Nutraceuticals and phytoceuticals in the treatment of schizophrenia: a systematic review and network meta-analysis

Background: Sub-optimal response in schizophrenia is frequent, warranting augmentation strategies over treatment-as-usual (TAU). Methods: We assessed nutraceuticals/phytoceutical augmentation strategies via network meta-analysis. Randomized controlled trials in schizophrenia/schizoaffective disorder were identified via the following databases: PubMed, MEDLINE, EMBASE, Scopus, PsycINFO, CENTRAL, and ClinicalTrials.gov. Change (Standardized Mean Difference=SMD) in total symptomatology and acceptability (Risk Ratio=RR) were co-primary outcomes. Secondary outcomes were positive, negative, cognitive, and depressive symptom changes, general psychopathology, tolerability, and response rates. We conducted subset analyses by disease phase and sensitivity analyses by risk of bias and assessed global/local inconsistency, publication bias, risk of bias, and confidence in the evidence. Results: The systematic review included 49 records documenting 50 studies (n=2,384) documenting 22 interventions. Citicoline (SMD=−1.05,95%CI=−1.85; −.24), L-lysine (SMD=−1.04,95%CI=−1.84;−.25), N-acetylcysteine (SMD=−.87,95%CI=−1.27;−.47) and sarcosine (SMD=−.5,95%CI=−.87−.13) outperformed placebo for total symptomatology. High heterogeneity (tau2=.10, I2=55.9%) and global inconsistency (Q=40.79, df=18, p=.002) emerged without publication bias (Egger’s test, p=.42). Sarcosine improved negative symptoms (SMD=−.65, 95%CI=−1.10; −.19). N-acetylcysteine improved negative symptoms (SMD=−.90, 95%CI=−1.42; −.39)/general psychopathology (SMD=−.76, 95%CI=−1.39; −.13). No compound improved total symptomatology within acute phase studies (k=7, n=422). Sarcosine (SMD=−1.26,95%CI=−1.91; −.60), citicoline (SMD=−1.05,95%CI=−1.65;−.44), and N-acetylcysteine (SMD=−.55,95%CI=−.92,−.19) outperformed placebo augmentation in clinically stable participants. Sensitivity analyses removing high-risk-of-bias studies confirmed overall findings in all phases and clinically stable samples. In contrast, the acute phase analysis restricted to low risk-of-bias studies showed a superior effect vs. placebo for N-acetylcysteine (SMD=−1.10,95%CI=−1.75,−.45), L-lysine (SMD=−1.05,95%CI=−1.55,−.19), omega-3 fatty acids (SMD=−.83,95%CI=−1.31,−.34) and withania somnifera (SMD=−.71,95%CI=−1.21,−.22). Citicoline (SMD=−1.05,95%CI=−1.86,−.23), L-lysine (SMD=−1.04,95%CI=−1.84,−.24), N-acetylcysteine (SMD=−.89,95%CI=−1.35,−.43) and sarcosine (SMD=−.61,95%CI=−1.02,−.21) outperformed placebo augmentation of TAU (“any phase”). Drop-out due to any cause or adverse events did not differ between nutraceutical/phytoceutical vs. placebo+TAU. Conclusions: Sarcosine, citicoline, and N-acetylcysteine are promising augmentation interventions in stable patients with schizophrenia, yet the quality of evidence is low to very low. Further high-quality trials in acute phases/specific outcomes/difficult-to-treat schizophrenia are warranted.


Introduction
Schizophrenia and related disorders affect 23,6 million people worldwide, accounting for a signi cant burden 1,2 .People with schizophrenia often require long-term treatment with complex pharmacotherapy 3 , especially in treatment-resistant cases 4 .Nevertheless, up to 40% of patients fail to respond to standard antipsychotic treatment 5 .While signi cant research is ongoing, currently, no novel mechanism-of-action medications for schizophrenia have been approved besides postsynaptic dopamine receptor antagonism 6 .Hence, even small additional bene ts from safe augmentation strategies are of potential value 7 .Although 42 different pharmacologic augmentation strategies of antipsychotic agents were assessed in a previous umbrella review, the results were inconclusive due to the limited quality of the meta-analyzed studies and signi cantly greater effect sizes in lower-quality studies 8 .Nutraceuticals and phytoceuticals are increasingly used among non-pharmacological treatments due to their favorable safety and tolerability 9,10 .Yet, conclusive results are needed to establish their potential role in treating schizophrenia 11 .Historically, the term nutraceutical has had various de nitions 12,13 .
Per previous umbrella reviews in this eld, we refer to any nutrient-based intervention, such as vitamins, minerals, amino acids, and fatty acids 14 .We also de ne phytoceuticals as any intervention that uses plant-based compounds such as herbal formulations 15 .
Existing meta-analytic evidence 14,15 and the most current World Federation of Societies of Biological Psychiatry (WFSBP) and Canadian Network for Mood and Anxiety Treatments (CANMAT) clinical guidelines 11 inform about the e cacy, safety, and tolerability of different nutraceuticals and phytoceuticals in the treatment of schizophrenia.However, no network meta-analysis (NMA) has been conducted to assess these compounds' e cacy, acceptability, and tolerability in schizophrenia directly, except for a recent Bayesian NMA appraising selected compounds 16 .Therefore, the present NMA aimed to appraise the e cacy, acceptability, and tolerability of multiple nutraceuticals and phytoceuticals, either as mono-or augmentation therapy, in treating schizophrenia.

Methods
The present systematic review (SR) and NMA evaluated nutraceutical and phytoceutical interventions in schizophrenia using randomized controlled trials (RCTs).The study followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension statement for network meta-analysis (PRISMA-NMA) 17 .The protocol was registered via PROSPERO (CRD42022375946).

Search strategy and selection criteria
We systematically searched PubMed/MEDLINE, EMBASE, Scopus, PsycINFO, Cochrane CENTRAL, and ClinicalTrials.govfrom journal inception to October 17th, 2023.No language restriction was applied.The following search terms were adopted and augmented across different databases: psychosis, schizo*, nutraceuticals, phytoceuticals, and randomized controlled trials.The appendix (S1) details the search strategy.The electronic search was supplemented by a manual search of the reference list of all retrieved trials and relevant SRs and meta-analyses to identify additional relevant RCTs.Two authors (MB and MF) independently screened the papers and extracted data.Any disagreement was resolved by consensus with a third author (CC).
Inclusion criteria were: i) RCT; ii) inclusion of any nutraceutical/phytoceutical treatment, either as monotherapy or augmentation of treatment as usual (TAU), compared to either another nutraceutical/phytoceutical compound, placebo, TAU, or placebo + TAU; iii) involving adult (age ≥ 18 years) in-or outpatients with a primary diagnosis of schizophrenia or schizoaffective disorder, according to the Diagnostic and Statistical Manual of Mental Disorders (DSM), any edition/text revision, or International Classi cation of Disease (ICD), any edition; iv) providing quantitative data about the change in disease-speci c symptomatology, measured through adequate rating scales, treatment response, all-cause drop-outs (acceptability) and, intolerability-related drop-outs (tolerability).
Excluded were studies with i) observational or non-randomized design; ii) without a suitable control group (e.g., studies comparing two different doses of the same nutraceutical/phytoceutical molecule); iii) involving < 10 patients per arm, consistent with most WFSBP/CANMAT guidelines 11 ; iv) crossover design not providing pre-cross-over data (otherwise included in the SR portion only); v) special populations, like rst-episode psychosis (since a de nitive diagnosis of schizophrenia is unclear), high-/ultra-high risk populations, pediatric populations (age < 18 years); vi) unstrati ed results for different diagnostic populations (e.g., schizophrenia and bipolar disorder merged altogether).

Study selection, extraction, and outcomes
The following data were extracted: author(s), year, nutraceutical/phytoceutical and control interventions, mean dose of the active intervention, study design (parallel-group or crossover), trial duration, sample size (total, cases, controls), mean age and % of females in each arm, diagnostic criteria and (semi)structured interview, acceptability, tolerability, e cacy measures as reported in RCTs for the primary outcome (mean difference in scoring at endpoint from baseline, response rate).We also recorded whether the outcomes were reported per intention-to-treat (ITT) (i.e., last observation carried forward [LOCF]) or per-protocol (i.e., completers).All outcomes were measured at the study endpoint.For crossover RCTs, only the pre-cross-over effect was considered.The following were deemed co-primary outcomes: i) e cacy, as the mean change in the total score of schizophrenia-speci c rating scales, and ii) acceptability.Secondary outcomes were: i) mean change in positive, negative, cognitive, and depressive symptoms and general psychopathology, ii) intolerability-related drop-out, and iii) treatment response.Contact with study authors was attempted when data were unavailable.

Risk of bias, con dence in the evidence, and quality appraisal
The risk of bias was assessed using the Cochrane Risk of Bias tool, second edition (RoB2) 18 .Con dence in the evidence for the primary outcomes was evaluated within the Con dence In Network Meta-Analysis (CINeMA) framework 19 .Finally, an adapted version of the AMSTAR plus content total score 8 was employed to rate the quality of the included RCTs, weighted for the number of comparisons (score range = 0-7, owing to the following adapted items: 1) Blindness yes/no (score range 0-1); 2) sample size (n < 20 score = 0, n = 20-100 score 1, n > 100 score = 2); 3) ndings replicated in at least one comparison (1 point) and n > 20 in at least one study arm (1 point); 4) ITT (no = 0, yes = 1); 5) High (score = 0) or low risk (score = 1) of bias at the Rob2 regressed against the primary e cacy outcome effect size (change in total PANSS score: SMD, 95%CI).

Data analysis
We performed a random-effect NMA within the frequentist framework using the netmeta package v.2.1-0, RStudio version 4.2.1.We computed standardized mean difference (SMD) for continuous outcomes and risk ratio (RR) for binary ones.For studies failing to report the mean changes but only baseline and endpoint data, the SMD and standard deviation (SD) were estimated for each study arm using the methods outlined in the Cochrane Handbook 20 .
The homogeneity of direct evidence (within each pairwise comparison) was assessed using tau 2 .Global inconsistency was evaluated by considering a fulldesign-by-treatment model.Local inconsistency was measured with a node-splitting approach to assess the agreement between direct and indirect estimates for each outcome.The study node was the nutraceutical/phytoceutical molecule.Publication bias was measured for each outcome by visually examining the funnel plot and Egger's test for studies with ≥ 10 participants 21 .
The relative ranking of different nutraceutical/phytoceutical interventions was estimated using the surface under the cumulative ranking curves (SUCRA).The higher the SUCRA value, the better the rank of the intervention within the accounted series of interventions.
The following subgroup analyses were planned for e cacy outcomes: acute vs. maintenance phase schizophrenia and clozapine vs. non-clozapine antipsychotics.The following sensitivity analyses were planned for the co-primary outcomes: i) excluding studies with a high risk of bias and ii) focusing on schizophrenia studies only.

Results
The systematic search yielded 12,744 results, reduced to 11,754 after 990 duplicates were automatically detected and removed.Title and abstract screening led to 135 records eligible for the full-text screening.After the full-text assessment, 50 records were included in the SR  . One rcord provided two independent studies 34 .An exhaustive list of excluded studies with reasons is reported in the appendix, S8. Th PRISMA ow diagram is summarized in Fig. 1.
All included studies compared nutraceutical/phytoceutical compounds + TAU to placebo + TAU.Three studies did not enter the NMA synthesis: two 37,39 due to the lack of TAU and one 22 due to unreliable TAU (Table 1).
All included studies were parallel-group RCTs, except two crossover design studies 22,68 .The median number of participants was n = 60 (interquartile range, n = 40), range, n = 20-157).Most studies ranged from four to 24 weeks, except for three that lasted 26 weeks 36 , 52 weeks 32 , and two years 39 , respectively.The latter 39 was a maintenance study involving patients who discontinued the antipsychotic treatment, undergoing randomization to omega-3 + α-lipoic acid or placebo, thus not being appropriate for the quantitative synthesis.
Most of the included studies were conducted in Asia (29 studies), followed by North America (12 studies), Europe ( ve studies), Africa (two studies), and Australia (two studies).
In all included studies, the e cacy of nutraceuticals or phytoceutical compounds was assessed through the score on the Positive and Negative Syndrome Scale (PANSS) 71 , Scale for the Assessment of Positive Symptoms (SAPS) 72 , Scale for the Assessment of Negative Symptoms (SANS) 73 , and the Clinical Global Impression-Improvement module (CGI-I) 74 , different cognitive batteries, and functioning scales.
Most studies were conducted on patients with chronic schizophrenia, with baseline PANSS total scores ranging from 48 60 to 114 31 .The phase of illness and the setting of the study have been detailed in Table 1.Finally, the lack of corresponding data precluded a subgroup analysis for clozapine studies.
A qualitative synthesis of the results of all included RCTs is reported in Table 1.The main results are summarized in Figs.2-4.
Regarding the risk of bias, 11 studies (22% of the records) were rated as having some concerns, 27 studies as low-risk (54%), and 12 as high risk of bias (24%) (e-Table 1).
Network meta-analyses of nutraceuticals and phytoceuticals as an augmentation to treatment as usual

PANSS total symptomatology
Thirty-six studies with 22 treatments encompassing 2,384 patients were included in the NMA for change in total symptomatology.Speci cally, the following augmentations of TAU relied on one RCT only: cerulysine, citicoline, D-alanine, DHA, folate + Vitamin B12, glycine, L-lysine, mangosteen, sarcosine + sodium benzoate, sarsapogenein, sulforaphane, vitamin D, vitamin D + probiotics, whitania somnifera, and yokukansan.The following augmentations of TAU relied on two RCTs: curcumin and folate.The following augmentations of TAU relied on three RCTs: D-serine.The following augmentations of TAU relied on four RCTs: NAC, EPA, and omega-3.Sarcosine relied on six RCTs.

Positive symptoms
Regarding positive symptoms (measured through PANSS positive subscale or SAPS), 40 studies with 27 treatments encompassing 2,555 patients were included in the main analysis, encompassing all phases of the disease.Speci cally, the following augmentations of TAU relied on one RCT only: cerulysine, citicoline, D-alanine, DHA, folate, folate + Vitamin B12, ginseng, L-lysine, L-theanine, mangosteen, sarcosine + sodium benzoate, sarsapogenein, sulforaphane, vitamin D + probiotics, whitania somnifera, and yokukansan.The following augmentations of TAU relied on two RCTs: curcumin and glycine.The following augmentations of TAU relied on three RCTs: D-serine and ginkgo biloba.The following augmentations of TAU relied on four RCTs: NAC and omega-3.Sarcosine relied on ve RCTs.

Negative symptoms
The following augmentations of TAU relied on one RCT only: cerulysine, citicoline, D-alanine, D-cycloserine, EPA, folate + Vitamin B12, ginseng, L-lysine, Ltheanine, mangosteen, sarcosine + sodium benzoate, sarsapogenein, sulforaphane, vitamin D + probiotics, whitania somnifera, and yokukansan.The following augmentations of TAU relied on two RCTs: curcumin and vitamin D + folate.The following augmentations of TAU relied on three RCTs: glycine.The following augmentations of TAU relied on four RCTs: NAC and omega-3.D-serine and sarcosine relied on ve and six RCTs, respectively.

Acceptability and Tolerability Analysis
Regarding acceptability, 37 studies were included, accounting for 2,636 patients.None of the 24 nutraceutical or phytoceutical augmentation strategies signi cantly differed from the placebo regarding the RR of drop-out due to any cause (e-Figure 29 & e-Table 24).No heterogeneity was found (tau 2 = 0, I 2 = 0%).Global inconsistency was not statistically signi cant (Q = 9.33, df = 15, p = .86).No publication bias was detected (Egger's test, p = .90)(e-Figure 30).Similar ndings emerged from the sensitivity analyses without high risk-of-bias studies (e-Figure 31 & e-Table 25) and with people with schizophrenia only (e-Figure 32 & e-Table 26).
Concerning tolerability (k = 8, n = 747), none of the eight assessed augmentation strategies differed from the placebo regarding the RR of drop-out due to adverse events (e-Figure 33 & e-Table 27).Finally, there was insu cient information to compute the ine cacy (discontinuation due to lack of e cacy), set as a secondary outcome.

Con dence in evidence
The CINeMA con dence in evidence and its methods is summarized in Appendix S6.Regarding both PANSS total symptom and acceptability analyses, con dence in evidence was low/very low throughout all the comparisons (e-Tables 29-30).Likewise, meta-regression analysis consistently documented a trend for lower SMDs as the adapted AMSTAR content score increased across different outcomes (e.g., PANSS total symptoms change = β=-0.0245,95%CI = .0.0866, -0.1943, p = .77,R 2 = .34,N = 22).

Discussion
This NMA concurrently appraised 22 nutraceutical and phytoceutical interventions for adjunctive treatment of schizophrenia and related conditions, expanding the series of compounds already appraised by previous work (although a handful of compounds could nonetheless not be included in the present report due to the stringent inclusion/exclusion criteria) 14,16 .Speci cally, the present NMA also included N-Methyl-D-Aspartate receptor (NMDAR) modulators, such as D-alanine, D-serine, D-cycloserine, glycine, sarcosine (N-substituted glycine), and plant-derived compounds such as curcumin, in addition to vitamins (e.g.Folate, B12, D), N-acetylcysteine, and omega-3 fatty acids, otherwise included in the WFSBP and CANMAT clinical guidelines 11 .
While the NMA showed that most of the appraised nutraceutical and phytoceutical compounds failed to provide a signi cant bene t compared to placebo augmentation of TAU, speci c compounds showed promising bene cial effects as an antipsychotic-augmentation strategy, with favorable acceptability and tolerability pro les.N-acetylcysteine and sarcosine improved total and negative symptoms.PANSS general symptomatology improved with curcumin and Nacetylcysteine.Citicoline improved total symptomatology only.
While mechanistic understanding of these nutraceutical/phytoceutical interventions is ongoing, the e cacy of select agents seen in our NMA is concordant with the currently understood mechanisms of action of these interventions.
Speci cally, translational evidence suggests a relevant role of the nitric oxide (NO) signaling pathway in schizophrenia pathogenesis, affecting glutamate and dopamine storage, uptake, and release, also increasing oxidative stress 75 .In this regard, L-lysine competes with L-arginine for the cationic amino-acid transporter 76 on the blood-brain barrier, thus reducing the in ux of L-arginine and, consequently, the overall arginine-based NO synthesis 76 .Altered regulation of oxidative stress and neuroin ammatory mechanisms may be relevant to the pathogenesis and the progression of schizophrenia and related medical and psychiatric disorders, likely due to membrane and neuronal damage 77,78 .Moreover, a protracted in ammatory state activates indoleamine 2,3-dioxygenase, producing neuroactive metabolites that in uence the dopaminergic and glutamatergic neurotransmission, inducing neurotoxic stimuli 79,80 .Furthermore, an imbalance in glutamate neurotransmission is increasingly relevant in pathogenic theories of schizophrenia 81 .Such neurobiological pathways may explain why the appraised bene cial compounds (i.e., L-lysine, N-acetylcysteine, citicoline, sarcosine, and curcumin) appear to be e cacious in schizophrenia, considering that they share direct/indirect antioxidant, anti-in ammatory and glutamate modulatory activity.
Curcumin exerts antioxidant, neuroprotective, and anti-in ammatory properties 82 , concordant with its bene cial effects on general symptoms.Citicoline is a precursor of acetylcholine and membrane phospholipids with neuroprotective, antioxidant, and anti-excitotoxic properties 83 .Moreover, citicoline may exert anti-in ammatory activity via unclear mechanisms, likely reducing blood-brain barrier permeability 84 .N-acetylcysteine is a glutathione precursor with antioxidant and anti-in ammatory properties 85 and is already recommended for adjunctive treatment in schizophrenia (primarily for negative symptoms) at 1-3 g/day 11 .N-acetylcysteine in uences glutamate neurotransmission by enhancing NMDAR activity 86 , and several studies showed that its supplementation improved the processing speed, attention, working memory, and executive functioning of patients with schizophrenia 87-89 , consistent with results from a recent umbrella review of adjunctive therapies in schizophrenia 90 .
Finally, sarcosine is a glycine transporter type-I inhibitor and NMDAR allosteric modulator, directly binding to the NMDAR glycine co-agonist site, enhancing NMDAR activity, and possessing neuroprotective properties 91,92 .However, previous studies (not included in our study due to our exclusion criteria) suggested that sarcosine may not be effective when used as an augmentation agent with clozapine.This lack of e cacy is likely due to the potential glycinergic and glutamatergic activity of clozapine itself 53,93 , resulting in a "ceiling" effect when additional modulation of NMDA receptors is unlikely 92 .
Hypothetically, a high-in ammatory/oxidative stress subset of people with schizophrenia may bene t from such supplementations.Since oxidation, in ammatory, and genetic biomarkers could guide the strati cation of people with schizophrenia 94,95 , further nutraceutical/phytoceutical RCTs may adopt biomarkers as inclusion criteria to evaluate different e cacy pro les across various subsets 90 .

Limitations of the study
First, the main limitation of the present NMA is the low number of studies for each meta-analyzed agent.For example, for the primary e cacy outcome, for only ve agents, more than one or two studies could be meta-analyzed, i.e., sarcosine: k = 5; and N-acetylcysteine, D-serine, EPA, and omega-3 (k = 4 each).
Second, our ndings rely on a limited number of comparisons, with most going via the placebo node, warranting future studies to draw rmer conclusions regarding the results.Third, high global inconsistency emerged across most comparisons, re ecting the trial duration heterogeneity and the underlying antipsychotic drugs and doses comprising TAU.Fourth, the limited number of studies also precluded feasible subgroup and sensitivity analyses to explain the observed inconsistency.Fifth, many studies lacked information regarding illness characteristics that could affect outcomes and inform subgroups, including illness phase (i.e., acute vs. clinically stable).Sixth, only 54% of the studies had a low risk of bias.Seventh, low and very low con dence in evidence emerged in the CINeMA, underscoring again that more high-quality studies are needed and that the results of this NMA need to be interpreted with caution, especially considering the trend for larger effect sizes documented by records with lower AMSTAR plus content score (adapted).Finally, caution in translating the results from this NMA to clinical care is also warranted because nutraceutical/phytoceuticals are often heterogeneous in composition, and the bioavailability and the effects of the same compound may vary across different preparations.No clear-cut guidance is often available because FDA and other regulatory agencies' requirements for nutraceuticals and phytoceuticals differ from traditional drugs 96 .Conclusions Among many nutraceuticals tested as an augmentation treatment for schizophrenia, sarcosine (2gr/day), citicoline (2.5gr/day), and N-acetylcysteine (1.2-3.6gr/day)consistently show a bene cial effect for total and speci c symptoms, almost exclusively in stable patients, with good overall acceptability.However, multiple limitations encourage caution in interpreting and translating the results into clinical practice.Larger and high-quality RCTs with a low risk of bias should investigate the effects of nutraceuticals in schizophrenia, especially using heterogeneous daily doses and formulations.Note: data refers to any phase of the disease.

Figures Figure 1 PRISMA
Figures
AIMS=Abnormal Involuntary Movements Scale; BARS=Barnes Akathisia Rating Scale; BDNF=Brain Derived Neurotrophic Factor; BPRS=Brief Psychiatric Rating Scale; CANTAB=Cambridge Automated Neuropsychological Test Battery; CDSS=Calgary Depression Scale for Schizophrenia; CGI=Clinical Global Impression; DHA=docosahexaenoic acid; EPA=eicosatetraenoic acid; ESRS=Extrapyramidal Symptoms Rating Scale; GAF=Global Assessment of Functioning; GAS=Global Assessment Scale; HDRS=Hamilton Depression rating Scale; LIFE-RIFT=Range of Impaired Functioning Tool; MATRICS=Measurement and Treatment Research to Improve Cognition in Schizophrenia; MOAS= Modi ed Overt Aggression Scale; MoCA= Montreal Cognitive Assessment; PSS=Perceived Stress Scale; Q-LES-Q18=Quality of Life Enjoyment and Satisfaction Questionnaire; QLS=Quality of Life Scale; QOL=Quality Of Life; RBANS=RepeaTable Battery for Assessment of Neuropsychological Status; SANS=Scale for Assessment of Negative Symptoms; SAPS=Scale for Assessment of Positive Symptoms; SAS=Simpson Angus Scale; SCZ=schizophrenia; SOFAS=Social and Occupational Functioning Assessment Scale; TRS=Treatment Resistant Schizophrenia; WAIS=Wechsler Adult Intelligence Scale.
*Study entering the SR portion only.