New Pharmacological Treatment Approaches for Schizophrenia: Navigating the Post-iclepertin Landscape

Abstract

Schizophrenia is a complex psychiatric disorder, with cognitive impairment representing a critical unmet medical need. Current treatments primarily address positive symptoms, failing to effectively improve cognitive function and negative symptoms. This narrative review examines the current landscape of novel schizophrenia therapeutics following the discontinuation of iclepertin development, critically evaluates the most promising current candidates, and identifies priority research areas for future drug development. A comprehensive literature review was conducted covering developments from 2020 to 2025, with emphasis on Phase II and Phase III clinical data for novel mechanisms beyond traditional dopamine D₂ receptor antagonism. The landscape has been dramatically reshaped by two pivotal developments: the U.S. Food and Drug Administration (FDA) approval of xanomeline/trospium (KarXT, Cobenfy^®) in September 2024 as the first nondopaminergic antipsychotic, and the failure of iclepertin in Phase III trials in January 2025. The most compelling therapeutic approaches continue to be serotonin–dopamine activity modulators (brilaroxazine) and emerging M₄ selective agonists (NBI-1117568). Moreover, recent evidence suggests a more cautious outlook for TAAR1 agonists, with systematic analyses revealing significant placebo response challenges in recent trials. While iclepertin’s discontinuation represents a significant setback for GlyT1 inhibition strategies, the approval of KarXT and advancing pipeline candidates offer potential for therapeutic advances. Critical challenges include escalating placebo responses that compromise trial assay sensitivity, underscoring the need for improved patient stratification, refined trial methodology, and outcome measures that accurately capture real-world benefit.

Key Points

Schizophrenia treatment is at a turning point, marked by the approval of KarXT (the first nondopaminergic antipsychotic) and iclepertin’s Phase III discontinuation.

This review highlights the most promising novel therapies in development, including muscarinic modulators and serotonin–dopamine activity modulators.

Genetic and neurophysiological biomarkers for patient stratification, refined trial methodology to address escalating placebo responses, and functional outcome measures are essential to ensure new treatments translate into real-world benefits.

Introduction

Current Unmet Medical Needs in Schizophrenia Treatment

Schizophrenia affects approximately 24 million individuals worldwide, representing a complex psychiatric disorder characterized by positive symptoms (hallucinations, delusions), negative symptoms (social withdrawal, lack of motivation), and cognitive impairment that significantly impacts daily functioning and quality of life [1, 2]. The economic burden of schizophrenia, including direct healthcare costs and indirect costs from lost productivity, is substantial, further underscoring the critical need for effective treatments, especially for cognitive impairment associated with schizophrenia (CIAS).

Despite decades of therapeutic advancement, current antipsychotic medications—the gold standard treatment for schizophrenia—primarily address positive symptoms through dopamine D₂ receptor antagonism while failing to effectively treat CIAS and negative symptoms [3, 4]. Cognitive impairment, affecting approximately 80% of individuals with schizophrenia, often precedes the onset of positive and negative symptoms and persists throughout the illness course [5]. CIAS encompasses deficits in processing speed, episodic memory, working memory, and executive function, making it the strongest predictor of long-term functional outcomes and representing a critical unmet medical need [6, 7]. A recent systematic review and network meta-analysis by Feber et al. [8], including 68 trials with 9525 participants, confirmed that antipsychotics do not act as cognitive enhancers. The modest improvements in cognitive performance observed compared with placebo likely reflect secondary benefits from reduced disordered thinking following improvement in positive symptoms, rather than genuine cognitive enhancement [8]. Moreover, no pharmacotherapies specifically approved to target these domains were available, creating a substantial void in comprehensive schizophrenia care [9, 10].

Historical Context and Recent Paradigm Shift

Glutamate Hypothesis

The limitations of purely dopaminergic approaches have spurred investigation into alternative neurotransmitter systems, most notably the glutamatergic and cholinergic pathways [4, 11]. The glutamate hypothesis of schizophrenia emerged from observations that N-methyl-d-aspartate (NMDA) receptor antagonists can induce schizophrenia-like symptoms, including cognitive impairment [12]. This framework has driven the development of multiple therapeutic strategies aimed at enhancing NMDA receptor (NMDAR) function, e.g. via the coagonist glycine, including glycine transporter 1 (GlyT1) inhibition or muscarinic agents modulating the NMDA receptor stringency [13–15].

Recent Pivotal Developments

The therapeutic landscape has been fundamentally altered by two pivotal developments: The U.S. Food and Drug Administration (FDA) approved xanomeline/trospium chloride (KarXT, Cobenfy^®) in September 2024 as the first and only nondopaminergic antipsychotic, and iclepertin (BI 425809) failed in Phase III trials announced in January 2025. The near-simultaneous occurrence of these events marks a unique turning point: a concurrent validation of nondopaminergic pathways through KarXT and a profound cautionary tale regarding a specific glutamatergic approach via iclepertin. This dual nature necessitates re-evaluation of target validation and translational science, requiring innovative mechanisms and more rigorous assessment of preclinical models and early clinical signals.

Other Setbacks and Strategic Implications

These pivotal shifts occurred amidst several other setbacks for promising compounds, including roluperidone and pimavanserin (both 5-HT_2A-related mechanisms targeting negative symptoms), and various trace amine-associated receptor 1 (TAAR1) agonists, even as new muscarinic modulators continue to advance their clinical development programs. In addition, luvadaxistat (NBI-1065844), a D-amino acid oxidase (DAAO) inhibitor targeting CIAS, failed to replicate positive Phase II cognitive signals [16] in a confirmatory Phase II trial [17] and was discontinued in September 2024, further exemplifying the challenges facing NMDAR-modulating approaches for cognitive enhancement. Collectively, these events have fundamentally reshaped the strategic direction of schizophrenia drug development.

Scope and Objectives

This narrative review provides a comprehensive assessment of the current treatment landscape for schizophrenia, analyzing the implications of recent setbacks and successes. Our objectives are: (i) to analyze lessons learned from iclepertin’s failure and the significance of KarXT’s approval; (ii) to evaluate current promising therapeutic candidates across diverse mechanisms including novel digital therapeutics; (iii) to examine the broader pattern of novel mechanism failures and their implications for future development strategies with particular emphasis on the systematic increase in placebo response rates that has compromised trial assay sensitivity; and (iv) to identify critical research gaps and priority areas for future investigation, particularly those related to precision medicine approaches.

Literature Search and Data Synthesis

Literature Search Strategy

A comprehensive literature search was conducted using PubMed, Embase, clinical trial registries, and pharmaceutical press releases covering January 2020 to January 2025, with emphasis on developments from 2024 to 2025. Search terms included combinations of “schizophrenia,” “novel antipsychotics,” “cognitive impairment,” “TAAR1 agonists,” “GlyT1 inhibitors,” “muscarinic modulators,” “negative symptoms,” “5-HT2A modulators,” “precision medicine,” “biomarkers,” “DAAO inhibitors,” “digital therapeutics,” “placebo response,” and specific compound names. Priority was given to Phase II and Phase III clinical data, systematic reviews, meta-analyses, regulatory communications, and mechanistic studies elucidating translational challenges.

In addition, press releases and investor communications from pharmaceutical companies were systematically reviewed to capture up-to-date information on pipeline developments and trial outcomes not yet published in peer-reviewed literature. This pragmatic approach ensures timeliness in a rapidly evolving field, although it acknowledges the inherent tension between rapid dissemination via corporate communications and the verified scientific rigor of peer-reviewed publications.

Data Synthesis Approach

Information was synthesized and categorized on the basis of mechanism of action, with critical evaluation of each compound’s clinical development stage, efficacy signals, safety profiles, and regulatory pathway clarity. This critical evaluation involved assessing the methodological quality of studies, identifying potential biases, and considering the generalizability of findings. While a precise count of screened and included articles would be beyond the scope of this review, the search yielded over 400 relevant publications, from which the most pivotal Phase II and Phase III data, systematic reviews, and mechanistic insights were selected for detailed analysis. Temporal discrepancies were resolved by prioritizing the most recent evidence-based sources, with particular emphasis on systematic reviews and meta-analyses providing comprehensive efficacy assessments.

The Iclepertin Case Study: Lessons Learned and Implications

Iclepertin’s Development Timeline and Phase III Failure

Iclepertin (BI 425809), developed by Boehringer Ingelheim, represented the most advanced GlyT1 inhibitor in clinical development for CIAS. Its mechanism was predicated on enhancing NMDA receptor function by increasing synaptic glycine concentrations through selective inhibition of glycine transporter 1 [18]. Phase I studies confirmed target engagement through dose-dependent increases in cerebrospinal fluid glycine levels in both rodents and healthy volunteers [19]. Phase II results (NCT02832037, n = 509, 12 weeks) demonstrated statistically significant pro-cognitive effects with 10 mg and 25 mg iclepertin versus placebo on the MATRICS Consensus Cognitive Battery (MCCB) overall composite T-score [20]. Based on similar efficacy and safety profiles [21], the 10 mg once-daily dose was selected for Phase III development.

The Phase III CONNEX program comprised three similarly designed, multinational, randomized, double-blind, placebo-controlled trials (CONNEX-1 [NCT04846868], CONNEX-2 [NCT04846881], and CONNEX-3 [NCT04860830]) conducted at 338 specialist psychiatric centers across 41 countries between September 2021 and April 2024. A total of 1835 patients received at least one dose of trial medication (918 iclepertin, 917 placebo), with 1602 completing the 26-week treatment period. The study population was broadly representative, with mean age 34.7 years (SD 8.8), 66% male, and mean time since first diagnosis 10.6 years. Patients were randomized 1:1 to receive 10 mg oral iclepertin or placebo once daily as adjunctive therapy to standard-of-care antipsychotics (excluding clozapine). Importantly, the trials did not require specific cognitive impairment thresholds at baseline, enrolling a broad population of adults aged 18–50 years with DSM-5 schizophrenia and functional impairment. Treatment adherence was high (89% by tablet counting, 63% by AiCure digital monitoring), and multiple quality control measures were implemented, including rater training and independent expert review of cognitive assessments [22].

In January 2025, Boehringer Ingelheim announced that iclepertin failed to meet its primary and key secondary end points in the CONNEX program, with no statistically significant effects on cognition or functioning compared to placebo after 6 months of treatment [23]. The full publication of results in November 2025 provided comprehensive confirmation of the program’s failure [21]. The primary end point—change from baseline at week 26 in MCCB overall composite T-score—showed no statistically significant difference between iclepertin and placebo in any individual trial or in the pooled population (adjusted mean difference 0.127, 95% CI − 0.396 to 0.650; p = 0.63). At week 26, the adjusted mean change from baseline was 2.293 (95% CI 1.922–2.663) for iclepertin versus 2.166 (95% CI 1.796–2.535) for placebo. Similar null findings were observed at week 12 (adjusted mean difference 0.11, 95% CI − 0.37 to 0.58; p = 0.66). All secondary end points, including the Schizophrenia Cognition Rating Scale (SCoRS) total score (adjusted mean difference 0.620, 95% CI − 0.027 to 1.268; p = 0.061), Virtual Reality Functional Capacity Assessment Tool (VRFCAT) adjusted total time T-score (adjusted mean difference − 0.445, 95% CI − 1.688 to 0.798; p = 0.48), and all further prespecified efficacy measures, failed to demonstrate significant treatment effects. Comprehensive subgroup analyses stratified by sex, age, race, region, time since diagnosis, baseline cognitive impairment severity (MCCB T-score categories), treatment adherence, and concomitant antipsychotic use revealed no evidence of efficacy in any subpopulation. Responder analyses showed similar proportions of patients in both treatment groups meeting predefined improvement thresholds across all cognitive and functional measures [21].

A critical observation from the CONNEX program was the magnitude of placebo response at week 26, which exceeded that observed in the Phase II trial at week 12. While placebo responses at week 12 were comparable between Phase II and Phase III studies, the additional improvement observed with placebo at week 26 in Phase III was consistent with known practice effects from repeated MCCB administration. The defining feature distinguishing Phase II from Phase III outcomes was the substantially smaller cognitive improvement observed with iclepertin in Phase III compared with Phase II, suggesting that repeated testing had a greater effect on MCCB performance than iclepertin treatment itself [21].

Despite the complete absence of cognitive or functional efficacy signals, iclepertin demonstrated a favorable safety and tolerability profile consistent with Phase II findings. On-treatment adverse events were reported in 604 (66%) patients receiving iclepertin versus 633 (69%) receiving placebo. Drug-related adverse events occurred in 169 (18%) versus 189 (21%), respectively, with headache being the most common (2% versus 1%). Serious adverse events were infrequent (iclepertin 36 [4%] versus placebo 46 [5%]), with no clinically relevant changes in physical examinations, vital signs, ECG parameters, or laboratory values. Notably, schizophrenia relapse rates were low and comparable between groups (both 1%), with no worsening of psychotic symptoms assessed by PANSS total score. Three deaths occurred during the trials (one placebo on-treatment, two iclepertin during safety follow-up), all deemed unrelated to trial treatment. Extrapyramidal symptoms, suicidal ideation and behavior, and discontinuation rates due to adverse events were similar between treatment groups [21].

The CONNEX program represents the largest clinical trial conducted for CIAS to date. The trial authors concluded that “targeting GlyT1 might be insufficient for treating CIAS in a broad population of patients with schizophrenia” and recommended that “future trials targeting CIAS might have a higher likelihood of success if the trial population was less broad compared with CONNEX.” Specifically, they suggested preselecting patients with significant cognitive impairment at screening, either through specific MCCB eligibility criteria or by enriching trial populations with potential responders based on underlying disease etiology [21].

Implications for GlyT1 Inhibitor Strategies

Iclepertin’s failure follows earlier disappointments with other GlyT1 inhibitors, such as bitopertin, which failed Phase III trials despite favorable effects in Phase II [24]. This recurring pattern suggests fundamental challenges with the GlyT1 inhibition approach that extend beyond compound-specific issues, necessitating a fundamental re-evaluation of the GlyT1 target itself, not just the specific compounds. Recent theoretical work by Singer and Yee [25] provides mechanistic insights into these challenges, proposing three key hypotheses for why GlyT1 inhibition may paradoxically impair rather than enhance NMDAR function. While these hypotheses are gaining traction, it is important to note that they represent a current theoretical framework that continues to be debated and refined within the scientific community.

Their model suggests that astrocytic GlyT1 blockade may disrupt the serine shuttle mechanism, thereby depleting the l-serine available for neuronal d-serine synthesis. Furthermore, because astrocytic GlyT1 can operate in reverse to release glycine during synaptic activation, its pharmacological blockade may reduce this activity-dependent glycine supply. Finally, the authors propose that sustained blockade may disrupt the essential cyclic operation of astrocytic GlyT1, which normally switches between uptake and release modes [25].

These insights help explain the insufficient dose optimization for bitopertin, limited EEG biomarker effects [26, 27], and unpredictable directions of GlyT1 inhibition. The failure underscores the complexity of glycine neurotransmission and the critical importance of neuron-glia interactions in NMDAR regulation [28].

Luvadaxistat: The Challenge of Replicating DAAO Inhibition Efficacy

Paralleling iclepertin’s failure, luvadaxistat (NBI-1065844/TAK-831) represented an alternative NMDAR enhancement strategy through D-amino acid oxidase (DAAO) inhibition, designed to elevate synaptic D-serine levels rather than glycine. The Phase II INTERACT study (NCT03382639, n = 256, 14 weeks) failed its primary end point for negative symptoms but showed exploratory cognitive benefits at 50 mg (BACS composite p = 0.031, effect size 0.4; SCoRS p = 0.011, effect size 0.4), prompting advancement focused on CIAS [16].

However, the Phase II ERUDITE study (NCT05182476) failed to replicate these cognitive findings. On 12 September 2024, Neurocrine Biosciences announced complete discontinuation of luvadaxistat development, citing failure to replicate the cognitive battery improvements observed in the INTERACT Phase II trial. The company attributed the replication failure “in part to the large variability seen in the cognitive measures across the population studied and a potential imbalance in the baseline characteristics of subjects enrolled across the treatment arms,” highlighting the substantial challenges in CIAS drug development [29].

A defining characteristic of luvadaxistat’s pharmacology was its inverted U-shaped dose-response curve, wherein efficacy signals at 50 mg were lost at higher doses (125 mg, 500 mg) across preclinical models, human electrophysiology (mismatch negativity), and clinical cognitive measures. This pattern, analyzed comprehensively by Terry-Lorenzo et al. [16], likely reflects narrow therapeutic windows created by high baseline NMDAR glycine-site occupancy, with overcorrection at higher doses potentially triggering receptor internalization. Regional brain distribution issues—DAAO expression is highest in cerebellum with moderate cortical levels—may further limit d-serine elevation in cognition-relevant regions.

The luvadaxistat experience adds to the accumulating evidence that DAAO inhibition, similar to GlyT1 inhibition, faces fundamental translational challenges in CIAS treatment, including narrow therapeutic windows, complex dose–response relationships, inadequate understanding of system-level NMDAR regulation, and heterogeneity in patient populations regarding baseline receptor occupancy and coagonist status.

Broader Implications for Glutamatergic Approaches

While the discontinuation of iclepertin represents a setback for direct glutamatergic enhancement strategies, it should not be interpreted as definitive refutation of all NMDA receptor approaches. Alternative strategies, including DAAO inhibition and direct NMDA receptor modulation, remain in development. The discontinuations of both iclepertin (GlyT1 inhibitor) and luvadaxistat (DAAO inhibitor) represent significant setbacks for NMDAR enhancement strategies targeting CIAS. Both compounds demonstrated Phase II cognitive signals that failed to replicate in larger confirmatory trials, revealing a systematic pattern that transcends individual compound properties and suggests fundamental challenges with current approaches to NMDAR coagonist modulation: The convergent failures of GlyT1 and DAAO inhibition illuminate several critical translational barriers. Both mechanisms revealed narrow and unpredictable therapeutic windows, with bitopertin and luvadaxistat each demonstrating inverted U-shaped dose-response relationships where higher doses lost efficacy [16, 24]. Inadequate target engagement was documented through limited effects on electrophysiological biomarkers [26, 27]. The mechanistic complexity of neuron-glia interactions governing NMDAR coagonist availability suggests that simple elevation of synaptic coagonist levels may produce counterintuitive systemic effects. In addition, high measurement variability in cognitive batteries and substantial placebo responses in larger trials create methodological obstacles to detecting modest treatment effects, even if biologically present.

Despite these setbacks, alternative glutamatergic strategies warrant continued investigation with refined approaches. Notably, the experience with pomaglumetad (LY2140023)—a metabotropic glutamate receptor 2/3 agonist that initially failed in traditional clinical trials [30] but is now being redeveloped using precision medicine approaches by Denovo Biopharma [31]—illustrates how glutamatergic strategies may require biomarker-guided patient selection rather than broad population treatments. It should be noted that mGlu2/3 agonists have been extensively studied and have consistently failed at lower doses. Substantial biomarker work has suggested that pomaglumetad was likely underdosed in prior trials, which may explain the initial failures and supports the rationale for biomarker-guided redevelopment at optimized doses. Similarly, direct NMDA receptor modulators that bypass coagonist site limitations, or approaches targeting downstream plasticity mechanisms, may circumvent the system-level complications revealed by GlyT1 and DAAO inhibition failures.

In addition, clinical subgroups, disease stages, and specific combination strategies might be promising directions of research. The iclepertin and luvadaxistat experiences collectively underscore that successful NMDAR enhancement for CIAS will likely require: (i) precision medicine frameworks identifying patients with specific NMDAR-related pathophysiology, including assessment of baseline glycine-site occupancy, d-serine/glycine ratios, and genetic variants affecting NMDAR function; (ii) functional biomarkers capable of demonstrating target engagement and predicting therapeutic response in early development to de-risk later stages; (iii) more sophisticated preclinical models that capture system-level complexity of neurotransmitter regulation and neuron-glia metabolic coupling; (iv) trial designs accounting for high cognitive measure variability and escalating placebo responses; and (v) potentially, focus on mechanisms that preserve rather than disrupt the dynamic regulation of NMDAR coagonist systems. Without these strategic refinements, even mechanistically sound NMDAR-modulating compounds risk continued failures in late-stage development despite early promise.

The KarXT Success Story: Validating Non-dopaminergic Mechanisms

Xanomeline/Trospium Development and Approval

The FDA approval of xanomeline/trospium chloride (KarXT, Cobenfy^®) in September 2024 represents a significant milestone, establishing it as the first and only nondopaminergic antipsychotic approved for schizophrenia [32, 33]. This combination product leverages xanomeline’s muscarinic M₁/M₄ receptor agonist activity while using the peripherally restricted antagonist trospium to mitigate peripheral cholinergic side effects [34, 35].

Mechanism of Action and Clinical Efficacy

Xanomeline is a selective muscarinic acetylcholine receptor agonist that primarily targets M₁ and M₄ receptors. M₁ receptors are involved in learning, memory, and cognitive processes within the cerebral cortex and hippocampus, whereas M₄ receptors modulate striatal dopamine release and provide antipsychotic effects [36, 37]. Trospium’s peripheral restriction prevents the cholinergic adverse events that limited the monotherapy development of xanomeline.

KarXT’s approval was based on positive results from two Phase III trials (EMERGENT) which demonstrated significant improvements in Positive and Negative Syndrome Scale (PANSS) total scores compared with placebo [38]. Importantly, post-hoc analyses of Phase III data suggested cognitive benefits in patients with impaired baseline cognition in meta-analysis indicating a moderate effect size (Cohen’s d = 0.54), though these findings require prospective confirmation [39, 40]. Whether these cognitive improvements represent direct enhancement or secondary benefits from positive symptom reduction as observed with traditional antipsychotics [8], requires prospective validation.

Regulatory and Commercial Implications

The approval of KarXT validates nondopaminergic mechanisms and establishes a regulatory precedent for novel approaches. While KarXT demonstrates a lower incidence of extrapyramidal symptoms (EPS) and metabolic neutrality compared with traditional D₂ antagonists, it is associated with cholinergic adverse events that require consideration in clinical practice. A systematic review and meta-analysis by Kishi et al. [41] examining three randomized controlled trials confirmed that KarXT demonstrated superior PANSS improvements with discontinuation rates comparable to placebo. The compound was associated with cholinergic adverse events (dry mouth, gastrointestinal symptoms, hypertension) but showed incidence of extrapyramidal symptoms, weight gain, and drowsiness similar to placebo, confirming its favorable metabolic and motor profile compared with traditional dopaminergic antipsychotics. This demonstrates that for novel antipsychotics, a differentiated side-effect profile may influence treatment adherence and clinical acceptance. However, recent data from the ARISE trial showed that adjunctive KarXT failed to meet the primary end point when added to existing atypical antipsychotics, highlighting difficulties in demonstrating incremental benefits beyond established treatments [42]. Despite this limitation, the approval provides an alternative to traditional antipsychotics, distinguished by its novel mechanism and distinct side-effect profile while establishing a regulatory precedent for nondopaminergic approaches.

Commercially, KarXT is also being explored for treatment refractory schizophrenia, as well as a broader range of neuropsychiatric conditions, including cognitive symptoms (MINDSET program), agitation (ADAGIO program) and psychosis (ADEPT program) in Alzheimer’s disease [43], and manic episodes in bipolar I disorder (BALSAM program) [44], representing significant market opportunities beyond schizophrenia.

Current Promising Pipeline Candidates

TAAR1 Agonists: Reassessing Prospects following Comprehensive Evidence

Ulotaront (SEP-363856, Sumitomo Pharma) remains the most advanced TAAR1 agonist, representing a fundamentally different mechanism from traditional antipsychotics [45, 46]. As a full agonist at TAAR1 with additional 5-HT_1A receptor agonist activity, ulotaront modulates dopamine, serotonin, and glutamate systems without direct D₂ receptor blockade [47], and shows robust effects on negative symptoms as well as moderate improvements in acute schizophrenia.

However, a recent comprehensive meta-analysis offers a more cautious perspective, revealing that ulotaront and ralmitaront demonstrated only modest improvements in acute schizophrenia (pooled SMD 0.15, 95% CI: − 0.05 to 0.34), substantially smaller than traditional antipsychotics, which typically show effect sizes ranging from 0.24 to 0.89 [48, 49]. In direct comparison, ralmitaront was significantly less efficacious than risperidone (SMD = − 0.53, 95% CI − 0.86 to − 0.20) [50]. Despite these meta-analytic findings, ulotaront’s Phase II studies showed robust efficacy (PANSS improvement − 17.2 versus − 9.7 placebo, p = 0.001, effect size = 0.45) with sustained benefits in a 26-week extension [51, 52]. Large placebo responses in Phase III DIAMOND trials may have compromised assay sensitivity [48, 53], reflecting the methodological crisis described in Sect. 7.

Importantly, the contrasting outcomes between ulotaront and ralmitaront within the TAAR1 class provide crucial insights into mechanism optimization. Ralmitaront (RO6889450/RG-7906), developed by Roche, was terminated in October 2023 following failure of both Phase II trials. Critical pharmacological differences distinguish these compounds: ralmitaront exhibits significantly lower efficacy at TAAR1 receptors, approximately 30-fold slower association kinetics, and—most critically—lacks detectable activity at serotonin 5-HT_1Areceptors. Expert analysis suggests that dual TAAR1/5-HT1A activity may be essential for clinical efficacy, explaining why ralmitaront’s TAAR1-selective approach proved insufficient despite sound mechanistic rationale [54].

Large placebo responses in the Phase III DIAMOND trials may have compromised assay sensitivity [48, 54], reflecting the methodological crisis described in Sect. 7. Preclinical evidence suggests TAAR1 activation preferentially modulates dopamine under hyperdopaminergic conditions, potentially explaining both favorable safety and modest efficacy in unselected populations [55–57].

Serotonin-Dopamine Activity Modulators: Brilaroxazine’s Continued Success

Brilaroxazine (RP5063, Reviva Pharmaceuticals) represents a novel multitarget approach with activity as a D₂ partial agonist, 5-HT_1A agonist, 5-HT_2A antagonist, and 5-HT₇ antagonist [58]. Recent data from the Phase II RECOVER trial and its 1-year open-label extension study have solidified brilaroxazine’s position as a leading candidate. The RECOVER trial met all primary and secondary end points, demonstrating robust and sustained improvements across symptom domains. The 1-year extension showed dose-dependent efficacy and a favorable safety profile, with a 35% discontinuation rate over 1 year [59]. In preclinical studies, brilaroxazine was notably associated with reductions in multiple neuroinflammatory markers, including inflammatory cytokines, suggesting a broader disease-modifying potential beyond mere symptomatic relief [60].

M₄ Selective Agonists: NBI-1117568 Emerging as a New Contender

NBI-1117568 (Neurocrine Biosciences, formerly HTL-0016878) represents the first and only investigational oral muscarinic M₄ selective orthosteric agonist in clinical development [61]. Recent Phase II results showed statistically significant improvements in PANSS total scores with the 20 mg dose by week 3, maintained through week 6 [62]. The compound demonstrated a favorable safety profile with treatment discontinuation rates similar to placebo. Most common adverse events were somnolence and dizziness, with no weight gain observed. Based on these results, Neurocrine Biosciences initiated a Phase III registrational program in 2025 [63].

Next-Generation Muscarinic Modulators: ML-007 and NS-136

The success of KarXT has catalyzed development of next-generation muscarinic approaches, with ML-007 (MapLight Therapeutics) and NS-136 (NeuShen Therapeutics) representing distinct advances in muscarinic receptor modulation [35]. ML-007 employs a dual M₁/M₄ agonist mechanism combined with an innovative fixed-dose combination approach, pairing the agonist with a precision-matched peripherally acting anticholinergic (PAC) to mitigate peripheral side effects while maintaining central efficacy. The compound has completed three Phase I trials encompassing 270 healthy volunteers [64–66], demonstrating favorable safety with no serious adverse events and minimal extrapyramidal symptoms. The ZEPHYR Phase II study initiated in July 2025 represents a critical milestone and is expected to enroll approximately 300 patients with acute schizophrenia [67]. In parallel, MapLight Therapeutics is running a Phase II study program for ML-007 in Alzheimer’s disease psychosis [68, 69], highlighting the potential for muscarinic modulation across various psychotic conditions.

NS-136 (NeuShen Therapeutics) represents a mechanistically distinct approach as a selective M₄ positive allosteric modulator (PAM) developed using proprietary artificial intelligence (AI) technology. Unlike direct receptor agonists, PAMs offer enhanced selectivity and reduced peripheral effects by potentiating endogenous acetylcholine signaling [70]. NeuShen Therapeutics initiated a first-in-human trial in 2024 (NCT06345703) [71] which, as of November 2025, is still listed as actively recruiting.

Benzamide Antipsychotics: LB-102’s Promising Debut

LB-102 (LB Pharmaceuticals), a methylated derivative of amisulpride, is positioned as a potential first-in-class benzamide antipsychotic for the U.S. market [72]. The compound successfully completed its Phase II NOVA1 trial (NCT05055089, n = 359, 4 weeks), a randomized, double-blind, placebo-controlled, multi-center study in adults with acute schizophrenia, meeting its primary end point with statistically significant reductions in PANSS total scores across all dose levels (50 mg, 75 mg, 100 mg once daily). The 50 mg dose demonstrated an effect size of 0.61 with a favorable safety profile, including a low incidence of extrapyramidal symptoms and minimal metabolic effects [73].

Data presented at the 38th European College of Neuropsychopharmacology (ECNP) Congress in October 2025 expanded understanding of LB-102’s clinical profile across multiple symptom domains [74]. A post-hoc analysis of cognitive effects, evaluated as an exploratory end point using a validated battery of cognitive tests, demonstrated dose-dependent and statistically significant improvements in a completer population not enriched for baseline cognitive impairment. Treatment effect sizes versus placebo after 4 weeks were 0.26 (p = 0.0476) at 50 mg, 0.41 (p = 0.0027) at 75 mg, and 0.66 (p = 0.0018) at 100 mg. These effect sizes, particularly at higher doses, represent encouraging preliminary evidence that LB-102 may address cognitive impairment [75]. Additional post-hoc analyses presented at ECNP examined effects on negative symptoms in patients with prominent negative symptoms at baseline, further characterizing LB-102’s activity across symptom domains [76].

LB-102’s balanced clinical profile is characterized by robust efficacy on positive symptoms [77], a potentially class-leading tolerability profile among dopamine D₂/D₃ antagonists, and differentiated activity across cognitive and negative symptom domains. This multidimensional efficacy pattern positions LB-102 as a candidate for comprehensive disease management beyond traditional positive symptom control. The compound is advancing into Phase III trials for acute schizophrenia and Phase II development for bipolar depression, with planned expansion into additional indications including treatment-resistant negative symptoms and cognitive impairment associated with schizophrenia.

Glutamate Modulators: Evenamide’s Alternative Approach

Evenamide (NW-3509, Newron Pharmaceuticals) represents a distinct glutamatergic strategy, acting as a voltage-gated sodium channel inhibitor to modulate excessive glutamate release [78]. In a Phase II trial in treatment-resistant schizophrenia, it showed sustained benefits over 1 year, with 47.4% of patients achieving at least a 20% reduction in PANSS scores at 52 weeks [79]. A pivotal Phase III trial is planned for Q2 2025, backed by key licensing agreements [80]. Notably, evenamide’s mechanism differs fundamentally from the previously unsuccessful GlyT1 inhibition approach and the failed DAAO inhibition strategy.

Repurposed Therapies: Varenicline’s Dual Potential

Varenicline (Chantix^®/Champix^®), originally approved for smoking cessation, has emerged as a unique opportunity in the post-iclepertin landscape through its dual nicotinic acetylcholine receptor modulation. The compound demonstrates α₄β₂ receptor partial agonism combined with α₇ receptor full agonism [81], directly addressing cholinergic hypofunction associated with cognitive impairment in schizophrenia. Although initial studies reported small cognitive effects for varenicline [82], meta-analyses have failed to confirm any consistent benefit on cognition [83]. As such, varenicline does not appear to hold promise as a cognitive enhancer in schizophrenia. However, its robust efficacy in promoting smoking cessation [84] remains highly relevant, given that smoking prevalence in this population exceeds 70% [85]. Thus, with a favorable safety profile in psychiatric populations, varenicline represents a valuable intervention to address the substantial somatic burden associated with tobacco use in schizophrenia.

For an overview of emerging pharmacological pipeline candidates, see Table 1.

Table 1.

Emerging antipsychotic candidates for schizophrenia

Compound name (manufacturer)	Mechanism of action	Clinical trial registration number (most recent) and recruitment status	Development stage	Key efficacy highlights	Key safety/tolerability highlights	Current status/outlook
Brilaroxazine/RP5063 (Reviva)	D2 partial agonist, 5-HT1A agonist, 5-HT2A/5-HT7 antagonist	NCT05184335 (recruiting)	Phase II (NDA filing alignment) Phase III (RECOVER) open-label extension 1-year study	Robust, sustained improvements across symptom domains; met all primary/secondary end points	Favorable safety profile; 35% discontinuation over 1 year Preclinical studies showed reductions in neuroinflammatory markers	Leading candidate, advancing towards NDA filing
Evenamide (NW-3509, Newron Pharma)	Voltage-gated sodium channel inhibitor	NCT07184619 (not yet recruiting)	Phase II (completed), Phase III (planned Q2/3 2025)	Sustained benefits in treatment-resistant schizophrenia	Sustained benefits over 1 year with favorable tolerablility	Alternative glutamatergic strategy distinct from failed GlyT1 and DAAO inhibition, pivotal Phase III planned
LB-102 (LB Pharmaceuticals)	Methylated amisulpride (D2/3/5-HT7 inhibitor)	NCT06179108 (completed)	Phase II (NOVA-1) completed	Statistically significant PANSS reductions across all dose levels	Favorable safety, low EPS incidence, minimal metabolic effects	Promising Phase II debut; positioned as potential first-in-class benzamide for U.S. market; Phase III planned late 2025/early 2026
ML-007 (MapLight Therapeutics)	Dual M1/M4 agonist + PAC	NCT07038876 (recruiting)	Phase II (ZEPHYR) initiated July 2025	Favorable safety in Phase I (no serious AEs, minimal EPS)	Favorable safety, no serious AEs, minimal EPS	Critical Phase II study initiated; innovative fixed-dose combination approach with precision-matched PAC to mitigate peripheral side effects
NBI-1117568 (Neurocrine Biosciences)	M4 selective orthosteric agonist	NCT07114874 (recruiting)	Phase III (initiated 2025)	Statistically significant PANSS improvements, maintained by week 6	Favorable safety profile; most common side effects: somnolence / dizziness. No weight gain.	New contender; Phase III registrational program initiated
NS-136 (NeuShen Therapeutics)	Selective M4 PAM	NCT06345703 (recruiting)	First-in-human (initiated 2024)	Enhanced selectivity, reduced peripheral effects (theoretical)	Actively recruiting, early stage safety data pending	Mechanistically distinct; developed using proprietary AI technology; early clinical development
Ulotaront / SEP-363856 (Sumitomo/Otsuka)	TAAR1 agonist, 5-HT1A agonist	NCT06894212 (recruiting)	Phase II/III (DIAMOND) Meta-analysis	Modest improvements in acute schizophrenia; Phase II showed improvements in PANSS total score and negative symptoms when used as adjunctive treatment	Preclinical; beneficial metabolic profile; favorable safety profile in Phase II	Reassessing path; dual TAAR1/5-HT1A activity critical. Large placebo responses in DIAMOND trials may have compromised assay sensitivity
Varenicline (Chantix®/Champix®-Pfizer)	α4β2 partial agonist, α7 full agonist	n/a	Repurposed	Small cognitive effects initially; meta-analyses (four trials, n = 339 patients) failed to confirm consistent cognitive benefit	Favorable safety profile in psychiatric populations; robust efficacy for smoking cessation	Valuable for smoking cessation in SCZ, limited promise for CIAS. Should be focused on smoking cessation rather than cognitive improvement

AE adverse event, CIAS cognitive impairment associated with schizophrenia, D2/3 dopamine receptor subtypes 2 and 3, EPS extrapyramidal symptoms, FDA U.S. Food and Drug Administration, n/a not applicable, NDA New Drug Application (regulatory filing for approval), PAC peripherally acting anticholinergic, PAM positive allosteric modulator, PANSS Positive and Negative Syndrome Scale, SCZ schizophrenia, TAAR1 trace amine associated receptor 1, 5-HT serotonin receptor subtypes

Digital Therapeutics: CT-155’s Breakthrough for Negative Symptoms

Beyond pharmacological interventions, digital therapeutics represent an emerging modality. CT-155 (BI 3972080), a prescription digital therapeutic developed by Boehringer Ingelheim and Click Therapeutics [86], targets experiential negative symptoms through smartphone-delivered psychosocial interventions. The Phase III CONVOKE trial (NCT05838625, n = 457) met its primary end point with a 2.6-point CAINS-MAP improvement over digital control (p = 0.0003, Cohen’s d = − 0.36) [87]. The trial employed centralized blinded rating, reducing site-level variability. Safety was favorable (8.3% adverse events versus 13.4% in control; zero serious adverse events). CT-155 received FDA Breakthrough Device Designation in January 2024 as the first prescription digital therapeutic with positive Phase III evidence for a core schizophrenia symptom domain [88]. Its success demonstrates that digital therapeutics can achieve clinically meaningful effect sizes comparable to pharmacological trials while offering scalability, treatment fidelity, and access to evidence-based interventions without medication-related side effects. The centralized rating methodology employed suggests approaches that may reduce site-level variability affecting pharmacological trials. Notably, CT-155’s effect size (Cohen’s d = − 0.36) falls within the range of recent pharmacological trials for negative symptoms, demonstrating that digital therapeutics can achieve competitive efficacy while offering advantages in scalability and treatment fidelity.

Failed and Discontinued Programs: Learning from Setbacks

Emraclidine’s Phase II Failure

Emraclidine (CVL-231, Cerevel Therapeutics & AbbVie), an M₄ muscarinic positive allosteric modulator (PAM) that had shown promise in Phase Ib trials [89], failed to demonstrate efficacy in two Phase II trials (EMPOWER 1 and 2), as it did not separate from placebo on PANSS total scores [90]. Specifically, in EMPOWER-1 (n = 379), placebo demonstrated a 13.5-point PANSS improvement (95% CI: − 17.0, − 10.0), while emraclidine 10 mg achieved − 14.7 points and 30 mg achieved − 16.5 points—differences that failed to reach statistical significance. In EMPOWER-2 (n = 373), placebo showed an even larger 16.1-point PANSS improvement (95% CI − 19.4, − 12.8), compared with − 18.5 points for 15 mg and − 14.2 points for 30 mg emraclidine—again without significant separation. These exceptionally high placebo responses (13.5–16.1 points) eliminated drug-placebo separation despite drug effects comparable to Phase Ib [91]. The multicenter design and COVID-19 timing likely contributed to these magnitudes, a pattern explored comprehensively in Sect. 7. This failure contributed to AbbVie’s decision to scale back investment in psychiatric medicines following their acquisition of Cerevel Therapeutics.

Roluperidone: Nondopaminergic Negative Symptom Approach

Roluperidone (MIN-101, Minerva Neurosciences) represented a paradigm shift in targeting negative symptoms through triple receptor antagonism (5-HT_2A, sigma-2, and α_1A-adrenergic receptors) while completely avoiding dopamine receptor binding [92]. The compound demonstrated dose-dependent efficacy in a Phase IIb study (12 weeks, 244 patients, 36 sites) with the 32 mg dose achieving an effect size of 0.45 (p ≤ 0.024) and the 64 mg dose an effect size of 0.57 (p ≤ 0.004) on negative symptom measures versus placebo. Personal and Social Performance (PSP) improvement with 64 mg yielded an effect size of 0.59 (p ≤ 0.003), indicating clear drug-placebo separation with moderate placebo response [90]. Despite promising Phase II results showing significant improvements in negative symptoms (Cohen’s d = 0.57) [93], the Phase III trial (n = 513) narrowly missed its primary end point (p ≤ 0.064) [94]. In the Phase III study (MIN-101C07; 12 weeks, 513 patients, 61 sites), effects were attenuated compared with Phase IIb. In the intent-to-treat analysis, the 64 mg dose versus placebo yielded a treatment difference of − 0.85 on NSFS (Negative Symptom Factor Score) (95% CI − 1.75, 0.05), marginally missing significance with p = 0.064 (effect size 0.26). After excluding one site with 17 patients showing implausible data (no symptom variation, minimal adverse events), the modified ITT analysis achieved nominal significance: − 0.96 (95% CI − 1.89, − 0.03), p = 0.044. Secondary end point PSP showed significant separation in both ITT (p = 0.021) and mITT (p = 0.017) analyses [95]. The FDA issued a Complete Response Letter in February 2024 [96], citing insufficient evidence of effectiveness and requiring additional positive studies. This rejection underscored the substantial regulatory challenges in developing negative symptom-specific treatments, particularly given the lack of established outcome measures and unclear definitions of clinically meaningful change.

Pimavanserin: 5-HT2A Selective Approach Discontinued

Pimavanserin (Nuplazid^®, Acadia Pharmaceuticals), the first FDA-approved medication lacking dopamine D₂ receptor antagonism [97, 98], demonstrated mixed results in schizophrenia development before complete program discontinuation in March 2024. While the initial ADVANCE Phase II study (n = 403) met its primary end point for negative symptoms (p = 0.043, effect size = 0.21) [99], the definitive ADVANCE-2 Phase III trial (n = 454) failed to replicate these findings (p = 0.4825, effect size = 0.07) [100]. A separate Phase III ENHANCE study (6-week duration) also failed its primary end point of PANSS total score change (p = 0.094, LS mean difference − 2.1, 95% CI − 4.5, 0.4), though trends emerged for PANSS negative symptom subscale separation [101]. The compound’s highly selective 5-HT_2A receptor inverse agonist mechanism offered theoretical advantages for negative symptoms while avoiding traditional antipsychotic side effects, but the inability to demonstrate consistent efficacy led to development termination and reflects the recurring trajectory of Phase II success followed by Phase III failure observed across multiple negative symptom programs.

F17464: D₃-Selective Dopamine Approach

F17464 (Pierre Fabre) represented a pioneering approach as the first preferential dopamine D₃ receptor antagonist [102] to demonstrate clinical efficacy. Cariprazine, an established antipsychotic, also exhibits a notable D₃ partial agonism, highlighting the clinical relevance of modulating this receptor [103]. The compound F17464 achieved remarkable in vivo selectivity (79–98% D₃ occupancy while maintaining < 18% D₂ occupancy) and showed significant efficacy in Phase II trials (p = 0.014 for PANSS total score) with favorable safety characterized by no weight gain and minimal extrapyramidal symptoms [104]. Despite these positive results and validation of D₃-selective antagonism as a viable strategy, development was discontinued around 2019–2020, likely reflecting strategic and commercial considerations rather than scientific limitations.

Broader Patterns in Development Challenges

The comprehensive meta-analysis of TAAR1 agonists reveals a broader pattern of translational disalignment between preclinical promise and outcomes in clinical settings. While animal studies demonstrated substantial benefits (SMD = 1.01, 95% CI 0.74–1.27), these did not translate to comparable efficacy in human trials, where TAAR1 agonists underperformed relative to established antipsychotics [48].

Similarly, compounds including MK-8189 (PDE10A inhibitor) showed only modest trends toward improvement [105], highlighting that even mechanistically sound approaches may yield insufficient clinical benefit. This disconnect highlights fundamental challenges in modeling complex psychiatric conditions and underscores the need for improved preclinical-to-clinical translation strategies.

A summary of failed or discontinued schizophrenia drug development programs is provided in Table 2.

Table 2.

Notable failed or discontinued schizophrenia drug development programs

Compound name (manufacturer)	Mechanism of action	Clinical trial registration number (most recent) and recruitment status	Primary reason for failure/discontinuation	Key lesson learned
Emraclidine/CVL-231 (Cerevel/AbbVie)	M4 PAM	EMPOWER 1-3 program NCT05443724 (completed)	Failed Phase II (no separation from placebo); exceptionally high placebo responses	Allosteric modulation complexity; clinical efficacy not guaranteed despite theoretical advantages. Multicenter design and COVID-19 timing may have contributed to excessive placebo response
F17464 (Pierre Fabre)	Preferential dopamine D3 receptor antagonist	NCT02151656 (completed)	Discontinued despite positive Phase II results and favorable safety profile (likely strategic/commercial considerations)	Commercial realities of drug development: Scientific promise alone is insufficient for continued development. Strategic and commercial factors heavily influence development decisions
Iclepertin/BI 425809 (Boehringer Ingelheim)	GlyT1 inhibitor	CONNEX 1-3 program; NCT04860830 (completed)	Failed Phase III primary/secondary endpoints (cognition/functioning)	Translational gap: Cognitive test improvements do not always translate to functional outcomes; complex mechanistic challenges of GlyT1 inhibition (neuron-glia interactions); limited effects on EEG biomarkers suggest inadequate target engagement
Luvadaxistat/NBI-1065844/TAK-831 (Neurocrine Biosciences)	DAAO inhibitor	INTERACT (NCT03382639, completed) and ERUDITE (NCT05182476; terminated)	INTERACT: Failed negative symptoms primary endpoint, exploratory cognitive benefits at 50 mg ERUDITE: Failed to replicate cognitive signals from INTERACT	Narrow therapeutic window, regional DAAO limits, and individual variability hinder efficacy; simple co-agonist elevation may yield paradoxical effects
Pimavanserin/ATC N05AX17 (Acadia Pharmaceuticals)	5-HT2A inverse agonist	ADVANCE (NCT04531982, completed) and ENHANCE (NCT03121586, completed)	Phase III failed to replicate Phase II findings; program discontinued	Inconsistent efficacy; challenges in targeting specific symptom domains (negative symptoms) without broader D2 modulation
Ralmitaront/RO6889450 (Roche)	Selective TAAR1 agonist	NCT04512066 (completed)	Both Phase II trials failed to demonstrate efficacy; significantly less efficacious than risperidone in direct comparison	Critical pharmacological differences within TAAR1 class: ralmitaront exhibits ~30-fold slower association kinetics and lacks detectable 5-HT1A receptor activity compared to ulotaront; dual TAAR1/5-HT1A activity may be essential for clinical efficacy; TAAR1-selective approach insufficient despite mechanistic rationale
Roluperidone/MIN-101 (Minerva Neurosciences)	5-HT2A, Sigma-2, α1A-adrenergic antagonist (non-D2)	NCT06107803 (completed)	Narrowly missed Phase III primary end point; FDA Complete Response Letter (insufficient evidence)	Persistent challenge of negative symptom endpoints; lack of established outcome measures and clear definitions of clinically meaningful change Effect attenuation from Phase IIb to Phase III illustrates systematic erosion of effect size with increased trial scale; Pattern of Phase II success followed by Phase III failure

AE adverse event, DAAO d-amino acid oxidase, D2/3 dopamine receptor subtypes 2 and 3, EEG electroencephalography, EPS extrapyramidal symptoms, FDA U.S. Food and Drug Administration, GlyT1 glycine transporter 1, NDA New Drug Application (regulatory filing for approval), PAC peripherally acting anticholinergic, PAM positive allosteric modulator, PANSS Positive and Negative Syndrome Scale, PDT Prescription digital therapeutic (smartphone-delivered psychosocial interventions), SCZ schizophrenia, TAAR1 trace amine associated receptor 1, 5-HT serotonin receptor subtypes

The Systematic Challenge of Escalating Placebo Response Rates

Temporal Trends and Regional Differences

A critical FDA analysis by Gopalakrishnan et al. [106] documented the temporal trajectory of placebo response in schizophrenia trials, revealing a fundamental shift in trial dynamics over the past 15 years. Pre-2009, mean placebo response was − 6.4 PANSS points with treatment effects of − 8.6 points, yielding robust drug-placebo separation. Post-2009, mean placebo response increased to − 10.5 PANSS points (a 64% increase) while treatment effects declined to − 5.8 points (a 33% decrease). Trial success rates fell from 78% pre-2009 to 57% post-2009, directly reflecting this erosion of assay sensitivity.

Regional differences are particularly stark and have important implications for trial design. North American trials showed dramatic increases with placebo response rising from − 4.3 to − 8.5 points and treatment effects declining from − 9.0 to − 3.4 points (a 62% decrease in treatment effect). Multiregional trials remained more stable, with placebo responses of − 10.0 to − 11.3 and treatment effects of − 8.1 to − 6.4. These regional variations suggest that factors specific to North American trial conduct—including regulatory environment, patient expectations, recruitment practices, and site management quality—may be driving disproportionate placebo inflation.

Factors Contributing to Elevated Placebo Response

A meta-analysis by Czobor and colleagues examining negative symptom trials specifically identified multiple factors predicting higher placebo response [107]: industry sponsorship, larger total sample size, more recent publication year, smaller screening-to-baseline improvement, less severe baseline negative symptoms, more severe depression/anxiety symptoms, four versus three treatment groups, and greater numbers of post-baseline assessments. Their reanalysis corrected previous methodological overcorrections, demonstrating that placebo effects in negative symptom trials are moderate rather than large, but nonetheless sufficient to be problematic for detecting drug efficacy in adequately powered trials.

The mechanistic drivers of increased placebo response likely include: (1) proliferation of study sites with variable quality control and investigator experience; (2) increased patient expectations driven by direct-to-consumer advertising and internet access to trial information; (3) regression to the mean amplified by enrollment of patients during acute exacerbations; (4) rater drift and expectancy effects in open-label baseline periods preceding double-blind randomization; (5) improved standard care and supportive trial environments providing nonspecific therapeutic benefits; (6) selection of less severely ill patients with greater capacity for spontaneous improvement; and (7) coronavirus disease-19 (COVID-19) pandemic effects on recruitment practices, site monitoring, and patient engagement.

The Kantrowitz 2025 Meta-Regression: Differential Scaling and Effect-Size Deflation

A recent comprehensive meta-regression analysis by Kantrowitz et al. [108] provides mechanistic explanation for the pattern of Phase II success followed by Phase III failure. Examining 159 active/placebo comparisons across 13,000+ participants and eight mechanisms, the analysis demonstrates that placebo responses scale disproportionately with increasing study size and site numbers, while active treatment responses remain stable. This differential scaling creates systematic effect-size deflation in larger multicenter trials, fundamentally challenging assumptions underlying current power calculations.

Statistical modeling reveals that significance is most reliably achieved in smaller studies (≤ 150 participants) with fewer sites (≤ 10) and higher recruitment density per site. As trials scale beyond these parameters, excessive Type II error risk emerges from the nonlinear relationship between trial scale and differential placebo variance—not from insufficient statistical power per se. This explains the pattern observed across emraclidine, roluperidone, pimavanserin, and ulotaront, where Phase II efficacy signals failed to replicate in larger Phase III studies despite similar absolute drug effects.

Implications for Trial Design and Drug Development Strategy

The findings reveal that current power analytic approaches are fundamentally inadequate because they fail to consider consequences of sample size-dependent effect-size deflation. Results indicate excessive Type II error risk with increasing sample size, increasing site numbers, and decreased recruitment density per site. This explains why promising treatments demonstrating efficacy in Phase II trials fail in larger Phase III studies despite similar or even larger absolute drug effects—the issue is disproportionate placebo response growth, not loss of drug efficacy.

Kantrowitz and colleagues recommend focusing on smaller numbers of high-quality study sites with higher recruitment per site rather than proliferating sites to achieve target enrollment. Increasing recruitment density per site could minimize differential placebo scaling while maintaining adequate statistical power. A related published article by Yoon et al. [109], including Kantrowitz, addresses practical recruitment strategies enabling higher per-site enrollment as a step toward trial designs that mitigate differential placebo scaling. This work analyzed 554 individuals from their research database to identify recruitment factors conducive to higher-quality samples and more efficient site operations.

For drug development strategy, the analysis implies that site management quality, recruitment density, and site selection criteria should receive equal or greater attention than sample size calculations. The excessive Type II error risk in current large multicenter designs may be leading to premature abandonment of promising compounds. Re-evaluation of power calculations to account for effect-size deflation, combined with focus on fewer high-quality sites, represents a paradigm shift for future trial methodology. The work demonstrates that several mechanisms of action retain meta-analytic evidence of efficacy despite individual large trial failures, supporting continued investigation with improved trial designs rather than abandonment of mechanistically sound approaches.

Practice Effects and Cognitive Test Batteries

For cognitive endpoints specifically, an additional consideration is the contribution of practice effects to apparent placebo response. A systematic evaluation by Keefe et al. [110] demonstrated small practice effects (< 0.2 SD) in the placebo arm of cognitive trials using the MCCB, indicating that true learning from repeated test administration contributes modestly to score improvements. However, these practice effects are insufficient to explain the magnitude of placebo responses observed in recent large CIAS trials such as ERUDITE (luvadaxistat) and CONNEX (iclepertin). The larger placebo responses in these trials likely reflect the same design and operational drivers affecting symptom-based trials: site proliferation, increased heterogeneity, regression to the mean, and variable quality control rather than simple test familiarity. Furthermore, standardized measurement of cognitive function in clinical trials remains essential [8], as inconsistent assessment methodology across studies contributes to variability and complicates interpretation of cognitive outcomes. The finding that antipsychotics do not function as true cognitive enhancers, combined with high measurement variability and modest practice effects, underscores the substantial challenge of developing and validating treatments specifically targeting cognitive impairment in schizophrenia.

Synthesis: A Systematic Crisis in Trial Methodology

The collective evidence reveals that the field faces not merely a series of unfortunate compound failures, but rather a systematic methodological crisis. The pattern transcends individual mechanisms and compounds, affecting GlyT1 inhibitors, DAAO inhibitors, M₄ PAMs, 5-HT_2A modulators, and TAAR1 agonists. The common thread is not scientific invalidity of the mechanisms, but rather the erosion of assay sensitivity through escalating placebo responses that scale disproportionately with trial size and site proliferation.

This realization has profound implications: mechanistically sound compounds with demonstrated Phase II efficacy may be prematurely abandoned based on Phase III failures that reflect methodological limitations rather than true lack of efficacy. Conversely, the field’s declining success rate may be partly attributable to operational and design factors amenable to correction through modified trial architecture emphasizing site quality over quantity, higher recruitment density, improved rater training and oversight, and potentially biomarker stratification to reduce sample heterogeneity.

The success of CT-155 in this context is particularly instructive: the use of centralized blinded rating, a consistent digital intervention immune to site-level delivery variability, and a well-defined patient population may have contributed to its ability to demonstrate efficacy where pharmacological trials have struggled. These methodological elements—rather than fundamental superiority of digital therapeutics over pharmacology—may explain its success and offer lessons for future pharmacological trial design.

Research Gaps and Priority Areas

Biomarker Development for Patient Stratification

The failure of multiple promising compounds in late-stage trials, combined with systematic review evidence of modest effect sizes across diverse mechanisms, highlights the critical need for validated biomarkers for patient stratification and treatment response prediction [111]. The heterogeneity of schizophrenia in a longitudinal as well as cross-sectional perspective likely contributes to high failure rates, as demonstrated by both iclepertin’s inability to translate Phase II cognitive improvements to Phase III success, luvadaxistat’s failure to replicate INTERACT cognitive signals in the ERUDITE trial, and TAAR1 agonists’ limited overall efficacy despite mechanistic promise.

The precision medicine approach being pursued with pomaglumetad by Denovo Biopharma exemplifies the potential for biomarker-guided redevelopment. Their proprietary Denovo Genomic Marker (DGM) platform identified 23 single nucleotide polymorphisms (SNPs) in the HTR2A gene associated with treatment response [112], suggesting that genetic stratification could potentially rescue compounds that failed in broad population trials. It is important to note that mGlu2/3 agonists, including pomaglumetad, have been extensively studied and have consistently failed at lower doses across multiple trials. Substantial biomarker work has suggested that pomaglumetad was likely underdosed in prior trials, which may explain the initial failures. The biomarker-guided redevelopment strategy aims to identify optimal dosing and responsive patient subgroups, potentially addressing the underdosing issue that limited earlier development attempts.

Recent preclinical evidence suggests that TAAR1-mediated effects may be most pronounced under specific pathological conditions, such as hyperdopaminergic states, rather than in all patients [113]. This selectivity could explain both the favorable safety profile and modest efficacy signals observed in clinical trials. Similarly, insights into GlyT1 inhibitor failures suggest that the therapeutic window may depend on complex neuron-glia interactions and regional brain differences in glycine versus d-serine dependency for NMDAR activation [114]. Individual differences in baseline NMDAR glycine-site occupancy and d-serine metabolism may determine therapeutic response to DAAO inhibition, as suggested by luvadaxistat’s narrow therapeutic window [16].

Priority areas for future research include development of neurophysiological, neuroimaging, and molecular biomarkers that can identify patients most likely to respond to specific mechanisms [115]. For muscarinic agonists, biomarkers related to cholinergic dysfunction could guide patient selection. For TAAR1 agonists, despite limited overall efficacy, biomarkers related to presynaptic dopaminergic dysfunction or elevated dopamine synthesis capacity might identify responder subpopulations. For glutamatergic approaches, biomarkers distinguishing patients with primarily glycine-dependent versus d-serine-dependent NMDAR function could inform patient selection strategies. Near-term implementable biomarkers include CSF d-serine/glycine ratio measurement and genetic variants in DAAO and GlyT1, while longer-term development is needed for neuroimaging markers of NMDAR occupancy and in vivo assessment of neuron-glia metabolic coupling.

Functional Outcome Measures beyond Cognitive Batteries

The MCCB [116] may not adequately capture clinically meaningful improvements that translate into real-world functional gains. Therefore, the development of more sensitive and ecologically valid assessments is essential [117]. The experience with roluperidone, where Phase II negative symptom improvements failed to translate to Phase III success, further emphasizes the need for outcome measures that better predict real-world functional benefit. The high variability observed in cognitive measures across CIAS trials, explicitly cited as a factor in luvadaxistat’s ERUDITE failure, underscores the measurement challenges in this domain: While practice effects contribute modestly (< 0.2 SD) to score improvements in placebo arms, the larger issue appears to be genuine measurement instability and sensitivity to baseline characteristics, test administration conditions, and patient engagement. Moreover, the absence of standardized approaches to cognitive assessment introduces substantial variability across studies, hindering comparability and limiting the validity of meta-analytic synthesis. Establishing consensus guidelines for cognitive evaluation—encompassing standardized administration protocols, rater training, and rigorous quality control—remains an essential priority for advancing the field.

While objective neurocognitive end points, measured in controlled test situations, can serve as valuable biomarkers for demonstrating a drug’s mechanism of action and efficacy, they may not fully reflect real-world functional improvements. Consequently, greater emphasis should be placed on real-world functional outcomes, including independent living, social functioning, and quality of life, which may provide a better assessment of therapeutic benefit and are essential for estimating the true value of an effect. Digital health technologies offer opportunities for continuous monitoring and more sensitive outcome measurement, which can provide a more holistic and ecologically valid picture of therapeutic benefit, crucial for regulatory approval and clinical acceptance. The success of CT-155 using the CAINS-MAP scale, which specifically assesses motivation and pleasure domains with direct functional relevance, demonstrates that well-targeted, clinically meaningful outcome measures can detect treatment effects even in the challenging domain of negative symptoms. The scale’s focus on experiential negative symptoms with clear functional implications may have contributed to its sensitivity to therapeutic change.

Mechanism-Specific Considerations

Muscarinic Modulators

While KarXT’s approval demonstrates the feasibility of muscarinic approaches, optimal patient selection remains unclear. Understanding which patients benefit most from M₁/M₄ modulation versus other mechanisms requires further investigation. The failed adjunctive trial suggests limitations in combining muscarinic agonists with dopaminergic antipsychotics. The development of next-generation muscarinic approaches including ML-007 and NS-136 offers opportunities to address limitations of first-generation compounds through improved formulation strategies and enhanced selectivity.

TAAR1 Agonists

The disconnect between robust preclinical effects and limited clinical benefit (see Sect. 5.1) may reflect several factors: (i) species differences in TAAR1 expression and function, (ii) the selectivity of TAAR1 effects for pathological versus normal dopaminergic states, and (iii) challenges in translating preclinical models of psychosis to human disease complexity. For CIAS, further research is needed to determine if TAAR1 agonists offer specific cognitive benefits in defined patient subgroups. The documented large placebo effects in the DIAMOND trials, occurring during the COVID-19 pandemic era, suggest that at least part of the Phase III failure may reflect methodological challenges rather than fundamental lack of efficacy, supporting the case for biomarker-guided redevelopment rather than complete mechanism abandonment.

Glutamatergic Approaches

Post-iclepertin insights reveal that GlyT1 inhibition strategies face fundamental mechanistic challenges beyond simple dose optimization. The three hypotheses proposed by Singer and Yee highlight critical considerations: disruption of the astrocytic serine shuttle, interference with activity-dependent glycine release, and perturbation of cyclic GlyT1 operation [25]. Moreover, the failure of luvadaxistat yields critical information specific to DAAO inhibition, in particular (1) the inverted U-shaped dose-response relationship; (2) regional brain differences in DAAO expression; (3) differential effects on synaptic versus extrasynaptic NMDARs, mediated by d-serine versus glycine preferential binding; and (4) individual differences in baseline d-serine levels and NMDAR glycine-site occupancy. These insights suggest that successful glutamatergic enhancement may require approaches that preserve neuron-glia metabolic coupling and respect the dynamic regulation of NMDAR coagonists, combined with patient stratification based on baseline coagonist levels, receptor occupancy status, and genetic variants affecting NMDAR function and d-serine metabolism.

Negative Symptom-Specific Approaches

The failures of roluperidone and pimavanserin highlight fundamental challenges in developing treatments specifically for negative symptoms. Both compounds demonstrated mechanistic rationale and favorable safety profiles but failed to achieve consistent efficacy in adequately powered trials. These failures underscore the need for better understanding of negative symptom heterogeneity, improved outcome measures that capture clinically meaningful change, and combination approaches guided by robust mechanistic rationale, with particular focus on how these improvements might indirectly or directly impact cognitive function.

Trial Design and Operational Excellence as Research Priorities

Beyond Compound Development—Methodological Innovation

The systematic pattern of Phase II success followed by Phase III failure across multiple mechanisms necessitates recognition that trial methodology itself represents a critical research gap. The Kantrowitz meta-regression [108] demonstrating differential scaling of placebo versus active responses with trial size and site number provides an evidence-based framework for methodological reform. Priority research areas include: (1) optimal site selection criteria emphasizing investigator experience, recruitment capacity, and quality metrics over geographic distribution; (2) determination of the maximum number of sites compatible with maintaining assay sensitivity; (3) strategies for achieving higher per-site recruitment density, including community partnerships, patient registries, and early identification programs; (4) enhanced centralized rating and monitoring systems to reduce site-level variability; (5) novel statistical approaches accounting for effect-size deflation in power calculations; and (6) implementation science research on disseminating best practices for site management and quality control.

The success of CT-155’s centralized blinded rating approach offers a template potentially applicable to pharmacological trials. Remote assessment technologies, standardized video rating, and centralized expert raters may reduce the site-level variability that drives placebo response inflation. Similarly, adaptive trial designs allowing early identification and exclusion of sites with aberrant placebo responses could improve signal detection.

COVID-19 Pandemic Effects and Postpandemic Trial Conduct

Multiple failed trials occurred during the COVID-19 pandemic (emraclidine, ulotaront), with sponsors attributing failures partly to pandemic effects on trial conduct. While pandemic disruptions likely contributed to operational challenges and placebo inflation, distinguishing pandemic-specific effects from fundamental compound limitations remains difficult, particularly when Phase II signals were modest or mechanisms face theoretical challenges. Research quantifying pandemic impacts on trial parameters would inform whether reevaluation of pandemic-era failures is warranted.

Future Directions and Strategic Considerations

Short-Term Development Priorities (2025–2027)

Top-priority candidates for advancement include: (1) brilaroxazine—progressing toward regulatory filing based on sustained RECOVER trial efficacy and favorable 1-year safety data; (2) NBI-1117568—initiating Phase III trials following positive Phase II results demonstrating statistically significant PANSS improvements with favorable tolerability; (3) CT-155—advancing through the regulatory approval pathway as the first prescription digital therapeutic with positive Phase III evidence for negative symptoms, representing potential validation of an entirely novel treatment modality; and (4) next-generation muscarinic approaches (ML-007, NS-136)—advancing through Phase II development incorporating lessons learned from KarXT’s approval and clinical experience. Secondary priorities include: reassessing ulotaront’s development path considering meta-analytic findings and pandemic-era trial challenges, progressing evenamide toward Phase III trials for treatment-resistant schizophrenia, and advancing LB-102 into Phase III as a potential first-in-class benzamide for the USA market with emerging evidence of differentiated activity across multiple symptom domains including cognition and negative symptoms.

Understanding KarXT’s real-world effectiveness and identifying patient populations deriving the most benefit represents a critical short-term goal. Moreover, the integration of repurposed therapies such as varenicline into treatment algorithms offers immediate opportunities to address the substantial smoking burden in schizophrenia populations. However, given the accumulating evidence that varenicline does not provide consistent cognitive enhancement, its role should be focused specifically on smoking cessation rather than cognitive improvement.

Another critical short-term priority is implementation of methodological reforms in ongoing and planned trials. Adoption of the Kantrowitz recommendations—emphasizing fewer, higher-quality sites with greater recruitment density—should be incorporated into Phase III trial designs for compounds currently in Phase II development. This includes brilaroxazine, NBI-1117568, ML-007, and evenamide. Pharmaceutical sponsors should prioritize site selection based on demonstrated recruitment capacity, investigator experience, and historical quality metrics rather than primarily geographic distribution. Enhanced centralized rating systems and remote assessment technologies should be implemented where feasible to reduce site-level variability.

Furthermore, compounds that demonstrated Phase II efficacy but failed in Phase III trials conducted during the COVID-19 pandemic or in excessively large multicenter designs warrant reconsideration. Biomarker-guided redevelopment strategies, similar to the pomaglumetad approach, should be explored for mechanistically sound compounds such as ulotaront, emraclidine, and potentially even negative symptom treatments including roluperidone and pimavanserin. The meta-analytic evidence of mechanism-level efficacy despite individual trial failures supports continued investigation with improved trial designs rather than complete abandonment.

Medium-Term Innovation Areas (2027–2030)

The development of companion diagnostics and biomarker-guided treatment selection represents a critical opportunity for implementing precision medicine approaches. The heterogeneity of schizophrenia necessitates more sophisticated patient and disease stage stratification, particularly given the evidence that broad-spectrum interventions often yield modest effect sizes. The precision medicine redevelopment of pomaglumetad using genomic biomarkers provides a template for rescuing failed compounds through biomarker-guided patient selection.

Specific biomarker development priorities include: (1) for NMDAR modulators, assessment of baseline glycine-site occupancy, d-serine/glycine ratios, and genetic variants affecting NMDAR function to identify patients with adequate therapeutic window; (2) for TAAR1 agonists, neuroimaging markers of presynaptic dopamine synthesis capacity and striatal dopamine release to identify hyperdopaminergic subgroups; (3) for muscarinic modulators, markers of cholinergic dysfunction including acetylcholine levels, muscarinic receptor expression, and cognitive profiles suggesting cholinergic deficiency; and (4) for negative symptom treatments, distinction between primary deficit syndrome and secondary negative symptoms, with potential neuroimaging markers of reward pathway dysfunction and anhedonia-specific circuits.

Furthermore, while initial attempts have been challenging, strategic combinations of therapies targeting complementary mechanisms may offer significant benefits if guided by robust mechanistic rationale and appropriate biomarkers. These combinations could be advantageous by addressing multiple pathways, treating different symptom domains simultaneously, and overcoming monotherapy limitations, while carefully managing challenges such as drug–drug interactions and increased complexity. The failed adjunctive trial of KarXT added to existing antipsychotics highlights that not all combinations will provide additive benefits, and that interference between mechanisms (e.g., muscarinic agonism combined with dopaminergic antagonism) may produce antagonistic rather than synergistic effects. Future combination strategies should be guided by mechanistic compatibility, preclinical synergy data, and potentially sequential rather than simultaneous administration.

Finally, medium-term priorities should also include expansion of digital therapeutic approaches beyond negative symptoms. Given CT-155’s demonstrated efficacy, development of digital therapeutics targeting cognitive impairment, social functioning, medication adherence, and early psychosis intervention represents a promising avenue. The advantages of digital platforms—consistent delivery, scalability, absence of medication side effects, and reduced site-level variability—may be particularly valuable for symptom domains with high measurement variability and placebo sensitivity. Integration of digital therapeutics with pharmacological treatments in a comprehensive care model, rather than viewing them as competing approaches, offers potential for optimized outcomes across multiple symptom domains.

Long-Term Vision (2030+)

The long-term vision for schizophrenia treatment must embrace the neurodevelopmental origins of the disorder, creating opportunities for early intervention during critical developmental windows. This also includes exploring advanced therapeutic modalities such as gene therapy and targeted neurostimulation procedures as emerging fields of research. The convergence of pharmacotherapy with digital therapeutics, integrating cognitive training, real-time monitoring, and virtual reality applications, represents a potentially transformative frontier. This suggests a future where pharmacological treatments integrate into a broader, technology-enabled care system, enhancing treatment adherence, providing continuous data on patient progress, and offering personalized interventions (e.g., cognitive rehabilitation via VR). This implies the need for interdisciplinary collaboration among pharmaceutical companies, technology developers, and healthcare providers. Ultimately, the field should aspire to develop effective prevention strategies by intervening in at-risk populations before the onset of psychosis.

Regulatory and Commercial Landscape

FDA Guidance Evolution

The approval of KarXT signals regulatory receptivity to novel, nondopaminergic mechanisms and validates cognitive improvement as a meaningful end point within the context of broader symptom reduction. This development provides a clearer regulatory pathway for future drug development. The FDA Breakthrough Device Designation granted to CT-155 further signals regulatory recognition of novel modalities beyond traditional pharmaceuticals, establishing that digital therapeutics meeting appropriate evidentiary standards can achieve regulatory milestones comparable to pharmacological interventions. However, the modest effect sizes observed for several novel mechanisms in systematic reviews underscore the continued importance of demonstrating robust efficacy. The FDA’s rejection of roluperidone despite positive Phase II results emphasizes increasingly stringent evidence standards, particularly for negative symptom treatments where clinically meaningful change definitions remain unclear. Key regulatory trends include growing emphasis on the end point of real-world functioning and greater willingness to consider biomarker-supported development programs for patient stratification. Biomarker-supported development programs could streamline regulatory approval by reducing required trial sizes, increasing success rates, and identifying specific responder populations, thereby making drug development more efficient and targeted.

Market Access and Commercial Considerations

KarXT’s launch price of approximately US$22,500 annually [118] highlights the pricing pressures for novel therapies and the economic expectations of companies. Demonstrating clear differentiation from existing agents and establishing cost-effectiveness will be essential for securing market access. Besides, the role of health insurers and their increasing demand for value-based care will significantly impact market access, requiring therapies to demonstrate not only efficacy but also tangible improvements in patient outcomes relative to cost. The limited efficacy reported in systematic reviews for some novel mechanisms makes the demonstration of a clear clinical value paramount. Furthermore, the failure of specific adjunctive strategies suggests inherent challenges in proving incremental value, requiring future development programs to carefully consider market positioning and target patient populations. The commercial model for digital therapeutics including CT-155 remains to be established, but the prescription status and Breakthrough Device Designation suggest potential for reimbursement pathways similar to medical devices. The absence of manufacturing costs and potential for scalability may enable different pricing strategies compared with pharmaceuticals, potentially offering cost-effectiveness advantages despite demonstrated efficacy. However, reimbursement mechanisms for digital therapeutics remain less established than those for pharmaceuticals, creating commercial uncertainty that may affect investment in this modality despite clinical success.

Key translational and methodological barriers and their potential solutions are summarized in Table 3.

Table 3.

Translational and methodological barriers and solutions in schizophrenia drug development

Research gap	Underlying challenge	Strategic priority/proposed approach
Lack of patient stratification biomarkers	Heterogeneous illness; late-stage trial failures (iclepertin, luvadaxistat, TAAR1 agonists, emraclidine)	Use biomarker-guided stratification: NMDAR modulators—glycine-site occupancy, d-serine/glycine ratios, NMDAR variants; TAAR1 agonists—dopamine synthesis/release imaging; muscarinic agents—cholinergic markers and cognitive profiles; negative symptom therapies—reward-pathway imaging and deficit differentiation. Precision-medicine models (e.g., HTR2A SNP-guided pomaglumetad) exemplify this approach
Outcome measures not reflecting real-world functioning	Cognitive batteries (MCCB) and negative symptom scales fail to translate into functional benefit	Prioritize ecologically valid endpoints (independent living, social/occupational functioning); use digital health tools for continuous assessment; establish consensus guidelines for cognitive evaluation including standardized administration protocols, rater training, and quality control; greater emphasis on real-world functional outcomes over objective neurocognitive test performance
Negative symptom drug development failures	Unclear definition of clinically meaningful change; regulatory bar high (roluperidone, pimavanserin) with lack of established outcome measures	Standardizes outcome measures; explore combination therapies rather than monotherapy approaches; distinguish between primary deficit syndrome and secondary negative symptoms for patient selection; consider digital therapeutics as complementary modality offering scalability and treatment fidelity
Translational gap between preclinical and clinical findings	Robust effects in animals not reproduced in humans (TAAR1 agonists, GlyT1 inhibitors); complex system-level effects not captured by current preclinical models (neuron-glia interactions, astrocytic serine shuttle disruption, activity-dependent co-agonist release); Species differences in receptor expression and function	Improve preclinical models of glia-neuron interactions; adopt precision medizing approaches with biomarker-guided patient selection (e.g. pomaglumetad with biomarker-guided selection); develop functional biomarkers demonstrating target engagement early in development to de-risk later stages; focus on mechanisms preserving rather than disrupting dynamic regulation of receptor systems; consider state-dependent drug effects (e.g., TAAR1 preferentially modulates hyperdopaminergic conditions)
Escalating placebo response compromising trial assay sensitivity	Temporal increase in placebo response with declining treatment effects Differential scaling: placebo response increases disproportionately with trial size and site number while active treatment effects remain stable	Consolidate high-quality sites; strengthen investigator and site selection; use centralized/remote ratings; adjust for effect-size deflation; apply adaptive designs to manage placebo effects; and pursue biomarker-guided redevelopment of prior failed mechanisms
Combination strategies underexplored	Monotherapies yield modest effect sizes; adjunctive trials often negative; mechanistic interference possible (e.g. muscarinic agonism + dopaminergic antagonism may be antagonistic rather than synergistic)	Explore rational polypharmacy with mechanistic complementarity, guided by robust preclinical synergy data and biomarkers and careful safety monitoring; consider sequential rather than simultaneous administration to avoid mechanistic interference; focus on combinations targeting complementary pathways for different symptom domains
Trial methodology and operational excellence gaps	COVID-19 pandemic effects on trial conduct; variable quality control across proliferating study sites; rater drift and expectancy effects; site-level delivery variability in interventions; inadequate power analytic approaches failing to consider sample size-dependent effect-size deflation	Prioritize methodological innovation: optimize site selection and recruitment density, expand centralized monitoring, refine statistical models, apply implementation science for best practices, assess pandemic effects on trials, and treat trial design as equally vital as compound development

GlyT1 glycine transporter 1, HTR2A serotonin 2A receptor gene, MCCB MATRICS Consensus Cognitive Battery, NMDAR N-methyl-d-aspartate receptor, PANSS Positive and Negative Syndrome Scale, SMD standardized mean difference, SNP single nucleotide polymorphism, TAAR1 trace amine associated receptor 1

Conclusions

Schizophrenia drug development has reached an inflection point where methodological innovation matters as much as mechanistic innovation. The approval of KarXT validates nondopaminergic approaches and demonstrates that innovation succeeds when mechanistic plausibility aligns with optimized trial methodology. Conversely, the discontinuation of iclepertin, luvadaxistat, and multiple other mechanistically sound compounds reveals that the systematic erosion of trial assay sensitivity through escalating placebo responses has become a rate-limiting factor. Success in the post-iclepertin era will require not merely novel compounds, but fundamentally redesigned trials emphasizing site quality over quantity, biomarker-guided patient selection, and outcome measures reflecting real-world function.

Three strategic imperatives emerge for the clinical research community. First, precision medicine frameworks must replace broad population approaches: biomarkers—genetic, neurophysiological, and functional—are essential for identifying subgroups most likely to benefit from specific mechanisms, particularly for cognition and negative symptoms where unselected population effect sizes remain modest. Second, outcome measures must evolve beyond symptom rating scales to capture real-world functioning, independent living, and patient-centered goals that reflect genuine therapeutic benefit. The success of CT-155 using experiential negative symptom measures with direct functional relevance demonstrates that well-targeted assessments can detect treatment effects even in challenging domains.

Third, and perhaps most critical, trial methodology itself requires urgent reform. The Kantrowitz meta-regression demonstrates that placebo responses scale disproportionately with trial size and site proliferation while active treatment effects remain stable, creating systematic effect-size deflation that compromises even genuinely effective compounds. Implementation of methodological reforms—fewer high-quality sites with greater recruitment density, centralized rating systems, enhanced site selection criteria, and power calculations accounting for differential placebo scaling—represents an actionable pathway to preserve assay sensitivity. Without substantive reform, mechanistically sound compounds will continue to falter in late-stage trials, perpetuating high attrition rates and discouraging investment in innovation.

The therapeutic pipeline offers cautious optimism. Next-generation muscarinic modulators (NBI-1117568, ML-007, NS-136) build on KarXT’s validation of cholinergic approaches. Brilaroxazine’s multi-target serotonin–dopamine modulation demonstrates sustained efficacy with favorable safety. CT-155 establishes digital therapeutics as a viable modality for core symptom domains. Even compounds that failed in pandemic-era or oversized multicenter trials—including ulotaront, emraclidine, and negative symptom treatments—may warrant biomarker-guided reconsideration rather than complete abandonment, as meta-analytic evidence suggests mechanism-level efficacy despite individual trial failures.

Over the coming years, clinical researchers are uniquely positioned to close the translational gap by aligning three critical elements: mechanistic innovation targeting validated pathways, precision medicine frameworks identifying responsive patient subgroups, and reformed trial methodology that preserves signal detection. This integration, combined with outcome measures capturing real-world benefit, can move the field beyond symptomatic management toward interventions that genuinely restore function and quality of life in schizophrenia, ultimately improving outcomes for millions affected by this complex disorder.

Funding

Open Access funding enabled and organized by Projekt DEAL. This review received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Declarations

Conflicts of Interest

M.Z. has received research support from Reshape (Rovi, clinical trial) and travel grants and speaker fees from Otsuka Lundbeck, Rovi, and Boehringer Ingelheim. S.E. declares no conflicts of interest.

Ethics Approval

Not applicable.

Consent to Participate

Not applicable.

Consent for Publication

Not applicable.

Availability of Data and Materials

Not applicable.

Code Availability

Not applicable.

Authors’ Contributions

S.E. and M.Z. contributed equally to this manuscript, including conceptualization, literature review, analysis, drafting, and critical revision. Both authors have read and approved the final version of the manuscript and agree to be accountable for all aspects of the work.