Repetitive Noninvasive Brain Stimulation to Modulate Cognitive Functions in Schizophrenia: A Systematic Review of Primary and Secondary Outcomes

Abstract

Despite many years of research, there is still an urgent need for new therapeutic options for the treatment of cognitive deficits in schizophrenia. Noninvasive brain stimulation (NIBS) has been proposed to be such a novel add-on treatment option. The main objective of this review was to systematically evaluate the cognitive effects of repetitive NIBS in schizophrenia. As most studies have not been specifically designed to investigate cognition as primary outcome, we have focused on both, primary and secondary outcomes. The PubMed/MEDLINE database (1985–2015) was systematically searched for interventional studies investigating the effects of repetitive NIBS on schizophrenia symptoms. All interventional clinical trials using repetitive transcranial stimulation, transcranial theta burst stimulation, and transcranial direct current stimulation for the treatment of schizophrenia were extracted and analyzed with regard to cognitive measures as primary or secondary outcomes. Seventy-six full-text articles were assessed for eligibility of which 33 studies were included in the qualitative synthesis. Of these 33 studies, only 4 studies included cognition as primary outcome, whereas 29 studies included cognitive measures as secondary outcomes. A beneficial effect of frontal NIBS could not be clearly established. No evidence for a cognitive disruptive effect of NIBS (temporal lobe) in schizophrenia could be detected. Finally, a large heterogeneity between studies in terms of inclusion criteria, stimulation parameters, applied cognitive measures, and follow-up intervals was observed. This review provides the first systematic overview regarding cognitive effects of repetitive NIBS in schizophrenia.

Introduction

Cognitive impairments are core symptoms of schizophrenia, stable over time, they contribute to the debilitating effects, and are one of the main contributor to long-term impairments in social and vocational functioning.^1,2 Several separable cognitive factors have been identified to characterize the global cognitive deficit in schizophrenia: processing speed, attention and vigilance, working memory, verbal learning and memory, visual learning and memory, reasoning and problem solving, and verbal and social cognition.³ Despite the enormous importance of cognitive symptoms, there are still no satisfying treatment options in clinical practice.² The initial hope that second-generation antipsychotics would prove effective for the treatment of cognitive deficits was not fulfilled satisfactorily,² whereas recently one meta-analysis suggests at least a moderate effect.⁴ Especially in recent years, noninvasive brain stimulation (NIBS) techniques have been proposed as new treatment options for cognitive deficits. The term NIBS summarizes various techniques, including repetitive transcranial magnetic stimulation (rTMS) and its variants such as theta burst stimulation (TBS), and transcranial electric stimulation including transcranial direct current stimulation (tDCS), transcranial random noise stimulation, and transcranial alternating current stimulation. More specifically, magnetic and electric stimulation have fundamentally different physiological modes of action. TMS induces neuronal action potentials, whereas the static electric field induced by tDCS does not result in such a rapid depolarization but modulates spontaneous neuronal activity by a tonic depolarization or hyperpolarization.⁵ Common to all is the possibility to induce long-lasting changes in brain excitability, to modulate cortical connectivity and network plasticity.^6,7 One consensus paper⁸ provides evidence for a moderate clinical efficacy of rTMS for the treatment of persistent auditory hallucinations (temporal lobe stimulation) and predominant negative symptoms (frontal lobe stimulation), although recent studies with relevant sample sizes^9,10 are questioning these statements. One tDCS pilot study showed that tDCS can improve both auditory hallucinations as well as negative symptoms in severely affected schizophrenia patients.¹¹ Especially, the improvement of negative symptoms following tDCS deserves particular attention as negative symptoms and cognitive impairments are considered to be correlated.¹² Negative symptoms and cognitive impairments have separable but related etiologies, whereas alternative models describing common or independent etiologies are also deemed as possible.¹² However, a clear dividing line between both concepts cannot be easily drawn challenging the differentiation between both.¹²

In this context, the treatment of schizophrenia-associated cognitive deficits in patients suffering from predominant negative symptoms came early into the focus of interest. In 1999, one of the very first proof-of-concept trials in the field (open-label, N = 6) showed that 2 weeks of 20 Hz rTMS applied to the left dorsolateral prefrontal cortex (DLPFC) improved negative symptoms.¹³ Moreover, the authors also investigated a broad cognitive battery and showed a numeric improvement in various cognitive measures, whereas delayed visual memory achieved significance.¹³

From a principle point of view, 2 effects of NIBS on cognition in schizophrenia are possible. Firstly, a disruptive effect on cognitive functions^14,15 (mainly following temporal lobe stimulation) as a side-effect or, secondly, an improvement of cognitive function as a positive effect^6,16 were discussed. The latter beneficial effects have been linked to impaired DLPFC function and connectivity in schizophrenia¹⁷ and the overall idea is that neurostimulation techniques have the capacity to enhance neuroplasticity and to remodel connectivity.⁶ However, no systematic evaluation of the cognitive effects of the whole range of NIBS techniques in schizophrenia is yet available. One meta-analysis provided by the Cochrane group summarized for the comparison of active and sham rTMS that the cognitive state was reported in 3 studies using 39 different measures, but that the results were equivocal for all measures.¹⁸ However, as only few repetitive NIBS studies used cognitive measures as primary outcomes, but since many studies included cognitive measures as secondary outcome, systematic qualitative analyses are needed. The main objective was to gain a detailed overview on potential positive and negative effects of NIBS. Therefore, we performed a systematic review following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA)¹⁹ requirements and analyzed all cognitive measures reported in the extracted trials. As most available trials focused on persistent positive or negative symptoms, we hypothesized that despite the clinical importance and the clear pathophysiological rational, only few trials will be available that specifically investigate the efficacy of NIBS for the treatment of cognitive impairments. Furthermore, based on the possibility to modulate various stimulation parameters and outcome measures, we hypothesized that a large heterogeneity between studies will limit the generalizability of findings.

Methods

This systematic review was conducted via the internet database PubMed/MEDLINE taking into account recommendations for the standardized reporting of systematic reviews (PRISMA).¹⁹ The data source contained studies published from January 1, 1985 to August 15, 2015 (ENDNOTE X7 filter “entered between [1985/01/01:2015/12/31]”, with the last search conducted on August 15, 2015). We have started in 1985 as TMS was introduced to the field in 1985. Literature was searched manually and via the ENDNOTE X7 search tools using the following terms: “brain stimulation” and “schizophrenia”; “brain stimulation” and “schizophrenia” and “cognition”; “repetitive transcranial magnetic stimulation” and “schizophrenia”; “repetitive transcranial magnetic stimulation” and “schizophrenia” and “cognition”; “repetitive transcranial magnetic stimulation” and “schizophrenia” and “negative symptoms”; “repetitive transcranial magnetic stimulation” and “schizophrenia” and “auditory hallucinations”; “rTMS” and “schizophrenia”; “rTMS” and “schizophrenia” and “cognition”; “rTMS” and “schizophrenia” and “negative symptoms”; “rTMS” and “schizophrenia” and “auditory hallucinations”; “transcranial direct current stimulation” and “schizophrenia”; “transcranial direct current stimulation” and “schizophrenia” and “cognition”; “transcranial direct current stimulation” and “schizophrenia” and “negative symptoms”; “transcranial direct current stimulation” and “schizophrenia” and “auditory hallucinations”; “tDCS” and “schizophrenia”; “tDCS” and “schizophrenia” and “cognition”; “tDCS” and “schizophrenia” and “negative symptoms”; “tDCS” and “schizophrenia” and “auditory hallucinations”; “Theta Burst Stimulation” and “schizophrenia”; “Theta Burst Stimulation” and “schizophrenia” and “cognition”; “Theta Burst Stimulation” and “schizophrenia” and “negative symptoms”; “Theta Burst Stimulation” and “schizophrenia” and “auditory hallucinations”; “TBS” and “schizophrenia”; “TBS” and “schizophrenia” and “cognition”; “TBS” and “schizophrenia” and “negative symptoms”; “TBS” and “schizophrenia” and “auditory hallucinations.”

Results from all search terms were combined and duplicate records were removed using the ENDNOTE X7 deduplication feature. The titles and, if relevant, the abstracts of each citation were screened and the full text of each potentially relevant citation was retrieved and reviewed in detail. Only clinical trials with repetitive stimulation sessions and longitudinal design (controlled, uncontrolled, blinded, open-label) reported in English language were included. Case reports and case series, single-session studies, abstracts of congress presentations, reviews, and meta-analyses were excluded. Furthermore, the recent meta-analyses on persistent auditory hallucinations and negative symptoms,^20,21 as well the latest large-scale clinical trials in the field,^9,10 were also screened for citations not revealed by the systematic literature search. Full-text articles were then analyzed with respect to the presentation of cognitive measures as primary or secondary outcome parameters. This data item search was not restricted to any cognitive measure. Full-text articles from the same research groups were compared regarding the risk of duplicate populations. We further analyzed 2 databases (first: https://clinicaltrials.gov/ and second: http://apps.who.int/trialsearch/) to detect ongoing or planned trials with any cognitive parameters as primary outcome measure. For both databases, we used the terms “rTMS and schizophrenia,” “repetitive transcranial magnetic stimulation and schizophrenia,” “tDCS and schizophrenia,” and “transcranial direct current stimulation and schizophrenia.”

Results

Study Selection

On the basis of our search strategy, 2231 unique records could be identified. A total of 2155 records were excluded after screening of titles, abstracts, and article format reviews, resulting in 76 full texts that could then be retrieved. Of these 76 full texts, 5 were not detected through our search strategy^22–26 and included via other sources and targeted manual search. These 76 full-text articles were then analyzed regarding the presentation of any cognitive measures before and after NIBS intervention. In total, 33 primary studies published between 1999 and 2015^{9,13–15,22,23,26–52} were included for further analyses in this systematic review. All other full texts (N = 43) were excluded due to the lacking presentation of cognitive measures. Of these 43 studies not reporting cognitive measures, 28 focused on persistent auditory hallucinations, 10 on predominant negative symptoms, and 5 on other primary outcome measures (clinical global impression, affective symptoms, global symptoms). Please see figure 1 for the PRISMA¹⁹ flow diagram.

Fig. 1.

PRISMA diagram summarizing the flow of information through all phases of this systematic review.

Study Characteristics

Among the 33 studies included in this systematic review, 4 studies reported cognitive measures as primary outcome.^40,44,47,52 Out of these 4 studies, one used the Cambridge Neuropsychological Test Automated Battery (CANTAB),⁴⁰ one focused on working memory using a silent phonemic Verbal Fluency Task,⁴⁴ one focused also on working memory using a verbal n-back task,⁴⁷ and one used the MATRICS Consensus Cognitive Battery (MCCB).⁵² The remaining 29 studies had global symptoms (N = 12),^{13,27–30,38,39,41–43,48,49} negative symptoms (N = 11),^{9,22,26,32,34–37,45,50,51} or persistent auditory hallucinations (N = 6)^{14,15,23,31,33,46} as primary outcomes. All studies had a pre/post design and participants were at rest during the stimulation.

Apart from 3,^9,26,50 all publications reported single-centre results. The mean sample size across all publications (inclusion of all 33 studies regardless of a possible overlap of sample across publications [see table 1]) was 33.21 (SD: 28.03) with a minimum of N = 4 and a maximum of N = 157. Twenty-seven studies had a double-blind sham-controlled parallel group design and 6 studies were open-label studies. Thirty studies used rTMS (including deep TMS and TBS) and 3 studies used tDCS as noninvasive stimulation techniques. No studies were found using other transcranial electric stimulation methods. Regarding the localization of the respective stimulation sites, 17 studies used the electroencephalogram (EEG) 10/20 method, 12 studies used the 5-cm-anterior method (for DLPFC localization), 4 studies used neuronavigation based on individual structural MRI, and 1 study did not report the applied procedure. Regarding rTMS sham procedures, 10 trials tilted the coil away from the scalp at an angle of 45°,^{9,14,15,23,26,27,29,31,45,46} 7 trials tilted the coil at an angle of 90°,^{30,34,36,43,47,50,51} 6 trials used a specific sham coil,^{33,35,37,38,44,49} and 1 study did not describe the sham procedure.²² However, as the latter study used the Brainsway H1 deep-TMS coil,²² one could speculate that the built-in sham mode has been used. For tDCS, 3 trials used different variations of the ramp-up/ramp-down procedure,^41,48,52 but no study used the “built-in study-mode.” The stimulation parameters varied across studies dependent on the target symptoms resulting in a high interstudy heterogeneity for this parameter. Most studies investigated cognitive functioning immediately after the respective stimulation intervention (1–4wk). Only few studies included follow-up measures after the end of intervention. The longest follow-up periods were 8–12 weeks (for details, see table 1).

Table 1.

Cognitive Effects of Repetitive NIBS in Schizophrenia

Reference	Study Design; Total Sample Size; Follow-up Period; Study Population	NIBS Parameters; NIBS Device and Stimulation Site; Sham Parameters	Cognitive Outcome Measures; Primary Outcome Yes/No	Summary of Results (Cognitive Measures)
Cohen et al13	Open-label pilot study (N = 6); before and immediately after 2wk; chronic schizophrenia patients	20 Hz rTMS for 2s; once per minute for 20min at 80% MT for 2wk (5d/wk); MagPro magnetic stimulator with figure- of-eight coil; EEG 10/20; coil was placed tangential to the orbital area on the C3 and C4 point	Wechsler Adult Intelligence Scale, TMT A and B, the FAS Verbal Fluency Test, and 2 subtests of the WMS (the visual memory reproduction and the verbal paired associates subtests); cognitive measures were secondary outcomes	Numeric improvement in all cognitive measures, whereas only delayed visual reproduction achieved significance
Rollnik et al27	Double blind; sham-controlled crossover study (N = 12); before and after 1 and 2wk; chronic schizophrenia patients	20×2s; 20 Hz stimulations with 80% of MT intensity over 20min for 2wk (800 pulses/session; 10 sessions at 10 working days); Magstim Rapid stimulator with figure-of-eight-coil; coil over the DLPFC of the dominant hemisphere (5cm anterior M1); for sham rTMS, the coil was tilted at an angle of 45°	NCT; cognitive measures were secondary outcomes	NCT tended to improve during active rTMS and to worsen during sham stimulation without reaching significance
D’Alfonso et al28	Open-label trial (N = 8); before and after 1 and 2wk; medication-resistant schizophrenia patients	1 Hz rTMS for 20min for 2wk (10 sessions at 10 working days) with 80% MT; Neopulse magnetic stimulator; EEG 10/20; coil was placed approximately 2cm above T3 (no coil specifications described)	Auditory Imagery Test, Rey Auditory Verbal Learning Test, Token Test (short form), verbal fluency and phoneme detection, Judgment of Line Orientation, Line Bisection Test, Benton Visual Retention Test, and the Test for Facial Recognition; short form; cognitive measures were secondary outcomes	Only the auditory imagery tests revealed a significant improvement over time, whereas only 6 patients had a complete dataset for this test. All other cognitive measures did not show a significant difference before and after intervention
Hoffman et al15	Double blind; sham-controlled parallel group study (N = 24); before and after 9 d; medication-resistant AH patients	1 Hz rTMS for 16min for day 3–9 (day 1: 8min; day 2: 12min) with 90% MT; Magstim Super Rapid System with figure- of-eight coil; EEG 10/20 halfway between T3 and P3; for sham rTMS, the coil was tilted at an angle of 45°	California Verbal Learning Test, Controlled Oral Word Association Test, Semantic Fluency, Digit Distraction Task, WRAT-R, TMT A and B, Grooved Pegboard, Digit Symbol Task, Temporal Orientation, HVLT, Letters- Number Span Test (the latter 2 were administered serially during the trial); cognitive measures were secondary outcomes	No differences in cognitive measures over time apart from a marginal significant time × treatment effect for the HVLT
Huber et al29	Re-evaluation of Rollnik et al24 ; chronic schizophrenia patients	See Rollnik et al26	NCT; cognitive measures were secondary outcomes	Women (N = 4) improved in the NCT performance, whereas men (N = 8) did not improve during active rTMS
McIntosh et al23	Double blind; sham-controlled crossover study (N = 16); before and after 1wk; medication-resistant AH patients	1 Hz rTMS for 4min at day 1, 8min at day 2, 12min at day 3, and 16min until the end of 2wk with 80% MT; each minute of stimulation was followed by 15s of rest; Dantec magnetic stimulator with figure-of- eight coil; EEG 10/20 halfway between T3 and P3; for sham rTMS, the coil was tilted at an angle of 45°	AVLT; cognitive measures were secondary outcomes	No differences in AVLT
Holi et al30	Double blind; sham-controlled parallel group study (N = 22); before and after 2wk; chronic patients	10 Hz rTMS with 20 trains (each of 5s; 30s apart) with 100% MT; Magstim stimulator with figure-of-eight coil; left DLPFC (5cm anterior left M1); for sham rTMS, the coil was tilted at an angle of 90°	MMSE; cognitive measures were secondary outcomes	No change in the MMSE scores within either group
Fitzgerald et al31	Double blind; sham-controlled parallel group study (N = 33); before and after 2wk; medication-resistant AH patients	1 Hz rTMS for 15min for 10 working days with 90% MT; Magstim Super Rapid System with figure-of-eight coil; EEG 10/20 T3; for sham rTMS, the coil was tilted at an angle of 45°	HVLT (also assessed after the first session), Visuospatial Digit Span; cognitive measures were secondary outcomes	No changes in cognitive functions across time and groups. For the HVLT, a small but significant deterioration in performance was reported in both study groups
Sachdev et al32	Open-pilot study (N = 4); before and after 4wk; predominant negative symptoms (deficit syndrome)	15 Hz rTMS with 24 trains (each of 5s; 25s apart) with 90% MT; Magstim Super Rapid System with figure-of-eight coil; coil over the DLPFC of the left hemisphere (5cm anterior M1)	MMSE, Digit Forward and Backward, TMT A and B Symbol Digit Coding, Verbal Fluency for letter and category, and WCST; cognitive measures were secondary outcomes	No changes in any of the cognitive measures over time
Hoffman et al14	Double blind; sham-controlled parallel group study (N = 50; 27 active rTMS); before and after 9 working days; medication-resistant AH patients	1 Hz rTMS for 16min for day 3–9 (day 1: 8min; day 2: 12min) with 90% MT; Magstim Super Rapid System with figure- of-eight coil; EEG 10/20 halfway between T3 and P3; for sham rTMS, the coil was tilted at an angle of 45°	California Verbal Learning Test, Controlled Oral Word Association Test, Animal Naming, Digit Recall Task (nondistraction and distraction conditions), WRAT-R Reading Test, TMT A and B, Grooved Pegboard (dominant and nondominant), Digit Symbol Task, Temporal Orientation, HVLT, Letters-Number Span Test (the latter 2 were administered serially during the trial); cognitive measures were secondary outcomes	No differences between active and sham rTMS could be detected for the full neuropsychological battery before and after intervention. One patient receiving active rTMS and one receiving sham rTMS were removed from the trial due to drops in the Hopkins Verbal Learning Task
Brunelin et al33	Double blind; sham-controlled parallel group study (N = 24; 14 active rTMS); before and after 5 working days; medication-resistant AH patients	1 Hz rTMS for 17min twice/d for 5 working days with 90% MT; Medtronic MagPro with figure-of-eight coil; EEG 10/20 halfway between T3 and P3; for sham rTMS, a specific sham coil was used	Source Memory Tasks (discrimination between silent- and overt-reading words); cognitive measures were secondary outcomes	Active rTMS resulted in improvements in source monitoring. Reduction of AHs showed a trend-wise correlation with the improvement in the number of total confusion during source monitoring tasks
Novák et al34	Double blind; sham-controlled parallel group study (N = 16); before and after 8wk; predominant negative symptoms	20 Hz rTMS with 40 trains (each of 2.5s; 30s apart; 2000 stimuli) with 90% MT; Magstim Super Rapid System with figure- of-eight coil; coil over the DLPFC of the left hemisphere (5cm anterior M1); for sham rTMS, the coil was tilted at an angle of 90°	AVLT, TMT A and B, Rey- Osterrieth Complex Figure, CPT; cognitive measures were secondary outcomes	Between-group comparisons did not show any differences for any of cognitive measures after a follow-up of 8 wk
Mogg et al35	Double blind; sham-controlled parallel group study (N = 17; 8 active rTMS); before and after 2 and 4wk; predominant negative symptoms	10 Hz rTMS with 20 trains (each of 10s; 50s apart; 2000 stimuli) with 110% MT; Magstim Super Rapid System with figure- of-eight coil; coil over the DLPFC of the left hemisphere (5cm anterior M1); for sham rTMS, a specific sham coil was used	Controlled Oral Word Association Test, Stroop interference task, HVLT, Grooved Pegboard Test; cognitive measures were secondary outcomes	For the HVLT-delayed recall, there was a significant difference between groups in favor of active rTMS at 2-wk follow-up. For the Stroop Test, there was a trend towards a group effect in favor of active rTMS
Fitzgerald et al36	Double blind; sham-controlled parallel group study (N = 20); before and after 3wk; treatment-resistant negative symptoms	10 Hz rTMS with 20 trains (each of 5s; 25s apart; 1000 stimuli) with 110% MT; 20 trains were applied to each hemisphere at every session; Medtronic MagPro 30 magnetic stimulators with figure-of-eight coils; coils over the left and right DLPFC (5cm anterior M1); for sham rTMS, the coils were tilted at an angle of 90°	Stroop Test, Controlled Oral Word Association Test, TMT A and B; cognitive measures were secondary outcomes	There were no significant overall or between-group effects of rTMS on any of the cognitive measures
Schneider et al37	Double blind; sham-controlled parallel group study (N = 51; 34 active rTMS); before and after 2, 4, and 8wk; predominant negative symptoms	10 Hz rTMS or 1 Hz rTMS with 20 trains (each of 5s; 15s apart; 1000 stimuli) with 110% MT; 20 trains were applied for 4 consecutive weeks; Neotonus 1000 magnetic stimulator; coil not specified; coil over the left DLPFC (5cm anterior M1); for sham rTMS, one of 2 head covers on the magnet was fitted by a person not involved in the stimulation	WCST; cognitive measures were secondary outcomes	The 10 Hz group showed significant improvement at weeks 4 and 8 on the failure to maintain set and perseverative response errors; but between- group differences did not reach significance. The 1 Hz group demonstrated significant improvement on WCST only at week 2. The results of this cognitive measure were biased by a high noncompletion rate in all study groups
Mittrach et al38	Double blind; sham-controlled parallel group study (N = 32; 18 active rTMS); before and after 2wk; chronic schizophrenia patients	10 Hz rTMS with 20 trains (each of 5s; 55s apart; 1000 stimuli) with 110% MT; 20 trains were applied to each hemisphere at every session; Medtronic MagPro X100 magnetic stimulators with figure-of-eight coils; coil over the left DLPFC (5cm anterior M1); for sham rTMS, a specific sham coil was used	Kurztest für Allgemeine Intelligenz (German version of a short test of general intelligence), the Mehrfach- Wortwahl-Test (German verbal intelligence test), D2 attention task, TMT A and B, WCST; cognitive measures were secondary outcomes	No significant effect of rTMS on cognitive measures could be observed
Demirtas- Tatlidede et al39	Open-label pilot study (N = 8); before and immediately after 5 d; treatment- resistant schizophrenia patients	Intermittent TBS (3 pulses at 50-Hz repeated at a rate of 5 Hz; 20 trains of 10 bursts given with 8-s intervals; 600 pulses) with 100% active MT were applied twice a day to the cerebellar vermis; Medtronic MagPro X100 magnetic stimulators with figure-of-eight coil; MR navigation based on individual MRI	Auditory CPT, Letter-Number Span, WMS, 3rd Edition (WMS- III)-Spatial Span, Phonemic/Letter Fluency, Category Fluency, BACS- Symbol Coding, TMT A and B, WCST, Delis-Kaplan Executive Function System, California Verbal Learning Test, Rey-Osterrieth Complex Figure Test, Grooved Pegboard; cognitive measures were secondary outcomes	A significant improvement over time could be detected for the CPT (memory and interference condition), for the spatial span forward performance, and for the organization score of the Rey-Osterrieth Complex figure. No negative effects of the intervention were shown
Levkovitz et al40	Open-label pilot study (N = 15); before and after 2 and 4wk; predominant negative symptoms	20 Hz Deep-TMS (42 trains of 2s; 40 pulses each; with 20-s intertrain interval) with 120% MT was applied once a day for 20 d; Magstim Super Rapid stimulator connected to an H1 coil; coil over the left DLPFC (5cm anterior M1)	CANTAB with 4 domains: psychomotor speed (RTI), visuospatial memory (spatial recognition memory), sustained attention (RVP), and executive functions (SOC, SWM, spatial span, and intra-/extradimensional shift); cognitive measures were together with the SANS scale primary outcomes	A significant improvement in RVP and SOC (problems solved), SWM (between errors), SWM (strategy), and spatial span length could be observed
Mattai et al41	Double blind; sham-controlled pilot study (N = 12); before and after 2wk; childhood-onset schizophrenia	Bilateral anodal DLPFC stimulation (n = 8) or bilateral cathodal STG stimulation (n = 5); 20-min stimulation once a day for 10 working days; Phoresor II Auto Model PM850 with 25cm2 sponge electrodes; EEG 10/20 FP1 and FP2 for left and right DLPFC and T3 for STG – reference electrode at nondominant forearm; for sham tDCS; the current was turned on for 1min and then ramped down	MMSE, California Verbal Learning Test, WMS; cognitive measures were secondary outcomes	No changes in cognitive measures in favor of any stimulation condition could be observed
Oh et al42	Open-label study (N = 10); before and after 1 and 4wk; treatment-resistant schizophrenia patients	10 Hz rTMS with 20 trains (each of 3s; 30s apart) to the left DLPFC and 1 Hz rTMS with 20 trains (each of 30s; 30s apart) to the left TPC with 80%–100% MT; Magstim Rapid stimulator with figure-of-eight coil; EEG 10/20 F3 for left DLPFC and P3-T3 midpoint for TPC	RKMT (K-AVLT), Complex Figure Test, Stroop Test, WCST; cognitive measures were secondary outcomes	Short-term auditory verbal memory improved over time. All other cognitive measures did not show a significant change over time
Blumberger et al43	Double blind; sham-controlled parallel group study (N = 51; 34 active rTMS); before and after 4wk; medication-resistant AH patients	1 Hz rTMS for 20min with 115% MR; primed 1 Hz rTMS was administered by 10min of 6 Hz rTMS (20 trains; 5s each; 25s apart) with 90% MT followed by 10-min 1 Hz rTMS with 115% rTMS; Medtronic Repetitive Magnetic Stimulator with figure-of-eight coil; MR navigation based on individual MRI; for sham rTMS, the coil were tilted at an angle of 90°	RBANS; cognitive measures were secondary outcomes	No difference in RBANS change could be observed across groups and between responders and nonresponders
Prikryl et al44	Double blind; sham-controlled parallel group study (N = 30; 19 active rTMS); before and after 3wk; predominant negative symptoms	10 Hz rTMS with 15 trains (each of 10s; 30s apart; 1500 stimuli) with 110% MT; Magstim Super Rapid stimulator with figure-of-eight coil; left DLPFC localization was not described; for sham rTMS, a specific sham coil was used	Silent Phonemic VFT during fMRI; cognitive measures were primary outcomes	No differences in the VFT performance could be observed between active and sham rTMS
Guse et al45	Double blind; sham-controlled parallel group study (N = 25; 13 active rTMS); before and after 3wk; predominant negative symptoms; subgroup of Wobrock et al7	10 Hz rTMS with 20 trains (each 30s apart; 1000 stimuli) with 110% MT; 20 trains were applied to each hemisphere at every session; Medtronic MagPro X100 magnetic stimulators with figure- of-eight coil; EEG 10/20 F3 for DLPFC localization; for sham rTMS, the coil was tilted at an angle of 45°	Verbal n-back WM task with parametrically varied WM load among 0, 1, and 2 items during fMRI; TMT A and B; divided attention (TAP); WCST; cognitive measures were secondary outcomes of the initially randomized population of schizophrenia patients7	No changes in cognitive performance in favor of any stimulation condition could be observed
Hoffman et al46	Double blind; sham-controlled parallel group study (N = 83; 55 active rTMS); before and after 3wk (third stimulation block); medication- resistant AH patients	1 Hz rTMS for 16min with 90% MT; Magstim Rapid-2 System with figure- of-eight coil; MR navigation based on individual MRI; Wernicke’s area and the right homologue; for sham rTMS, the coil was tilted at an angle of 45°	Hopkins verbal memory and letter- numbers working tasks (baseline and after the 3rd, 8th, and 13th session), full Neuropsychological Test Battery (baseline and during the third stimulation block), 2 laterality tasks at baseline: dichotic listening task and right-left pegboard difference; cognitive measures were secondary outcomes	As outlined in the “Results” section of the respective publication, aggregate neuropsychological data did not reveal any significant alterations, either improvements or declines—when contrasting rTMS with sham stimulation
Barr et al47	Double blind; sham-controlled parallel group study (N = 27; 13 active rTMS); before and after 4wk; chronic schizophrenia patients	20 Hz rTMS with 25 trains (each 30s apart; 1500 stimuli [750 pulses/ hemisphere]) with 90% MT; 20 trains were applied to each hemisphere at every session; Medtronic MagPro magnetic stimulators with figure-of-eight coil; MR navigation based on individual MRI; for sham rTMS, the coil was tilted at an angle of 90°	n-back WM task at 1- and 3-back conditions; cognitive measures were primary outcome	Active rTMS improved 3-back accuracy to targets compared with sham. Active rTMS normalized the performance in this condition to the level of a healthy control group. No effect on 1- and 2-back as well on correct responses to nontargets and on RTI in responses to both targets and nontargets could be observed
Fitzgerald et al48	Double blind; sham-controlled parallel group pilot studies (2 pilot studies in 1 publication) (total N = 24); before and after 3wk; medication-resistant AH patients and nonresponsive negative symptoms	tDCS (2 mA; 20min; ramp up of 120s and ramp down of 15s); 15 daily sessions over 3 consecutive weeks; neuroConn Eldith Stimulator Plus; EEG 10/20; unimodal tDCS: anode F3 and cathode TP3; bimodal tDCS: anode F3 and F4 and cathode TP3 and TP4; sham stimulation: ramp up of stimulation and 30s of stimulation prior to stimulation off set	Forwards and backwards digit span, block spatial span tests, N-back task, Tower of London planning task, FAS VFT, TMT A and B; cognitive measures were secondary outcomes	No changes in cognitive measures in favor of any stimulation condition could be observed
Wölwer et al49	Double blind; sham-controlled parallel group study (N = 32; 18 active rTMS); before and after 2wk; chronic schizophrenia patients (overlap to Mittrach et al35 )	10 Hz rTMS with 20 trains (each of 5s; 55s apart; 1000 stimuli) with 110% MT; Medtronic MagPro X100 magnetic stimulators with figure-of-eight coil; coil over the left DLPFC (5cm anterior M1); for sham rTMS, a specific sham coil was used	Facial affect recognition task during fMRI, Mehrfach- Wortwahl-Test (German verbal intelligence test), D2 attention task, TMT A and B, WCST; cognitive measures were secondary outcomes	Active rTMS was superior to sham rTMS in improving the facial affect recognition performance. No changes in the other cognitive measures could be observed
Rabany et al22	Double blind; sham-controlled parallel group study (total N = 30; 20 active deep TMS); before and after 4 and 5wk; predominant negative symptoms	20 Hz Deep-TMS (42 trains of 2s; 40 pulses each; with 20-s intertrain interval) with 120% MT was applied once a day for 20 d; Magstim Super Rapid stimulator connected to an H1 coil; coil over the left DLPFC (5.5cm anterior M1); the technique of sham stimulation was not described	CANTAB: psychomotor speed: MOT and RTI, visuospatial memory: PRM, sustained attention: RVP, executive functions: SOC and SWM; cognitive measures were secondary outcomes	Apart from one SOC measure (subsequent times for 5 move problems at week 4), no changes in cognitive measures in favor of any stimulation condition could be observed
Dlabac-de Lange50	Double blind; sham-controlled parallel group study (total N = 32); before and after 3wk and at 4-wk follow-up; predominant negative symptoms	10 Hz rTMS with 20 trains (each of 10s; 50s apart; 2000 stimuli) with 90% MT; 20 trains were applied to each hemisphere at every session; Medtronic MagPro X100 magnetic stimulators with figure-of-eight coils; EEG 10/20; coils over the left (F3) and right (F4) DLPFC; for sham rTMS, the coil was tilted at an angle of 90°	Verbal Learning Test, Digit Symbol Substitution Test, TMT A and B, Dutch version of the Rey AVLT, Verbal Fluency Test, WCST; cognitive measures were secondary outcomes	Semantic verbal fluency improved significantly in the active group (n = 10) compared with the sham group (n = 9) up to 4-wk follow-up. No changes in the other cognitive measures could be observed
Dlabac-de Lange51	Double blind; sham-controlled parallel group study (total N = 24); before and after end of intervention; predominant negative symptoms (overlap to Dlabac-de Lange et al46 )	10 Hz rTMS with 20 trains (each of 10s; 50s apart; 2000 stimuli) with 90% MT; 20 trains were applied to each hemisphere at every session; Medtronic MagPro X100 magnetic stimulators with figure-of-eight coils; EEG 10/20; coils over the left (F3) and right (F4) DLPFC; for sham rTMS, the coil was tilted at an angle of 90°	Tower of London task during fMRI; cognitive measures were secondary outcomes	Brain activity in the active group increased in the right DLPFC and the right medial frontal gyrus. Further analyses revealed decreased activation in the active and increased activation in the sham group in the left posterior cingulate. Behavioral data did not show a change over time between groups
Wobrock et al9	Double blind; sham-controlled parallel group study (N = 157; 76 active rTMS); before and after 3, 6, and 12wk; predominant negative symptoms	10 Hz rTMS with 20 trains (each 30s apart; 1000 stimuli) with 110% MT; 20 trains were applied to each hemisphere at every session; Medtronic MagPro X100 magnetic stimulators with figure- of-eight coil; EEG 10/20 F3 for DLPFC localization; for sham rTMS, the coil was tilted at an angle of 45°	TMT A and B; cognitive measures were secondary outcomes	No differences in cognitive measures in favor of any stimulation condition could be observed
Smith et al52	Double blind; sham-controlled parallel group study (N = 33; 17 active tDCS); before and after 5 d or 1wk, 6wk, and 12wk; chronic patients	tDCS (2 mA; 20min; ramp up/ramp down not described); 5 sessions over 5 consecutive work days); tDCS device manufacturer not described; EEG 10/20; anode F3 and cathode FP2 (contralateral supraorbital ridge); sham stimulation: 40s with 2 mA	MCCB: composite score and the following domains MCCB domains: WM, attention- vigilance, reasoning/problems, verbal learning, visual learning, processing speed, social cognition; cognitive measures were primary outcomes	Active tDCS was superior to sham tDCS in improving MCCB with the greatest improvement in the composite score and the domain scores for WM and attention-vigilance
Hasan et al26	Double blind; sham-controlled parallel group study (N = 156; 77 active rTMS); before and after 3, 6, and 12wk; predominant negative symptoms (overlap to Wobrock et al7 )	See Wobrock et al7	Rey AVLT, TMT A and B, WCST, Digit Span Test, Regensburg Word Fluency Test, neuropsychological composite score; cognitive measures were secondary outcomes	No differences in cognitive measures in favor of any stimulation condition could be observed. However, effect sizes indicate a numeric but nonsignificant superiority of active rTMS in certain cognitive tests

Note: AH, auditory hallucinations; AVLT, Auditory Verbal Learning Test; BACS, Brief Assessment of Cognition in Schizophrenia; CANTAB, Cambridge Neuropsychological Test Automated Battery; CPT, Continuous Performance Test; DLPFC, dorsolateral prefrontal cortex; EEG, electroencephalogram; FAS, Verbal Fluency Test (uses the letters F, A, and S); fMRI, functional magnetic resonance imaging; HVLT, Hopkins Verbal Learning Test; K-AVLT, Korean versions of the Auditory Verbal Learning Test; MCCB, MATRICS Consensus Cognitive Battery; MMSE, Mini-Mental State Examination; MOT, Motor Screening Task; MT, motor threshold; NCT, Number Connection Test; NIBS, noninvasive brain stimulation; PRM, pattern recognition memory; RBANS, Repeatable Battery for the Assessment of Neuropsychological Status; RKMT, Rey-Kim Memory Test; RTI, reaction time; rTMS, repetitive transcranial magnetic stimulation; RVP, rapid visual information processing; SANS, Scale for the Assessment of Negative Symptoms; SOC, Stockings of Cambridge; STG, superior temporal gyrus; SWM, spatial working memory; TAP, Test of Attentional Performance; TBS, theta burst stimulation; tDCS, transcranial direct current stimulation; TMT, Trail Making Test, version A or B; TPC, temporoparietal cortex; VFT, Verbal Fluency Task; WCST, Wisconsin Card Sorting Test; WM, working memory; WMS, Wechsler Memory Scale; WRAT-R, Wide Range Achievement Test-Revised.

Most studies required a stable medication regime during the intervention^{9,13–15,26–32,34–45,49–51} and 6 studies did not include this information.^{22,23,33,47,48,52} With respect to the use and abuse of legal and illegal drugs, a large heterogeneity across studies could be observed. Ten studies did not describe whether legal or illegal drugs were permitted or not.^{23,27–30,32,33,37,41,52} All other studies^{9,13–15,22,26,31,34,36,38,40,42–46,50,51} described drug abuse or drug dependency as exclusion criterion. In more detail, legal and illegal drug dependency within 2 years,^38,49 5 years,¹³ or in general^{14,22,31,36,40,44,46,48} was an exclusion criterion in many studies. Other studies excluded the abuse/dependency 4 weeks,¹⁵ 3 months,³⁴ and 6 months prior to study inclusion.^{11,15,19,23,24,30} Finally, some studies^9,26,42,45 excluded a current drug abuse/dependency at the time of study start.

Regarding new trials, our database search revealed 3 trials investigating the efficacy of rTMS/TBS for the treatment of working memory deficit (NCT01880255, CTRI/2015/01/005448, and ChiCTR-TRC-13004017), 2 trials focusing on social cognition (NCT02440867 and NCT02479919), and 1 trial focusing on overall cognition according to BACS (Brief Assessment of Cognition in Schizophrenia) (NCT02131129). For tDCS, 3 trials are focusing on overall cognition measured by the MCCB (NCT02539797, NCT01971073, and NCT02085421), 2 trials are aiming to improve general cognition without further specification (NCT01733602 and NCT01602263), and 1 has working memory as primary outcome (ACTRN12612000217808).

Risk of Bias Within and Across Studies

The main bias within and across studies was that almost all studies did not access cognitive performance as primary outcome. Therefore, most studies were not designed and powered to detect differences between active and sham intervention for these parameters. Furthermore, the limited sample sizes and the monocentric design of most studies question the generalizability of the presented positive and negative findings regards cognition. A possible additional bias could be the procedures used to localize the target areas for NIBS. Various different methods, including the application of different stimulation parameters, were used across studies resulting in a high interstudy heterogeneity and hampering comparability.

Results of Individual Studies (Outcomes)

The cognitive measures used in the different studies were dependent on various factors including theoretical deliberations and/or pathophysiological hypotheses as well as safety questions. Studies with a stimulation focus on the frontal lobe followed the idea of improving cognition through neurostimulation (cognition as outcome parameter). As most studies did not investigate cognition as primary outcome and as the applied cognitive measures are subject to a high interstudy variability, no clear pattern across studies could be observed. Studies with a double-blind design that had cognitive measures as secondary outcomes could principally not overserve a beneficial effect of active stimulation compared to sham stimulation.^{9,22,26,27,30,34,36,41,45} Other studies^33,35,37,50 indicate a potential benefit of active neurostimulation applied to the DLPFC on verbal memory, source monitoring, cognitive flexibility, or verbal fluency, whereas the aforementioned negative studies failed to show a benefit in the same or related domains. Interestingly, one study using high-frequency rTMS applied to the left DLPFC focused on the safety effect of frontal neurostimulation and did not reveal any potential harmful effect on cognition.³⁸ The largest study investigated the efficacy of 10 Hz rTMS applied to the left DLPFC (EEG 10/20 – F3) on various cognitive domains in schizophrenia patients with predominant negative symptoms and did not show a superiority of active compared to sham rTMS. However, the reported effect sizes point towards a numeric, but nonsignificant superiority of the active intervention (positive effect sizes favoring active rTMS: global composite z score [0.28], Trail Making Test [TMT] domain z score [0.27], Verbal Learning and Memory Test [VLMT] domain z score [0.24], Wisconsin Card Sorting Test [WCST] domain z score [0.24]).²⁶ Studies focusing on cognition as primary outcome revealed a slightly different pattern. Three of the 4 studies^27,31,34,39 had a double-blind sham control and parallel group design and of these 3 studies, 2 used rTMS and 1 used tDCS. Barr et al⁴⁷ applied 20 Hz rTMS with 1500 stimuli to the left and right DLPFC (neuronavigated rTMS) for a treatment duration of 4 weeks. The DLPFC was targeted by individual structural MRI scans, 27 patients were randomized and the primary outcome was the mean magnitude of change in the n-back accuracy for target responses (verbal working memory). Active rTMS improved 3-back target accuracy to a level comparable with the performance of a healthy control group in contrast to sham rTMS (performance correct target responses: active rTMS: pre: 40.06% ± 18.57%, post: 49.86% ± 23.50%; sham rTMS: pre: 36.02% ± 24.34%, post: 33.33% ± 20.34%). However, no effect on 1- and 2-back, as well on correct responses to nontargets and on reaction time in responses to both targets and nontargets was observed.⁴⁷ Prikryl et al⁴⁴ applied 10 Hz rTMS with 1500 stimuli to the left DLPFC for a treatment duration of 3 weeks. The authors did not describe the procedure to detect the left DLPFC and analyses did neither show differences in the performance of the silent phonemic Verbal Fluency Task (working memory) nor in fMRI activation between groups. Smith et al⁵² applied tDCS with 2 mA (EEG 10/20: anode F3, cathode FP2) for 5 days and used the MCCB performance as primary outcome. Active tDCS was superior to sham tDCS with regard to the overall MCCB performance (composite score, effect size: 1.03) and with regard to the domain scores for working memory (effect size: 1.25) and attention vigilance (effect size: 0.84). Finally, Levkovitz et al showed in an open-label pilot study of 15 schizophrenia patients with predominant negative symptoms that 20 Hz Deep-TMS (target region left DLPFC, 5cm anterior left M1) improved rapid visual information processing, problem solving, spatial working memory, and spatial span length⁴⁰ (see table 1).

Cognition as safety parameter was accessed in studies with a stimulation focus on the temporal lobe of the dominant hemisphere. Various cognitive measures were evaluated before and after intervention following possible concerns that temporal stimulation might disrupt cognitive functioning. In this context, no evidence was actually revealed that temporal stimulation disrupts cognitive functioning.^{14,15,23,28,31,43,46}

Discussion

This is the first systematic review to investigate the impact of repetitive NIBS on different cognitive domains in schizophrenia patients. Following a standardized literature search and analysis, 76 full-text articles were assessed for eligibility and of these article, 33 studies were included in the qualitative synthesis. Remarkably, only 4 studies had different cognitive measures as primary outcomes making clear that the available evidence is very limited and that previous studies focused on the improvement of clinical symptoms. Different reviews are discussing possibilities and hypotheses to treat neurocognitive deficits in schizophrenia with NIBS (eg, Lett et al⁶, Demirtas-Tatlidede et al¹⁶), but until today the data basis for clear conclusions is still sparse. In principle, 2 different areas of interest need to be distinguished: (1) the possibility to enhance cognitive functions with NIBS (efficacy) and (2) the risk to disrupt cognitive functions with NIBS (safety).

At this point, no definite statement can be made regarding the efficacy of NIBS to improve cognitive functions in schizophrenia. The different trials used vast numbers of cognitive tests, different study populations, stimulation parameters, and stimulation sites, which makes comparisons and overall statements rather complicated. Moreover, in most studies, cognitive measures were only secondary outcomes, offering a limited amount of information on the efficacy of NIBS. Despite the situation that most studies failed to provide a beneficial effect of the respective intervention, there are at least some indications that active rTMS can improve visual and verbal memory as well as working memory and source monitoring (see table 1 for stimulation parameters and stimulation targets). From all rTMS trials assessing cognition as primary outcome, the pilot trial presented by Barr et al⁴⁷ provides the best evidence for a possible efficacy for the treatment of working memory impairments in schizophrenia. Whether the detected improvement in one cognitive domain (verbal 3-back task) is really clinically relevant is still an unanswered question. However, further trials with a focus on patients with clearly defined cognitive deficits as well as clinically meaningful cognitive outcome measures are needed to confirm these initial promising findings. Interestingly, one pilot tDCS study (EEG 10/20, anode F3 and cathode FP2)⁵² used the MCCB, the gold standard for the assessment of cognitive functioning in schizophrenia trials as primary outcome measure. After 5 days of tDCS (anode: left DLPFC), improvements in the MCCB composite score as well as in the domain scores of working memory and attention-vigilance in the active compared to the sham group could be observed.⁵² In terms of efficacy, one important question is the relationship between symptomatic and cognitive improvement. In antipsychotic trials, a clear relationship could not be established (for a more detailed discussion, see Bora et al⁵³) and the same question has to be raised in NIBS studies. One critical factor remains that almost every second NIBS trial in schizophrenia did not show a specific effect of the intervention of the targeted psychopathology. However, NIBS differs fundamentally in the underlying modes of actions compared to antipsychotics. Therefore, one could speculate that based on the possibility to modulate plasticity and improve neural synchrony, NIBS have the potential be a pathophysiological-orientated treatment specifically for cognitive impairments.⁶

The aforementioned safety discussion has been raised in the context of the first rTMS trials for the treatment of persistent auditory hallucinations through inhibitory stimulation of the left temporal lobe.^14,15 On the basis of these different trials, our analysis indicates no increased risk to worsen cognitive functions (especially short-term verbal memory, semantic processing, and clustering of verbal information)^14,15,46 with NIBS. However, one should note that many different trials focusing on the treatment of persistent auditory hallucinations did not include cognitive measures and that no single trial had been powered to detect differences in side-effects between interventions. Therefore, this safety statement is limited to the trials (and therefore stimulation configurations) summarized in this systematic review. In this context, only the 2009 published TMS safety guidelines⁵⁴ discuss these safety issues following repetitive sessions of therapeutic TMS for all indications and conclude that unequivocal unintentional cognitive deterioration during the observation periods has not been noted in any studies.⁵⁴ For tDCS, not enough data from clinical trials in schizophrenia is available to provide a statement regarding this safety question. The here analyzed tDCS trials^41,48,52 did not show any evidence for a cognitive deterioration following the intervention.

One important limitation of all analyzed trials are the limited follow-up periods. Most trials assessed cognitive function before and immediately after the intervention and only few trials had a subsequent follow-up period. The longest follow-up period covered 3 months which already needs to be considered as too short to evaluate changes in cognitive functioning. A network meta-analysis analyzing 9 randomized controlled antipsychotic trials provides evidence that differences in cognitive performance between interventions can be reliably observed after a delay of 6 months.⁴ Therefore, longer follow-up periods in studies investigating the efficacy of NIBS for the modulation of cognitive functioning are needed to exclude practice effects, to establish a stable relationship between active intervention and outcome, and to provide clinical meaningful information.

Several methodological aspects need to be considered in interventional NIBS trials to improve cognitive impairments in schizophrenia. First, the effect of placebo/sham and thus the limitations of uncontrolled trials need to be critically discussed. Interventional trials in schizophrenia are subject to high placebo rates and causal factors are, eg, expectancy bias, rater reliability, and data quality.⁵⁵ The settings of a clinical trial providing social and emotional support for study patients by the research team day by day might be another explanation.⁹ Moreover, one recently published meta-analysis (21 rTMS studies) showed that the effect size for sham rTMS is 0.29 for the treatment of auditory verbal hallucinations.⁵⁶ These high placebo rates question the promising findings from uncontrolled NIBS trials.⁴⁰ Second, cognitive measures are highly sensitive to practice effects. Several factors like exposure, familiarity, and procedural learning have been discussed to bias the measures cognitive performance in longitudinal schizophrenia trials.⁵⁷ Therefore, lacking differences between active and sham stimulation in many of the discussed trials may be explained by such practice effects. Especially tests applied frequently in clinical settings have a high risk for practice effects. Future trials need must give special attention to this bias by, eg, using several baselines before intervention, applying test with parallel versions or using tests not common in clinical settings. One randomized controlled antipsychotic trial showed that cognitive tests with simple directions, large numbers of trials, and a restricted set of stimuli had only limited practice effects.⁵⁷ Such tests might be more suitable to detect significant differences across time between active and sham stimulation.

The present systematic review has the following possible limitations: The scope of this review was restricted to cognitive measures (primary and secondary outcomes) and as 43 of the 76 (57%) identified full-text articles did not report these outcome parameters, the generalizability of our findings is limited. Furthermore, we have analyzed exclusively the cognitive outcome without taking into consideration the clinical outcome of the selected trials. It is likely that an intervention which failed to show differences in the primary outcomes (eg, improvement in auditory hallucinations, improvements in predominant negative symptoms) will consecutively fail to show differences in secondary outcomes (here: cognition). Further limitations are the exclusion of non-English literature, case reports, and single-session studies. However, since the most relevant literature is now published in English, to miss relevant studies by using this restriction is therefore rather unlikely. We have deliberately excluded case reports as this format will only be accepted for publication if positive results are reported. The exclusion of single-session studies (experiments) from the qualitative literature syntheses can be explained by the fact that these studies do not follow the strict designs and reporting requirements of interventional clinical trials. However, from single-session experiments, important information how NIBS modulates cognition in schizophrenia can be obtained. One single-session experiment applied 20 minutes of anodal tDCS with 2 mA to the left DLPFC of 20 schizophrenia patients (verum vs sham) and could not show that anodal tDCS improves probabilistic association learning but increases the variance indicating that a subgroup of patients had a response to one stimulating session.⁵⁸ Another single-session experiment applied tDCS oscillating at the frequency of the sleep slow oscillation to the left and right DLPFC in 14 schizophrenia patients during sleep and showed a greater verbal retention capacity following active tDCS compared to sham tDCS.⁵⁹ Concerning working memory, one single active tDCS session (20min, 2 mA, anode left DLPFC [F3], cathode right supraorbital) was more effective in improving n-back performance (d-prime) compared to sham tDCS and to 1 mA stimulation after 20 and 40 minutes, but not immediately after stimulation.⁶⁰ Another experimental study showed that active 20 Hz rTMS reduced frontal gamma oscillatory activity elicited during working memory performance (n-back) in 24 schizophrenia patients but enhanced activity in healthy controls.⁶¹ To investigate the impact of single-session tDCS (2 mA, left and right DLPFC) on social cognition, one experiment randomized 12 patients to either anodal, cathodal, or sham tDCS. Anodal stimulation resulted in an improvement in facial emotion identification, but not in managing emotions, perception and theory of mind (videotape test).⁶² These single-session experiments highlight the possibility to modulate cognitive functioning in schizophrenia with NIBS and show the possibility to induce at least short-lasting improvements in severely affected patients. The questions remain whether these single-session effects can sustained for a longer period.

In conclusion, the results of this systematic review show that there is as yet no sufficient data available for an exhaustive discussion of the potential of repetitive NIBS to modulate cognitive functions in schizophrenia. Although the recent discussions have focused on a potential beneficial effect of (pre)frontal neurostimulation on cognition (global cognitive capacity, working memory) in schizophrenia, a clear benefit of this intervention could, despite some promising patterns, not be established. However, the positive findings of 2 studies with cognitive outcome as primary outcome^47,52 and the reported effect sizes in the largest randomized controlled trial²⁶ are in the first instance encouraging and highlight the importance of further studies specifically designed to investigate this research question. Moreover, more research is needed to replicate the positive findings, to determine the most promising stimulation techniques and stimulation parameters, and to detect other moderators impacting the outcome (eg, antipsychotic treatment, legal and illegal drug abuse/dependency, stage-dependency, attention, age). Fortunately, there are many new trials ongoing focusing on different cognitive measures as primary outcome. While the ongoing rTMS/TBS trials are again subject to a large interstudy variability in terms of methodology and cognitive outcome domains, the tDCS trials are more homogenous and have mainly defined the MCCB as standardized outcome criterion.

From the current available literature, one could tentatively conclude for rTMS that long stimulation periods (4wk or even more) with high frequencies (eg, 20 Hz) are necessary. There is no evidence for the initially suspected potential cognitive disruptive effect (verbal memory, semantic procession, verbal clustering) of rTMS applied to the temporal lobe. For tDCS, more studies are needed to confirm the initial positive findings and to allow for a discussion regarding electrode placement, stimulation parameters, intervals, and total number of stimulations. Finally, more multicentric trials with sufficient statistical power and strict designs are essential to allow for guideline-relevant clinical recommendations.