Studies general characteristics

Supplementary Table 1 provides a summary of the 31 reviews included. These studies includes 33141 participants. Of these, 23 evaluated psychological therapies, 5 exercise therapy, 2 music therapy and 1 multiple approaches. The number of studies included in each review ranged between 2 and 95 with an average of 15.1 studies (median, 10). The objectives of the reviews varied with eight focusing specifically on negative symptoms^{Reference Velthorst, Koeter, van der Gaag, Nieman, Fett and Smit16,Reference Cramer, Lauche, Klose, Langhorst and Dobos30–Reference Cella, Preti, Edwards, Dow and Wykes36} and the rest investigating negative symptoms alongside other mental and physical health outcomes. Twenty-two reviews evaluated a single intervention approach while nine compared multiple therapeutic approaches.^{Reference Lutgens, Gariepy and Malla31,Reference Burlingame GMS, Hoppe, Hunt and Rosendahl37–Reference Jones, Hacker, Meaden, Cormac, Irving and Xia45}

Population

Participants number, used for the estimation of negative symptom treatment effect, ranged from 67 to 2878 and the average participants number was 872.1. Age ranged from 18 to 78 years. Most of the studies did not distinguish participants based on their illness stage (e.g. early or chronic phase). One study included only individuals with treatment-resistant schizophrenia.^{Reference Polese, Fornaro, Palermo, De Luca and de Bartolomeis42}

Interventions

Intervention approaches are described in Supplementary Table 1. Six reviews considered forms of exercise therapy including yoga, meditation and tai-chi,^{Reference Cramer, Lauche, Klose, Langhorst and Dobos30,Reference Sabe, Sentissi and Kaiser32} anaerobic and aerobic exercise,^{Reference Sabe, Kaiser and Sentissi33} mind–body exercise and resistance training.^{Reference Lutgens, Gariepy and Malla31,Reference Vogel, van der Gaag, Slofstra, Knegtering, Bruins and Castelein34} Studies generally defined exercise as any activity aimed at improving or maintaining physical fitness.^{Reference Firth, Cotter, Elliott, French and Yung46}

Two studies evaluated the effect of music therapy,^{Reference Geretsegger, Mössler, Bieleninik, Chen, Heldal and Gold47,Reference Jia, Liang, Yu, Lu, Wang and Wu48} considered as an intervention aiming to promote health in the context of music experience.^{Reference Geretsegger, Mössler, Bieleninik, Chen, Heldal and Gold47} The studies considered included active (e.g. music making) and receptive music (e.g. music listening techniques) therapy methods.

Psychological therapies included CBT interventions,^{Reference Velthorst, Koeter, van der Gaag, Nieman, Fett and Smit16,Reference Jauhar, McKenna, Radua, Fung, Salvador and Laws49} social skills training,^{Reference Kurtz and Mueser50,Reference Turner, McGlanaghy, Cuijpers, Van Der Gaag, Karyotaki and MacBeth51} acceptance and mindfulness-based approaches^{Reference Cramer, Lauche, Haller, Langhorst and Dobos38,Reference Jansen, Gleeson, Bendall, Rice and Alvarez-Jimenez39} and family interventions.^{Reference Rodolico, Bighelli, Avanzato, Concerto, Cutrufelli and Mineo52}

Three examined specifically cognitive remediation,^{Reference Cella, Preti, Edwards, Dow and Wykes36,Reference Turner, van der Gaag, Karyotaki and Cuijpers43,Reference Lejeune, Northrop and Kurtz53} two studies reported on group therapies,^{Reference Burlingame GMS, Hoppe, Hunt and Rosendahl37,Reference Orfanos, Banks and Priebe41} two reported on family-based therapy,^{Reference Lutgens, Gariepy and Malla31,Reference Ma, Chan, Chien, Bressington, Mui and Lee54} and one study focused on integrated neurocognitive therapy.^{Reference De Mare, Cantarella and Galeoto55}

Three studies compared multiple psychological approaches^{Reference Lutgens, Gariepy and Malla31,Reference Riehle, Böhl, Pillny and Lincoln35,Reference Turner, van der Gaag, Karyotaki and Cuijpers43} with Turner et al^{Reference Turner, van der Gaag, Karyotaki and Cuijpers43} reporting on multiple therapies including befriending, CBT, cognitive remediation, psychoeducation, social skills training and supportive counselling.

Control group characteristics

Most exercise therapy studies included treatment-as-usual (TAU) or waitlist as a control condition (see Supplementary Table 1); three studies also considered active control conditions, for example Cramer et al,^{Reference Cramer, Lauche, Klose, Langhorst and Dobos30} Vogel et al^{Reference Vogel, van der Gaag, Slofstra, Knegtering, Bruins and Castelein34} and Riehle et al.^{Reference Riehle, Böhl, Pillny and Lincoln35} For studies evaluating CBT and mindfulness-based interventions, control conditions ranged from active controls, non-active controls, waitlist, TAU or standard care. Two studies conducted different meta-analytic comparisons of psychological approaches, with each intervention being compared with the other interventions pooled.^{Reference Turner, van der Gaag, Karyotaki and Cuijpers43,Reference Jones, Hacker, Meaden, Cormac, Irving and Xia45} The reviews evaluating group therapies considered active control groups (psychotherapeutic treatments), passive control groups (waitlist controls, attention control) and TAU.^{Reference Burlingame GMS, Hoppe, Hunt and Rosendahl37,Reference Orfanos, Banks and Priebe41} For social skills training the majority of the control conditions were active controls and TAU.^{Reference Kurtz and Mueser50,Reference Turner, McGlanaghy, Cuijpers, Van Der Gaag, Karyotaki and MacBeth51} For cognitive remediation the control groups were TAU alone and TAU with the addition of an active control.^{Reference Cella, Preti, Edwards, Dow and Wykes36} All the studies evaluating family intervention had TAU as their control group.

Outcome measures

The most frequent negative symptoms outcome measure was the Positive and Negative Syndromes Scale (PANSS, 96.7%),^{Reference Kay, Fiszbein and Opler56} followed by the Scale for Assessment of Negative Symptoms (SANS; 43.4%)^{Reference Andreasen57} and the Brief Psychiatric Rating Scale (BPRS, 36.6%).^{Reference Velligan, Prihoda, Dennehy, Biggs, Shores-Wilson and Crismon58} Other measures used included the Brief Symptom Inventory,^{Reference Derogatis and Spencer59} Negative Symptoms Assessment,^{Reference Alphs, Morlock, Coon, van Willigenburg and Panagides60} Brief Negative Symptom Scale^{Reference Kirkpatrick, Strauss, Nguyen, Fischer, Daniel and Cienfuegos61} and Comprehensive Psychopathological Rating Scale.^{Reference Asberg and Schalling62}

Exercise interventions

Results of five of the meta-analyses suggested that exercise interventions had a significant and positive effect in reducing negative symptoms, with a small-to-medium effect size. One review did not report a significant effect of exercise on negative symptoms^{Reference Cramer, Lauche, Klose, Langhorst and Dobos30} (see Table 1).

Table 1 Meta-analyses results on the effect of exercise interventions on negative symptoms

g, Hedges g; AC, any comparator; RE, random effects; SMD, standardised mean difference; TAU, treatment-as-usual.

Two reviews evaluated the effect of the intervention compared with TAU and active controls separately. Cramer et al^{Reference Cramer, Lauche, Klose, Langhorst and Dobos30} did not find a significant effect for yoga when compared with either TAU or active controls.^{Reference Cramer, Lauche, Klose, Langhorst and Dobos30} Lutgens et al^{Reference Lutgens, Gariepy and Malla31} reported that the beneficial effect of exercise on negative symptoms was observed when the comparison group was TAU but there was no significant difference between exercise and active controls. Vogel et al^{Reference Vogel, van der Gaag, Slofstra, Knegtering, Bruins and Castelein34} also reported that exercise interventions were only effective when compared with TAU.

Three of the reviews considered mind–body exercises. One review evaluated yoga.^{Reference Cramer, Lauche, Klose, Langhorst and Dobos30} Two reviews included several mind–body exercise approaches including tai-chi, yoga and mindfulness.^{Reference Sabe, Sentissi and Kaiser32,Reference Vogel, van der Gaag, Slofstra, Knegtering, Bruins and Castelein34} Subgroup analyses revealed a small but significant effect favouring yoga, with high heterogeneity, which disappeared when studies at high risk of bias were excluded.^{Reference Sabe, Sentissi and Kaiser32}

Sabe et al^{Reference Sabe, Kaiser and Sentissi33} found that the effect of physical exercise on negative symptoms was driven by aerobic exercise (standardised mean difference (SMD) = −0.31, 95% CI −0.54 to −0.09; P < 0.01) but a different review showed this result may be affected by high risk of bias.^{Reference Vogel, van der Gaag, Slofstra, Knegtering, Bruins and Castelein34}

Cramer et al^{Reference Cramer, Lauche, Klose, Langhorst and Dobos30} showed no short-term benefits (i.e. 12 weeks after randomisation) on negative symptoms when comparing yoga to exercise or TAU.^{Reference Cramer, Lauche, Klose, Langhorst and Dobos30} Sabe et al^{Reference Sabe, Sentissi and Kaiser32,Reference Sabe, Kaiser and Sentissi33} reported intervention lengths between 14.5 and 20.8 weeks, however, Vogel et al^{Reference Vogel, van der Gaag, Slofstra, Knegtering, Bruins and Castelein34} showed that intervention length does not influence efficacy.

Music therapy

The two studies that examined music therapy showed moderate effect sizes in reducing negative symptoms compared with TAU (see Table 2). In a study by Geretsegger et al,^{Reference Geretsegger, Mössler, Bieleninik, Chen, Heldal and Gold47} music therapy combined with standard care showed a significant effect after intervention on negative symptoms when compared with TAU (SMD = −0.55, 95% CI −0.87 to −0.24, P < 0.001). Jia et al^{Reference Jia, Liang, Yu, Lu, Wang and Wu48} found that music therapy was able to reduce negative symptoms compared with TAU (SMD = −0.61, 95% CI −0.80 to −0.42, P < 0.05) and that interventions lasting >3 months were more effective in negative symptom reduction compared with those lasting < 3 months. Lutgens et al^{Reference Lutgens, Gariepy and Malla31} also evaluated arts-based treatments (including music therapy), which had no effect on negative symptoms, but in a sensitivity analysis they found a moderate effect of music therapy on reducing negative symptoms compared with TAU.

Table 2 Meta-analyses results on the effect of psychological interventions on negative symptoms

AC, any comparator; d = Cohen's d; FE, fixed effects; g = Hedges’ g; MD, mean difference; OPT, other psychological therapies; RE, random effects; SMD, standardised mean difference; TAU, treatment-as-usual.

a. SMD calculated see Supplementary Appendix 4 for details.

b. Review presented negative value as favouring control condition.

c. Review presented positive value as favouring experimental condition.

CBT

The effect of CBT was examined in ten studies (see Table 2). Overall, these studies showed small-to-moderate effects of CBT in reducing negative symptoms when compared with TAU and no differences between CBT and other psychological treatments.

Lutgens et al^{Reference Lutgens, Gariepy and Malla31} and Wykes et al^{Reference Wykes, Steel, Everitt and Tarrier63} showed CBT to be more effective compared with TAU with small effect sizes. The results by Velthorst et al^{Reference Velthorst, Koeter, van der Gaag, Nieman, Fett and Smit16} and Wykes et al^{Reference Wykes, Steel, Everitt and Tarrier63} suggest that these effects are smaller and non-significant in studies with lower risk of bias. Jones et al^{Reference Jones, Hacker, Meaden, Cormac, Irving and Xia45} found significant negative symptom reductions for CBT compared with TAU in studies using the PANSS as the outcome measure based on short- (up to 24 weeks), medium- (24–52 weeks) and long-term (over 52 weeks) follow-up periods, and no effect for studies using the SANS (only short-term follow-up available). Sarin et al^{Reference Sarin, Wallin and Widerlov64} found a small, approaching significance, effect at follow-up (3–15 months after treatment) but no immediate post-treatment effect.

In RCTs focusing on treatment-resistant schizophrenia, Polese et al^{Reference Polese, Fornaro, Palermo, De Luca and de Bartolomeis42} did not find an effect of CBT compared with TAU. In RCTs including only patients with elevated negative symptoms before therapy, Riehle et al^{Reference Riehle, Böhl, Pillny and Lincoln35} found a small, approaching significance, effect of negative symptom reduction for CBT compared with TAU. Secondary outcome analyses showed a small-to-moderate significant effect on motivational negative symptoms and no effect on expressive negative symptoms.

In meta-analyses comparing CBT with active controls or other psychological therapies, no study found a significant difference for the reduction of negative symptoms immediately post-treatment.^{Reference Lutgens, Gariepy and Malla31,Reference Riehle, Böhl, Pillny and Lincoln35,Reference Turner, van der Gaag, Karyotaki and Cuijpers43,Reference Jones, Hacker, Meaden, Cormac, Irving and Xia45,Reference Jauhar, McKenna, Radua, Fung, Salvador and Laws49} Sarin et al^{Reference Sarin, Wallin and Widerlov64} found a small effect favouring CBT over other psychological treatments at follow-up.

Social skills training

All the four reviews evaluating social skills training^{Reference Lutgens, Gariepy and Malla31,Reference Turner, van der Gaag, Karyotaki and Cuijpers43,Reference Kurtz and Mueser50,Reference Turner, McGlanaghy, Cuijpers, Van Der Gaag, Karyotaki and MacBeth51} found it to be effective in reducing negative symptoms. Turner et al^{Reference Turner, van der Gaag, Karyotaki and Cuijpers43} also conducted analyses excluding studies with high risk of bias and showed this finding to be robust (g = −0.32, P < 0.05).

Acceptance and mindfulness-based

Five reviews reported on acceptance and mindfulness-based therapies. Four of these found significant small-to-moderate effects for reducing negative symptoms when comparing mindfulness-based therapies with TAU with^{Reference Jansen, Gleeson, Bendall, Rice and Alvarez-Jimenez39,Reference Hodann-Caudevilla, Díaz-Silveira, Burgos-Julián and Santed65} and without an active control condition.^{Reference Tonarelli, Pasillas, Alvarado, Dwivedi and Cancellare66,Reference Liu, Li and Hsiao67} No difference with TAU was found in a smaller meta-analysis by Cramer et al.^{Reference Cramer, Lauche, Klose, Langhorst and Dobos30}

Group therapies

Group therapy was investigated in three reviews.^{Reference Lutgens, Gariepy and Malla31,Reference Burlingame GMS, Hoppe, Hunt and Rosendahl37,Reference Orfanos, Banks and Priebe41} The results showed a small and significant overall effect of group therapies compared with controls post-treatment. Burlingame et al,^{Reference Burlingame GMS, Hoppe, Hunt and Rosendahl37} conducted follow-up analyses showing that group social skills training and cognitive remediation had a small-to-medium significant reduction on negative symptoms (g = 0.23, 95% CI 0.03–0.34, P = 0.03; g = 0.56, 95% CI 0.29–0.84, P < 0.001), respectively.

Cognitive remediation

Three reviews investigated the effect of cognitive remediation.^{Reference Cella, Preti, Edwards, Dow and Wykes36,Reference Turner, van der Gaag, Karyotaki and Cuijpers43,Reference Lejeune, Northrop and Kurtz53} Cella et al^{Reference Cella, Preti, Edwards, Dow and Wykes36} and Lejeune et al^{Reference Lejeune, Northrop and Kurtz53} showed significant small-to-moderate effects for cognitive remediation compared with controls at post-treatment (see Supplementary Table 1). Sensitivity analysis in Cella et al^{Reference Cella, Preti, Edwards, Dow and Wykes36} found that studies with higher methodological quality had a larger effect size compared with those with lower quality. Turner et al,^{Reference Turner, van der Gaag, Karyotaki and Cuijpers43} found no significant difference between cognitive remediation and all other therapies on negative symptoms, and this remained non-significant after sensitivity analyses. Additionally, Riehle et al³⁴ found no difference between cognitive remediation and CBT in reducing negative symptoms in RCTs in which patients had elevated negative symptoms before therapy.

Family-based therapy

The review by Ma et al^{Reference Ma, Chan, Chien, Bressington, Mui and Lee54} found a significant effect of family-based therapies on negative symptoms whereas Lutgens et al^{Reference Lutgens, Gariepy and Malla31} found no effect regardless of the control comparison (see Supplementary Table 1). Long-term effects were not reported in either of the reviews. Rodolico et al^{Reference Rodolico, Bighelli, Avanzato, Concerto, Cutrufelli and Mineo52} found a significant overall effect on negative symptoms of family interventions compared with TAU. Their subgroup analyses with k ≥ 2 indicated superiority of ‘community-based supportive care interventions’, and the combination of family psychoeducation with patient behavioural skills training. Combinations with family behavioural skills training, mutual supportive skill training or emotional climate-focused interventions were not superior to TAU.

Other interventions

The review by De Mare et al,^{Reference De Mare, Cantarella and Galeoto55} reported short-term effects of integrated neurocognitive therapy at 15 weeks post-treatment and showed a significant reduction of negative symptoms (see Supplementary Table 1) with this result maintained at 9–12-month follow-up.

Turner et al,^{Reference Turner, van der Gaag, Karyotaki and Cuijpers43} found no evidence in support of befriending, psychoeducation or supportive counselling being superior to the other interventions pooled. Lutgens et al^{Reference Lutgens, Gariepy and Malla31} found a small beneficial effect for miscellaneous interventions on negative symptoms in comparison with TAU but not for active control.