it h ou t ou tl ie r 1 0 2 5 9 /2 5 9 .4 7 .3 0 - . 6 4 < .0 0 1 29 QoL Ph ysic al 9 2 3 8 /2 3 8 .5 0 .1 1 - . 8 9 < .0 5 6 3 Qo L P h ysic al w it h ou t ou tl ie r 8 2 2 0 /2 2 0 .3 9 .0 5 - .73 < .0 5 4 1 Q o L M e n ta l 9 197 /1 97 .3 8 -. 0 6 - . 8 2 .0 9 6 5 Q o L M e n ta l w it h o u t o u tl ie r 8 17 9/ 17 9 .3 3 -. 07 - . 7 2 .1 0 4 7 QoL S o ci al 5 13 9 /1 3 9 .6 7 .34 - 1 .0 0 < .0 0 1 0 Q o L E n vi ro n m e n ta l 4 10 8 /10 8 .6 2 .2 4 - 1 .0 0 < .0 1 0 Func ti oning 5 2 7 6 /2 7 6 .3 2 .1 1 - . 5 3 < .0 1 0 De pr e ssion 7 31 6 /2 9 6 .7 1 .33 - 1 .0 9 < .0 0 1 7 9 De pr e ssion w it h ou t ou tl ie r 6 2 7 7 /2 5 7 .6 4 .2 7 – 1 .0 2 < .0 1 7 0

Ta b le 1 : C o n ti n u e d Outc om e m e a sure S tudies (N) S ub je ct s /pa ti en ts (N) 1 H e d g e s’ g 9 5 % C I p -va lue I 2 (% ) Co g n it ion W o rk ing me mor y 6 2 6 2 /1 92 .2 3 -. 0 4 . 5 0 .0 9 4 9 L o n g-te rm m e m o ry 6 2 6 2 /2 3 3 .1 4 -. 0 7 - . 3 5 .1 9 1 4 P ro ce ss in g s p e e d 4 2 0 1 /1 8 0 .1 5 -. 1 0 - . 4 0 .2 4 0 A tt e n ti o n & e x e c u ti v e func ti oning 4 2 0 9 /1 8 8 .0 7 -. 17 - . 3 2 .55 5 9 1: T o ta l n u m b e r o f s u b je c ts ( p at ie n ts i n e x e rc is e g ro u p + c o n tro l g ro u p w it h o r w it h o u t s c h iz o p h re n ia ) / o n ly n u m b e r o f s c h iz o p h re n ia p at ie n ts (i n e x e rc is e a n d co n tro l g ro u p ). Bold Y DO X HVL Q G LF DW HV LJ Q LÀ FD Q WH II HF WV L] HV 

9

Figure 1. Meta-analysis of the effect of exercise on total symptom severity. Effect sizes are

grouped by the type of control patients as included in the individual studies. *With additional cognitive training.

The analysis of positive symptoms included 15 studies with a total of 641 patients (Table 1) (Behere et al., 2011; Dogan, 2008; Duraiswamy et al., 2007; Heggelund et al., 2011; Ikai et al., 2013; Kaltsatou et al., 2014; Malchow et al., 2015a, 2015b; Manjunath et al., 2013; Oertel-Knöchel et al., 2014a; Scheewe et al., 2013; Takahashi et al., 2012; Varambally et al., 2012; Visceglia and Lewis, 2011)(Kim and Jin, 2015). Exercise was VXSHULRUWRDFWLYHVWXGLHV1 +HGJHV·J SSDVVLYHVWXGLHV1  +HGJHV·J SDQGRYHUDOOFRQWUROVLQUHGXFLQJSRVLWLYHV\PSWRPV1  +HGJHV·J S)LJXUH+HWHURJHQHLW\ZDVKLJK,2_{=50%, Table 1). No outliers}

Figure 2. Meta-analysis of the effect of exercise on positive symptoms. Effect sizes are grouped

by the type of control patients as included in the individual studies. *With additional cognitive training

Regarding negative symptom scores, 18 studies could be retrieved, including a total of 765 patients (Table 1) (Behere et al., 2011; Dogan, 2008; Duraiswamy et al., 2007; Gholipour et al., 2012; Heggelund et al., 2011; Ikai et al., 2013; Kaltsatou et al., 2014; Malchow et al., 2015b, 2015a; Manjunath et al., 2013; Oertel-Knöchel et al., 2014a; R.E. et al., 2013; Scheewe et al., 2013; Takahashi et al., 2012; Varambally et al., 2012; Visceglia and Lewis, 2011)(Kim and Jin, 2015; Nath et al., 2001). Again, exercise turned RXWWREHVXSHULRURYHUDFWLYHVWXGLHV1 +HGJHV·J SSDVVLYH VWXGLHV1 +HGJHV·J SDQGRYHUDOOFRQWUROVLQLPSURYLQJQHJDWLYH V\PSWRPV1 +HGJHV·J S)LJXUH+HWHURJHQHLW\DPRQJVWXGLHV was high (I2_{=60%). The study by Gholipour et al (Gholipour et al., 2012) was considered}

DQRXWOLHU+HGJHV·J $IWHUUHPRYDOWKHHIIHFWVL]HDQGVLJQLÀFDQFHOHYHOIRUWKH active control group and overall analysis remained unchanged. Effect size for the effect RIH[HUFLVHRYHUSDVVLYHFRQWUROJURXSGHFUHDVHGVOLJKWO\ZLWKXQFKDQJHGVLJQLÀFDQFH +HGJHV·J S+HWHURJHQHLW\DPRQJVWXGLHVGHFOLQHGWRPRGHUDWH,2_=47%).

(JJHU·VWHVWUHPDLQHGQRQVLJQLÀFDQWEHIRUHDQGDIWHUH[FOXVLRQRIWKHRXWOLHU

Figure 3. Meta-analysis of the effect of exercise on negative symptoms. Effect sizes are grouped

by the type of control patients as included in the individual studies. *With additional cognitive training.

Concerning general symptom severity, the meta-analysis, including ten studies with 436 patients studied (Heggelund et al., 2011; Ikai et al., 2013; Kaltsatou et al., 2014; Malchow et al., 2015a, 2015b; Manjunath et al., 2013; Takahashi et al., 2012; Varambally HWDO9LVFHJOLDDQG/HZLV.LPDQG-LQVKRZHGDQRYHUDOOVLJQLÀFDQW HIÀFDF\RIH[HUFLVHWRFRQWUROVLQUHGXFLQJJHQHUDOV\PSWRPVHYHULW\+HWHURJHQHLW\ ZDVFRQVLGHUHGKLJK1 +HGJHV·J S,2_{=58%; Figure 4; Table 1). The}

moderator analysis showed a superior effect of exercise compared to passive controls VWXGLHV1 +HGJHV·J SZKHUHDVQRVLJQLÀFDQWGLIIHUHQFHZDVIRXQG between exercise and active controls (5 studies, N=222: Hedges’ g = .16, p=.24). Egger’s test showed no evidence for publication bias.

Figure 4. Meta-analysis of the effect of exercise on general symptoms. Effect sizes are grouped

by the type of control patients as included in the individual studies. * With additional cognitive training.

Additional moderator analyses for the type of exercise showed that both yoga and aerobic exercise were effective in reducing the total symptom severity (yoga: Hedges’ J   S DHURELF +HGJHV· J   S )LJXUH 6 7DEOH 6 DQG SRVLWLYH V\PSWRPV\RJD+HGJHV·J SDHURELF+HGJHV·J S)LJXUH6 7DEOH67KHVLJQLÀFDQFHRIWKHHIIHFWVL]HVIRUWRWDOV\PSWRPVHYHULW\UHPDLQHG unaffected after exclusion of the outlier study by Visceglia et al (Visceglia and Lewis, 2011) (Table S5). In addition, both yoga and aerobic exercise were comparable in UHGXFLQJQHJDWLYHV\PSWRPV\RJD+HGJHV·J SDHURELF+HGJHV·J  S)LJXUH67DEOH6)LQDOO\\RJDVKRZHGVLJQLÀFDQWHIIHFWLQUHGXFLQJJHQHUDO V\PSWRPVZKHUHDVDHURELFH[HUFLVHZDVQRQVLJQLÀFDQW\RJD+HGJHV·J S aerobic: Hedges’ g = .39, p=.14; Figure S5; Table S5).

3.2 Secondary outcome measures

([HUFLVHWXUQHGRXWWREHVXSHULRULQLPSURYLQJ4R/+HGJHV·J S)LJXUH 6JOREDOIXQFWLRQLQJ+HGJHV·J S)LJXUH6DQGUHGXFLQJGHSUHVVLRQ +HGJHV·J S )LJXUH6,QWKHRYHUDOODQDO\VLVH[HUFLVHZDVQRWVXSHULRU to control conditions in improving any of the cognitive subdomains (Attention & executive functioning: Hedges’ g = .07, p=.55; Processing speed: Hedges’ g = .15, p=.24; Working memory: Hedges’ g = .23, p=.09; Long-term memory: Hedges’ g = .14, p=.19;

+HGJHV·J S)LJXUH67DEOH6

Detailed results on the secondary outcome measures are shown in the supplementary.

4. DISCUSSION

The aim of this study was to quantitatively review all available controlled trials on the HIÀFDF\RISK\VLFDOH[HUFLVHLQSDWLHQWVZLWKDVFKL]RSKUHQLDVSHFWUXPGLVRUGHU7ZHQW\ nine studies, providing data on 1109 patients, were included in the analyses. Results VKRZHGDQRYHUDOOVLJQLÀFDQWHIIHFWRIH[HUFLVHRQFOLQLFDOV\PSWRPVZLWKPHGLXP effect sizes for total and negative symptoms, and small effect sizes for general and SRVLWLYHV\PSWRPV,QWHUHVWLQJO\RQO\\RJDVKRZHGVLJQLÀFDQFHLQLPSURYLQJJHQHUDO symptoms, while both yoga and aerobic exercise were effective in reducing the total symptom severity, and positive and negative symptoms. In addition, both yoga and aerobic exercise showed a similar effect in improving QoL. Furthermore, exercise was EHQHÀFLDOLQLPSURYLQJJOREDOIXQFWLRQLQJDQGGHSUHVVLYHV\PSWRPV1RWDEO\HYLGHQFH SRLQWVPRUHWRZDUGVWKHEHQHÀFLDOHIIHFWRI\RJDLQLPSURYLQJJOREDOIXQFWLRQLQJDQG UHGXFLQJGHSUHVVLYHV\PSWRPV([HUFLVHLQJHQHUDOVKRZHGDWUHQGWRZDUGVVLJQLÀFDQFH in improving the cognitive subdomain processing speed, while no effect on the other cognitive subdomains was demonstrated. However, a positive effect of yoga on the FRJQLWLYH VXEGRPDLQ ORQJWHUP PHPRU\ DQG D WUHQG WRZDUGV VLJQLÀFDQFH IRU WKH subdomain attention and executive functioning were observed.

Previous work examining the contribution of psychiatric symptoms to functional RXWFRPHUHYHDOVWKDWQHJDWLYHV\PSWRPVDUHVLJQLÀFDQWO\DVVRFLDWHGZLWKIXQFWLRQDO outcome in schizophrenia (Fervaha et al., 2014; Herbener and Harrow, 2004; Strassnig et al., 2015a; Ventura et al., 2009), while positive symptoms are less strongly correlated with functional impairment (Lin et al., 2013; Ventura et al., 2009). Moreover, negative symptoms are related to neurocognition and thereby also indirectly affect the outcome by mediating the relationship between neurocognition and outcome (Lin et al., 2013; Ventura et al., 2009). A reduction of negative symptoms in patients with schizophrenia

these symptoms (Soundy et al., 2015). A 2011 Cochrane review by Gorczynski and Faulkner (Gorczynski and Faulkner, 2011) including three randomized controlled trials 5&7IRXQGVLJQLÀFDQWHIIHFWRIH[HUFLVHRQQHJDWLYHV\PSWRPVEXWQRWRQSRVLWLYH symptoms (Gorczynski and Faulkner, 2011). However, since then many controlled VWXGLHVKDYHEHHQFRQGXFWHGWKDWSURYLGHHYLGHQFHIRUEHQHÀFLDOHIIHFWVRISK\VLFDO exercise on clinical symptoms in schizophrenia spectrum disorders (Behere et al., 2011; Chan, 2014; Gholipour et al., 2012; Heggelund et al., 2012, 2011; Ikai et al., 2013; Kaltsatou et al., 2014; Kim and Jin, 2015; Malchow et al., 2015b, 2015a; Manjunath et al., 2013; Nath et al., 2001; Oertel-Knöchel et al., 2014; R.E. et al., 2013; Scheewe et al., 2013; Takahashi et al., 2012; Varambally et al., 2012; Visceglia and Lewis, 2011). These individual studies are included in the present meta-analyses. Therefore, the SUHVHQW TXDQWLWDWLYH UHYLHZ LQFOXGLQJ  VWXGLHV VKRZV KLJKO\ VLJQLÀFDQW HIIHFW RI physical exercise on both positive and negative symptoms when compared to the control situation.

4.1 Type of exercise

7KHSUHVHQWPHWDDQDO\VHVVKRZEHQHÀFLDOHIIHFWVRIERWKDHURELFH[HUFLVHDQG\RJD on most outcomes measured, with better results for yoga in several occasions. Physical exercise in schizophrenia reduces psychological distress and state anxiety, while yoga, VSHFLÀFDOO\ KDV EHHQ SURSRVHG WR UHGXFH SRVLWLYH DQG QHJDWLYH V\PSWRPV 'DY\ Vancampfort et al., 2012). Eight studies included in the meta-analyses investigated yoga. All these studies applied the same type of yoga program consisting of asanas (i.e. postures or exercise as standing, twisting, sun salutation, balance, joint rotations), pranayama (i.e. breathing exercises), and relaxation exercises. None of the studies included meditation teachings in their yoga program.

Besides, exercise variables such as frequency, intensity, session duration, total intervention duration, and either or not supervision is present are non-trivial factors WKDQFDQVKDSHH[HUFLVHLQDZD\WKDWLVPRVWHIIHFWLYHIRUDVSHFLÀFSDWLHQWJURXS,Q the present study, exercise sessions ranged from 16 minutes once (Heggelund et al., 2014) to 360 – 720 minutes per week (Takahashi et al., 2012). The total intervention

most studies was around 12 weeks. For aerobic exercise, mostly 90 – 120 minutes of exercise per week were devoted, which is in line with the recommended amount of time patients with schizophrenia are advised to engage in physical exercise (Stanton et al., 2014). Although, the American College of Sports Medicine (ACSM) recommends a moderate-intensity cardiorespiratory exercise training of 150 minutes per week in adults (Garber et al., 2011a). This indicates that the current exercise interventions PD\QRWEHVXIÀFLHQWLQWKHSUHVHQWSDWLHQWSRSXODWLRQ+RZHYHUQRUPDOO\LQDFWLYH SHUVRQVHJVFKL]RSKUHQLDSDWLHQWVFDQDOVREHQHÀWIURPOHVVDPRXQWVRIH[HUFLVH than recommended by the ACSM (Garber et al., 2011a). On the other hand, almost all studies in the meta-analyses made use of group exercise under supervision as is also recommended by the ACSM (Garber et al., 2011a). Group exercises are cost-effective and aid in the development of sense of relatedness (Stanton et al., 2014). Therefore, qualitative assessment of the present data recommends clinicians to implement supervised group exercise programs of at least 30min/day, 3 times per week for minimally 12 weeks, but to consider long-term continuation for more robust effects.

4.2 Strengths, limitations & recommendations

The greatest strength of the present study is that it provides an up-to-date, and H[WHQVLYHTXDQWLWDWLYHDQGTXDOLWDWLYHRYHUYLHZRIWKHOLWHUDWXUHUHJDUGLQJWKHHIÀFDF\ of different exercise interventions in schizophrenia. A previous meta-analysis by Firth and colleagues (Firth et al., 2015) on the same topic as the present study showed only EHQHÀFLDOHIIHFWRIH[HUFLVHLQUHGXFLQJSRVLWLYHDQGQHJDWLYHV\PSWRPV)LUWKHWDO 2015). However, Firth et al investigated only English peer-reviewed RCTs, and excluded studies investigating yoga and/or combined weight-reduction and exercise programs as intervention. The present study, however, included all the studies meeting these criteria and therefore was able to perform meta-analyses in several clinical outcome measures. Furthermore, the present study showed that the best results of exercise are seen in the yoga intervention groups. These results are one of the main differences with the study by Firth et al. Furthermore, the current study aimed to compare exercise

JUHDWHUHIÀFDF\LQLPSURYLQJYDULRXVRXWFRPHPHDVXUHV7KHVWXG\E\)LUWKHWDO)LUWK et al., 2015) pooled data from studies with more than two non-exercise groups (Firth et al., 2015) and therefore couldn’t compare exercise with different types of control conditions. Finally, the present study also included uncontrolled studies on exercise to provide a comprehensive qualitative overview of the results found in these studies, whereas Firth and colleagues excluded these type of studies (Firth et al., 2015). These key differences with the study by Firth et al (Firth et al., 2015) emphasize the strength of the present results.

This study has also some limitations. First, only six studies could be included in the cognitive meta-analyses, so that the overall effect of exercise on cognition was underpowered and therefore not fully reliable. However, the results of the studies on FRJQLWLRQWKDWFRXOGQRWEHLQFOXGHGDOVRVKRZHGEHQHÀFLDOHIIHFWRIH[HUFLVH7DEOH S4). Second, publication bias is an important possible drawback in meta-analytical studies. To take this into account, funnel plots were visually inspected and the funnel plots asymmetry was tested with Egger’s test. However, none of the inspected and tested funnel plots showed asymmetry, increasing the validity of the found results. 7KLUGDQLQVXIÀFLHQWQXPEHURIVWXGLHVH[DPLQHGWKHHIIHFWRIRQO\DQDHURELFH[HUFLVH making a moderator analysis for this group not feasible. On the other hand, previous ÀQGLQJVUHJDUGLQJWKHLQHIÀFDF\RIDQDHURELFH[HUFLVHRQEUDLQIXQFWLRQLQJVXSSRUWLWV inferiority to other types of exercise (Firth et al., 2015).

)RUIXWXUHUHVHDUFKLQWRWKHHIÀFDF\RIH[HUFLVHLQSDWLHQWVZLWKVFKL]RSKUHQLDVSHFWUXP disorders, a few methodological concerns should be taken into account. First, at least two-arm studies should be performed instead of single-arm studies. Second, the use of an active control condition matched for time and personal contact is recommended. Blinded assessment of outcome measures, making use of standardized questionnaires and cognitive batteries are also recommendable (Lin et al., 2013), applying supervised interventions to decrease drop-out rates (Alexandratos et al., 2012), and providing full data on type of exercise and associated program variables used. By this, homogeneity RIVWXGLHVZLOOEHLQFUHDVHGDQGÀQGLQJVZLOOEHPRUHUREXVWDQGJHQHUDOL]DEOH

clinical symptoms, QoL, global functioning, and depressive symptoms, with notably evidence for yoga in improving clinical symptoms, depression, and global functioning. In addition, yoga may have a positive effect on cognition, while no overall effect of exercise could be shown for cognition. Qualitative assessment of data recommends clinicians to implement supervised group exercise programs of at least 30min/day, 3 times per week for minimally 12 weeks, but to consider long-term continuation for more robust effects.

Results

Quality of Life

For total QoL, eleven studies examining 277 patients could be retrieved (Battaglia et al., 2013; Dogan, 2008; Duraiswamy et al., 2007; Gomes et al., 2014; Heggelund et al., 2012, 2011; Ikai et al., 2013; Kaltsatou et al., 2014; Marzolini et al., 2009; Oertel-.Q|FKHOHWDOD9LVFHJOLDDQG/HZLV([HUFLVHVKRZHGVXSHULRUHIÀFDF\WR FRQWUROFRQGLWLRQ1 +HGJHV·J S)LJXUH6LQLPSURYLQJ4R/EXW studies were heterogeneous (I2_{=49%; Table 1). Considering moderator analyses in}

improving total QoL, exercise was found to be superior over performing no activities VWXGLHV1 +HGJHV·J SDQGVKRZHGDWUHQGWRZDUGVVXSHULRULW\ when compared to the active control group (4 studies, N=247: Hedges’ g = .36, p=.06). The study by Battaglia et al (Battaglia et al., 2013) was considered an outlier (Hedges’ g = 2.32). After exclusion, the effect size for exercise over controls in total, DQGVSHFLÀFDOO\RYHUWKHSDVVLYHFRQWUROJURXSGHFUHDVHGEXWUHPDLQHGVLJQLÀFDQW RYHUDOO+HGJHV·J SSDVVLYH+HGJHV·J S7KHHIIHFWVL]HIRU comparison with active control group remained unchanged (Hedges’ g = .36, p=.06). Heterogeneity among studies decreased, but remained moderate (I2_{=29%). Both,}

before and after exclusion of the outlier study, Egger’s test for funnel plot asymmetry ZDVQRQVLJQLÀFDQW

0RGHUDWRUDQDO\VHVIRUWKHW\SHRIH[HUFLVHVKRZHGVLPLODUHIÀFDF\IRUDHURELFH[HUFLVH +HGJHV·J SDQG\RJD+HGJHV·J SRYHUFRQWUROVLQLPSURYLQJ 4R/)LJXUH67DEOH67KHVLJQLÀFDQFHRIWKHHIIHFWVL]HVUHPDLQHGXQDIIHFWHG after exclusion of the outlier study by Battaglia et al (Table S5) (Battaglia et al., 2013). Since QoL has several indicators, subanalyses were performed for the physical, mental, social, and environmental subdomain of QoL. Exercise turned out to be superior over FRQWUROVIRUWKHSK\VLFDOVWXGLHV1 +HGJHV·J S)LJXUH6VRFLDO VWXGLHV1 +HGJHV·J S)LJXUH6DQGHQYLURQPHQWDOVWXGLHV 1 +HGJHV·J S)LJXUH64R/7DEOH6WXGLHVPHDVXULQJSK\VLFDO

trend over controls (9 studies, N=205: Hedges’ g = .38, p=.09; Figure S11), but the data were heterogeneous (I2_{=65%; Table 1).}

Moderator analyses for the social and environmental QoL were not possible. For the physical and mental subdomain, exercise was found to be superior over passive control JURXSSK\VLFDOVWXGLHV1 +HGJHV·V SPHQWDOVWXGLHV1  +HGJHV·J SEXWQRWRYHUDFWLYHFRQWUROJURXSSK\VLFDOVWXGLHV1  Hedges’ g = .30, p=.31; mental: 4 studies, N=92: Hedges’ g = .02, p=.95). Further, both mental and physical subdomains contained an outlier (Hedges’ g = 2.32) (Battaglia et al., 2013). After removal, effect sizes for exercise over passive control group for ERWKVXEGRPDLQVGHFUHDVHGEXWUHPDLQHGVLJQLÀFDQWSK\VLFDO+HGJHV·J S PHQWDO+HGJHV·J S7KHHIIHFWVL]HVIRUH[HUFLVHRYHUDFWLYHFRQWUROJURXS in both subdomains even as overall effect of exercise over controls in the mental subdomain remained unchanged. The latter was decreased in the physical subdomain ZLWKQRFKDQJHLQVLJQLÀFDQFH+HGJHV·J S$IWHUH[FOXVLRQRIWKHRXWOLHU heterogeneity decreased in both subdomains from high to moderate (Table 1). For all WKHVXEGRPDLQDQDO\VHV(JJHU·VWHVWVZHUHQRQVLJQLÀFDQWLQGLFDWLQJQRSXEOLFDWLRQ bias.

Global Functioning

Five studies, examining 276 patients on functioning, could be included in the analyses (Behere et al., 2011; Duraiswamy et al., 2007; Kaltsatou et al., 2014; Malchow et al., 2015b; Varambally et al., 2012). The meta-analysis showed an overall superiority of exercise over controls in improving global, social, and occupational functioning in VFKL]RSKUHQLDSDWLHQWV1 +HGJHV·J S)LJXUH6ZLWKGDWDEHLQJ KRPRJHQHRXV7DEOH6SHFLÀFDOO\H[HUFLVHZDVIRXQGWREHPRUHHIIHFWLYHRYHU SDVVLYHFRQWUROJURXSVWXGLHV1 +HGJHV·J SWKDQRYHUDFWLYH control group (4 studies, N=189: Hedges’ g = .23, p=.11). The Egger’s test was non-VLJQLÀFDQW

LPSURYLQJJOREDOIXQFWLRQLQJ\RJD+HGJHV·J SDHURELF+HGJHV·J  p=.24; Figure S13, Table S5).

Both, increased severity of negative and general (such as anxiety and depression) symptoms have been associated with poorer QoL (Hofer et al., 2006; Suttajit and Pilakanta, 2015; Woon et al., 2010). Contrary, the effect of positive symptoms on QoL is still inconclusive (Fitzgerald et al., 2001; Savilla et al., 2008; Suttajit and Pilakanta, 2015; Woon et al., 2010). Regarding cognitive symptoms, the subdomains attention, executive functioning and memory have been associated with impaired quality of life (Ritsner, 2007; Savilla et al., 2008; Woon et al., 2010). Moreover, a negative correlation has been described between QoL and functioning in schizophrenia patients (Cichocki HWDO:RRQHWDO7KHVHÀQGLQJVLQGLFDWHWKDWFOLQLFDOV\PSWRPV4R/ and functioning in schizophrenia are interrelated. Besides an indirect effect via improving clinical and cognitive symptoms, results of this meta-analysis also show a GLUHFWEHQHÀFLDOHIIHFWRISK\VLFDOH[HUFLVHRQ4R/DQGIXQFWLRQLQJ/DFNRISK\VLFDO DFWLYLW\GXULQJOHLVXUHWLPHLVDVLJQLÀFDQWSUHGLFWRUIRULPSDLUHGKHDOWKUHODWHG4R/ in patients with schizophrenia (Vancampfort et al., 2011b). A recent meta-analysis in SDWLHQWVZLWKPHQWDOLOOQHVVUHSRUWHGDPRGHUDWHVLJQLÀFDQWHIIHFWRISK\VLFDODFWLYLW\LQ improving QoL (Rosenbaum et al., 2014). Functional exercise capacity is also positively correlated with global functioning (D. Vancampfort et al., 2012b). In addition, low intensity exercise interventions such as yoga therapy seem to be more successful in improving QoL and wellbeing than other forms of exercise (Alexandratos et al., 2012). 7KHUHVXOWVRIWKHSUHVHQWVWXG\DUHLQOLQHZLWKWKHVHÀQGLQJVIRUWKHHIIHFWRI\RJD therapy on QoL and functioning, but also provide evidence for a positive effect of aerobic exercise on QoL.

Depression

Seven studies investigated depression in 296 patients and could be included in the analysis (Heggelund et al., 2014, 2011; Malchow et al., 2015b; Manjunath et al., 2013; 6FKHHZHHWDO&KDQ.LPDQG-LQ([HUFLVHVKRZHGVXSHULRUHIÀFDF\ RYHUSDVVLYHVWXGLHV1 +HGJHV·J SDFWLYHVWXGLHV1 

After exclusion of the outlier study by Scheewe et al,(Scheewe et al., 2013) the effect VL]HGHFUHDVHGEXWUHPDLQHGVLJQLÀFDQW+HGJHV·J S7KHKHWHURJHQHLW\ remained high (Table 1). Egger’s test, before and after exclusion of the outlier, remained QRQVLJQLÀFDQW

0RGHUDWRUDQDO\VLVIRUWKHW\SHRIH[HUFLVHVKRZHGRQO\VLJQLÀFDQFHIRU\RJDLQUHGXFLQJ GHSUHVVLRQLQVFKL]RSKUHQLD\RJD+HGJHV·J SDHURELF+HGJHV·J  S )LJXUH67DEOH67KHVLJQLÀFDQFHRIWKHHIIHFWVL]HUHPDLQHGXQDIIHFWHG after exclusion of the outlier study by Scheewe et al (Scheewe et al., 2013) (Table S5). Depression is well known to occur in the course of schizophrenia with a reported prevalence range of 25% to 81%. It has also been associated with worse outcome, functional impairment, higher rates of relapses and suicides (Bosanac and Castle, 2013; Siris, 2000). Besides antidepressants, both, atypical antipsychotics and clozapine have shown consistent effects in treating depressive symptoms in schizophrenia (Bosanac and Castle, 2013; Nakajima et al., 2015). On the other hand, comparison between HIÀFDF\RISKDUPDFRWKHUDS\DQGSK\VLFDOH[HUFLVHLQWUHDWLQJGHSUHVVLYHV\PSWRPV showed that clinically depressed patients that participated in physical exercise could reduce their medication dose, while other studies reported reduction in depressive V\PSWRPVDOUHDG\DIWHUZHHNVRIH[HUFLVH.Q|FKHOHWDO7KHÀQGLQJVRI WKHSUHVHQWVWXG\DUHLQOLQHZLWKWKHVHÀQGLQJVE\VKRZLQJDVLJQLÀFDQFHRISK\VLFDO exercise in reducing depressive symptoms in patients with schizophrenia.

Cognition

Fourteen studies investigating the effect of exercise on cognition were retrieved (Bhatia et al., 2012; Chan, 2014; Daumit et al., 2011; J. et al., 2011, 2013, Kimhy et al., 2014, 2015; Malchow et al., 2015; Oertel-Knöchel et al., 2014; Pajonk et al., 2010; R.C. et al., 2013; Rosenbaum et al., 2015; Strassnig et al., 2015; Zwick et al., 2010). From these, eight studies were excluded from the analyses because of the following reasons: cognition data not shown in the study, no full-text available or not enough data to calculate the effect sizes, single-arm study, and studies providing only baseline cognitive data (Daumit

et al., 2015; Strassnig et al., 2015b). Results from these studies are shown separately (Table S4). Six studies were included in the analyses (Bhatia et al., 2012; Malchow et al., 2015b; Oertel-Knöchel et al., 2014a; Pajonk et al., 2010; Zwick et al., 2010)(Chan, 2014). The different cognitive domains and/or cognitive tests from the latter studies ZHUHUHFODVVLÀHGIRUWKHDQDO\VHVLQWKHGRPDLQVDWWHQWLRQ H[HFXWLYHIXQFWLRQLQJ processing speed, long-term memory, and working memory including a total of 233 patients (Table 1).

Four studies investigated the cognitive subdomain attention & executive functioning (Bhatia et al., 2012; Malchow et al., 2015b; Zwick et al., 2010), and processing speed (Bhatia et al., 2012; Malchow et al., 2015b; Oertel-Knöchel et al., 2014a), while all the six studies investigated working memory and long-term memory (Table 1) (Bhatia et al., 2012; Malchow et al., 2015b; Oertel-Knöchel et al., 2014a; Pajonk et al., 2010; Zwick et DO7KHPHWDDQDO\VLVIRUFRJQLWLRQGLGQRWUHYHDODQRYHUDOOVLJQLÀFDQWGLIIHUHQFH between the exercise and control group for any of the subdomains (Attention & executive functioning (N=537): Hedges’ g = .07, p=.55; Figure S16; Processing speed (N=252): Hedges’ g = .15, p=.24; Figure S17; Working memory (468): Hedges’ g = .23, p=.09; Figure S18; Long-term memory (N=411): Hedges’ g = .14, p=.19; Figure S19) 7DEOH+RZHYHUH[HUFLVHVKRZHGDWUHQGWRZDUGVVLJQLÀFDQFHLQLPSURYLQJZRUNLQJ memory. The heterogeneity among studies varied from low to high (I2_{=59%, I}2_=0%,

I2_{=49%, I}2_{=14%, respectively). However, no outliers were detected. For all sub-analyses,}

IXQQHOSORWDV\PPHWU\WHVWHGZLWK(JJHU·VWHVWVZDVQRQVLJQLÀFDQWLQGLFDWLQJQR publication bias.

For the cognitive subdomains attention & executive functioning, working memory, and long-term memory, moderator analyses for the type of control group showed no VLJQLÀFDQWGLIIHUHQFHEHWZHHQVFKL]RSKUHQLDSDWLHQWVSHUIRUPLQJH[HUFLVHRYHUSDWLHQWV in the active or passive control group (Figure S16-S19). For the cognitive subdomain SURFHVVLQJVSHHGH[HUFLVHVKRZHGDWUHQGWRZDUGVVLJQLÀFDQFHRYHUSDVVLYHFRQWURO FRQGLWLRQ LQ LPSURYLQJ FRJQLWLRQ +HGJHV· J   S  ZKHUHDV QR VLJQLÀFDQW difference was found between exercise and the active control condition (Figure S17).

\RJDLQLPSURYLQJWKHVXEGRPDLQORQJWHUPPHPRU\<RJD+HGJHV·J S aerobic: Hedges’ g = .02. p=.90; Figure S20; Table S5) and a trend of yoga towards VLJQLÀFDQFHLQLPSURYLQJWKHFRJQLWLYHVXEGRPDLQDWWHQWLRQ H[HFXWLYHIXQFWLRQLQJ (Yoga: Hedges’ g = .38, p=.08; aerobic: Hedges’ g = -.04, p=.78, Figure S21; Table S5). Unfortunately, both yoga and aerobic exercise showed no effect on the other two cognitive subdomains (Figure S22 & S23; Table S5).

In schizophrenia, verbal memory is one of the most consistently impaired cognitive GRPDLQDQGIRUPVZLWKVRFLDOFRJQLWLYHGHÀFLWVWKHPRVWUREXVWPDUNHUVIRUFOLQLFDO RXWFRPH,QDGGLWLRQGHÀFLWVLQZRUNLQJPHPRU\DQGH[HFXWLYHIXQFWLRQLQJKDYHEHHQ associated with poor social and occupational functioning (Lepage et al., 2014). The effect of exercise on cognitive functioning in schizophrenia patients is less clear (Knöchel et al., 2012). The ‘age-dependence’ hypothesis proposed by Hotting et al (Hötting and Röder, 2013) states that physical exercise particularly affects cognitive functions at ages at which these functions undergo developmental changes or show decline (Hötting and 5|GHU&RQÀUPWKLVK\SRWKHVLVFDUGLRUHVSLUDWRU\IXQFWLRQKDVEHHQDVVRFLDWHG with cognitive performance in old adults (>55 years) but not in young adults (18-31 years) (Hayes et al., 2014). In schizophrenia, cognitive impairments are typically present in early phase of life (adolescence/young adulthood) and exercise may therefore be less effective than in elderly (Carrión et al., 2015; Vöhringer et al., 2013)(Fusar-Poli, 2014). However, attention and executive functioning have been located in (subregions of) the prefrontal cortex (Callahan and Terry, 2015; Talpos and Shoaib, 2015), which is fully developed in late adolescence (Best and Miller, 2010). Therefore, physical exercise in adolescence could improve attention and executive functioning in (early-onset) schizophrenia (Fusar-Poli, 2014; Vancampfort et al., 2015b).Low serum BDNF levels have also been associated with impaired neurocognitive functioning in schizophrenia (Zhang et al., 2012). A recent meta-analysis in schizophrenia obtained a positive VLJQLÀFDQWHIIHFWEHWZHHQSHULSKHUDO%'1)OHYHOVDQGFRJQLWLRQ$KPHGHWDO One of the retrieved studies in the present meta-analysis showed exercise-induced

(Kimhy et al., 2015). Therefore, BDNF has been described as a pathway by which SK\VLFDOH[HUFLVHLQÁXHQFHVEUDLQIXQFWLRQ/LHWDO9RVVHWDOD7KLVHIIHFW is mediated by both acute and chronic aerobic exercise, but not by anaerobic exercise (Huang et al., 2014). Remarkably, it should be noted that the mentioned results on brain function and cognition are the result of long-term exercise intervention (Erickson et al., 2011; Voss et al., 2010). Observational studies indicate that physical activity may need years to affect brain function and cognition (Andel et al., 2008; Lautenschlager et al., 6PLWKDQG%OXPHQWKDOZLWKVLJQLÀFDQWHIIHFWVUHSRUWHGIURPDPLQLPXP of one year intervention (Voss et al., 2010).

Lastly, cognitive functioning can also be enhanced by cholinergic, glutamatergic (stimulates N-methyl-D-aspartate receptors: NMDA), and dopaminergic agents (Li et al., 2014; Vöhringer et al., 2013). Recently, Vreeker et al (Vreeker et al., 2015) summarized in a review that only cholinergic agents show the most promising results in enhancing cognitive functioning in schizophrenia, while results of the latter two agents are inconclusive (Vreeker et al., 2015). Physical exercise, however, has been associated with enhancement in dopaminergic and cholinergic levels, and NMDA channel activity in the brain and therefore shows evidence for cognitive enhancement (Cotman et al., 2007; Li et al., 2014).

How to get schizophrenia patients into physical exercise?

Unfortunately, only one quarter of patients with schizophrenia meets the minimum public health recommendation of 150 minutes per week of moderate-intensity activity )DXONQHUHWDO1XPHURXVVWXGLHVKDYHDWWHPSWHGWRÀQGEDUULHUVWKDWGHWHU SDWLHQWVIURPSDUWLFLSDWLQJLQH[HUFLVHDQGLGHQWLÀHGIDFWRUVDVVLGHHIIHFWVRIPHGLFDWLRQ VRFLDOEDUULHUVDQGDQ[LHW\IHDURISDLQQHJDWLYHV\PSWRPVORZVHOIHIÀFDF\SRRU access to exercise facilities, negative past experiences in sport, lack of motivating atmosphere and individualized programs, and key role of staff in role modeling and promoting physical activity (Bassilios et al., 2015; Beebe et al., 2011; Gorczynski and Faulkner, 2011; Leutwyler et al., 2013; Soundy et al., 2015; D. Vancampfort et al., 2012a). +RZHYHUPRWLYDWLRQDOGHÀFLWVKDYHEHHQUHFRJQL]HGDVWKHPRVWLPSRUWDQWEDUULHU

behavior and the associated internal processes like interest, curiosity, and drive that causes a person to plan, initiate, and execute activities (Foussias et al., 2015; Kleinginna Jr and Kleinginna, 1981; Nakagami et al., 2008). Vancampfort et al (Vancampfort et al., 2015a) showed that negative symptoms are correlated with low autonomous PRWLYDWLRQWRZDUGVH[HUFLVH9DQFDPSIRUWHWDOD)XUWKHUPRUHGHÀFLWVLQ motivation have been shown to mediate the relationship between neurocognition and functional outcome in schizophrenia (Foussias et al., 2015; Nakagami et al., 2008; Ventura et al., 2009). Therefore, motivation can be a pertinent treatment target to get patients into exercise and improve functional outcome. The self-determination theory (SDT) and transtheoretical model (TTM), a motivation theory and behavior change model respectively, provide an explanatory framework why people pursue certain goals and behaviors (Bassilios et al., 2015; Gard et al., 2014). SDT proposes motivation as a continuum of self-determination with amotivation to change behavior at the lowest end of the continuum, and intrinsic motivation to engage in an activity for its own sake. TTM categorizes the readiness of a person to change its behavior DQGFRQVLVWVRIÀYHVWDJHV9DQFDPSIRUWHWDO3DWLHQWVZLWKVFKL]RSKUHQLDKDYH been shown to be more motivated towards physical activity in preparation, action, and maintenance stage of the TTM than in the pre-preparation stage (Vancampfort et DO)XUWKHUPRUHWKH\H[SHULHQFHPRUHGLIÀFXOW\LQHQJDJLQJLQJRDOGLUHFWHG behavior when a rewarding stimulus is lacking (Gard et al., 2014). Thus, mental health professionals should focus on enhancing autonomous motivation. According to the SDT, clinicians can facilitate autonomous motivation by providing an environment where patients can experience psychological freedom when participating in physical exercise, where patients can attain desired outcomes, and which gives patients a sense of social connectedness (Vancampfort et al., 2015a). Clinicians can accomplish this by SURYLGLQJH[HUFLVHWKDWLVWDLORUHGWRWKHFDSDELOLWLHVRIWKHSDWLHQWVSURYLGLQJVXIÀFLHQW but detailed exercise instructions and information to stimulate patients to change behavior towards physical activity, offering clear choices and group sessions to increase relatedness, supporting patients in their initiatives, and using autonomy supportive

sustained by frequent participation in exercise over a longer time period (Gard et al., 2014; Vancampfort et al., 2015a, 2014). Multidisciplinary assistance from professionals is required to maintain this engagement in exercise (Bassilios et al., 2015; Gorczynski and Faulkner, 2011).

T a b le S 5 : 2 YH UY LHZUH VX OW VI R UW KHP R G HU DW R UW \S HR IH [H UF LV HI R UD OOR X WFR P HP HD VX UH VV KR Z LQ JH II HF WV L] H& R Q À G HQ FH, Q WH UY DO S Y Ou tc o m e m e a sur e S tudi es ( n ) Yo g a /A e r o b ic Su b je c ts ( n ) Yo g a /A e r o b ic 1 Y o ga Aerobic ex ercise H e d g e s’ g 9 5 % C I p -va lue H ed ge s’ g 9 5 % C I p P A NSS To ta l 6 /7 3 8 0 /3 2 6 .4 4 .1 8 - .7 1 < .0 1 .5 9 .1 7 - 1 .0 1 < T o ta l w it h ou t ou tl ie r 5 /7 36 2 /3 2 6 .3 7 .1 6 - . 5 8 < .0 0 1 .5 9 .1 7 - 1 .0 1 < P osi ti ve 6 /9 3 9 5 /3 2 0 .3 1 .1 1 - . 5 0 < .0 1 .4 3 .0 3 - . 8 3 < Neg ativ e 6/ 9 3 9 5 /3 20 .4 6 .2 6 - . 6 6 < .0 0 1 .4 2 .0 9 - .7 5 < Gen e ra l 4 /6 2 6 1 /2 14 .4 4 .0 6 - . 8 2 < .0 5 .3 4 -. 14 - . 8 2 .1 Q u a li ty o f L if e To ta l 3 /7 3 3 4 /4 1 0 .5 8 .2 8 - . 8 9 < .0 0 1 .5 9 .3 0 - . 8 7 < T o ta l w it h ou t ou tl ie r 3 /6 3 3 4/3 7 4 .5 8 .2 8 - . 8 9 < .0 0 1 .4 6 .2 5 - . 6 7 < Func ti oning 3 /2 2 6 8 /7 4 .3 3 .0 9 - . 5 8 < .0 1 .2 7 -.18 . 7 2 .2 De pr e ssion 2 /5 13 8 /1 9 9 1 .1 5 .7 9 - 1 .5 1 < .0 0 1 .5 9 -.1 4 1. 3 2 .1 De pr e ssion w it h ou t ou tl ie r 2 /4 13 8 /1 6 0 1 .1 5 .7 9 - 1 .5 1 < .0 0 1 .2 4 -.18 . 6 6 .2 Co g n it ion W o rk in g m e m o ry 2 /4 2 0 9 /2 5 9 .2 1 -. 0 6 - . 4 8 .1 2 .2 7 -. 14 - . 6 8 .2 Long-ter m me mor y 2 /4 1 8 4 /2 2 7 .3 2 .0 4 - . 6 1 < .0 5 .0 2 -. 2 6 - . 3 0 .9 P ro ce ss in g s p e e d 2 /2 13 1 /1 2 1 .2 7 -. 07 - . 6 1 .1 2 .0 2 -. 3 3 - . 3 7 .9 A tt e n ti o n & e x e cu ti ve func ti oning 2 /2 1 8 4 /3 5 3 .3 8 -. 0 5 - . 8 1 .0 8 -. 0 4 -. 3 4 - . 2 6 .78 1: T o ta l n u m b e r o f s u b je c ts ( p at ie n ts i n e x e rc is e g ro u p + c o n tro l g ro u p w it h o r w it h o u t s c h iz o p h re n ia )

Figure S1:35,60$ÁRZFKDUWRIOLWHUDWXUHVHDUFK

Figure S2: Meta-analysis of the effect of exercise on total symptom severity. Effect sizes are

grouped by the type of exercise applied in the individual studies.

Figure S3: Meta-analysis of the effect of exercise on positive symptom subscale. Effect sizes

are grouped by the type of exercise applied in the individual studies.

Figure S4: Meta-analysis of effect of exercise on negative symptom subscale. Effect sizes are

Figure S5: Meta-analysis of the effect of exercise on general symptom subscale. Effect sizes

are grouped by the type of exercise applied in the individual studies.

Figure S6: Meta-analysis of the effect of exercise on total quality of life. Effect sizes are grouped

by the type of control patients as included in the individual studies. * With additional cognitive training.

Figure S7: Meta-analysis of the effect of exercise on total quality of life. Effect sizes are grouped

by the type of exercise applied in the individual studies.

Figure S8: Meta-analysis of the effect of exercise on the physical subdomain of quality of life.

Figure S9: Meta-analysis of the effect of exercise on the social subdomain of quality of life.

Figure S10: Meta-analysis of the effect of exercise on the environmental subdomain of quality

of life.

Figure S11: Meta-analysis of the effect of exercise on the mental subdomain of quality of life.

Effect sizes are grouped by the type of control patients as included in the individual studies.

* With additional cognitive training.

Figure S12: Meta-analysis of the effect of exercise on global functioning. Effect sizes are

Figure S13: Meta-analysis of the effect of exercise on global functioning. Effect sizes are

grouped by the type of exercise applied in the individual studies.

Figure S14: Meta-analysis of the effect of exercise on depression. Effect sizes are grouped by

the type of control patients as included in the individual studies.

Figure S15: Meta-analysis of the effect of exercise on depression. Effect sizes are grouped by

the type of exercise applied in the individual studies.

Figure S16: Meta-analysis of the effect of exercise on cognitive subdomain attention &

ex-ecutive functioning. Effect sizes are grouped by the type of control patients as included in the individual studies.

Figure S17: Meta-analysis of the effect of exercise on cognitive subdomain processing speed.

Effect sizes are grouped by the type of control patients as included in the individual studies. * With additional cognitive training.

Figure S18: Meta-analysis of the effect of exercise on cognitive subdomain working memory.

Effect sizes are grouped by the type of control patients as included in the individual studies.

* With additional cognitive training.

Figure S19: Meta-analysis of the effect of exercise on cognitive subdomain long-term memory.

Effect sizes are grouped by the type of control patients as included in the individual studies. * With additional cognitive training.

Figure S20: Meta-analysis of the effect of exercise on cognitive subdomain long-term memory.

Figure S21: Meta-analysis of the effect of exercise on cognitive subdomain attention &

execu-tive functioning. Effect sizes are grouped by the type of exercise applied in the individual studies.

Figure S22: Meta-analysis of the effect of exercise on cognitive subdomain processing speed.

Effect sizes are grouped by the type of exercise applied in the individual studies.

Figure S23: Meta-analysis of the effect of exercise on cognitive subdomain working memory.