INTERACT: a randomized phase 2 study of the DAAO inhibitor luvadaxistat in adults with schizophrenia

Deficits in N -methyl- D -aspartate receptor (NMDAR) signaling are implicated in the pathogenesis of schizophrenia. Luvadaxistat (TAK-831/NBI-1065844) is an investigational D -amino acid oxidase (DAAO) inhibitor that increases D -serine levels at NMDAR coagonist sites. INTERACT is a phase 2 randomized, placebo-controlled study that evaluated the efficacy and safety of three doses of luvadaxistat, covering a range of DAAO occupancy and D - serine levels, in patients with schizophrenia with persistent negative symptoms. The study included a 14-day, single-blinded placebo run-in period and a 12-week, double-blinded treatment period. The primary efficacy endpoint was the 12-week change from baseline in Positive and Negative Syndrome Scale — Negative Symptom Factor Score (PANSS NSFS). Secondary efficacy endpoints included the 12-week changes from baseline in Brief Assessment of Cognition in Schizophrenia (BACS) score and Schizophrenia Cognition Rating Scale (SCoRS) score. Safety endpoints included adverse event assessments. The full analysis set included all randomized patients ( N = 256 [placebo, n = 87; luvadaxistat 50 mg, n = 58; 125 mg, n = 56; 500 mg, n = 55]); 228 patients completed the study. No significant improvements in PANSS NSFS were observed at any dose versus placebo at week 12. Improvements were observed with luvadaxistat 50 mg versus placebo in cognitive endpoints: BACS composite score (nominal one-sided p = 0.031) and SCoRS interviewer total score (nominal one-sided p = 0.011). Luvadaxistat did not significantly improve negative symptoms of schizophrenia. However, luvadaxistat 50 mg met the pre-specified secondary endpoints for cognitive performance (BACS) and function (SCoRS), warranting further investigation in patients with cognitive impairment associated with schizophrenia. Luvadaxistat was well-tolerated in INTERACT, with no new safety signals observed.


Introduction
Schizophrenia is a complex psychiatric disorder characterized by positive, negative, and cognitive symptoms (Patel et al., 2014).Although the presence of negative or cognitive symptoms is not required for a schizophrenia diagnosis, most individuals with schizophrenia experience significant cognitive impairment and over 60 % experience ≥1 negative symptom (blunted affect, alogia, avolition, asociality, anhedonia) (Bobes et al., 2010;Patel et al., 2014;Zaragoza Domingo et al., 2015).Although positive symptoms tend to go through cycles of relapse and remission, cognitive impairment associated with schizophrenia (CIAS) and negative symptoms are often chronic, lack approved treatments, and contribute to ongoing functional disability in affected individuals (Maj et al., 2021;Mucci et al., 2021;Patel et al., 2014).
Deficits in N-methyl-D-aspartate (NMDA) receptor (NMDAR) signaling may underlie the pathophysiology of schizophrenia.NMDARs play a crucial role in learning and memory by regulating synaptic plasticity and are activated by the binding of glutamate and two obligatory coagonists: glycine and D-serine (Traynelis et al., 2010).Antagonism of NMDARs has been shown to induce schizophrenia-like psychosis and cognitive impairment in healthy individuals and worsen symptoms in schizophrenia (Domino and Luby, 2012).A meta-analysis of 20 studies found low D-serine levels in the blood of patients with schizophrenia compared with healthy individuals, and genetic studies have identified variants of NMDA signaling genes associated with an increased risk of schizophrenia (Cho et al., 2016;Hall et al., 2015).As such, approaches to restore NMDAR functionality remain at the forefront of schizophrenia treatment development.
Because direct agonism of NMDARs may lead to excitotoxicity, recent therapeutic efforts have focused on targeting D-serine and glycine levels (McCutcheon et al., 2020).Glycine transporter-1 (GlyT1) inhibitors increase NMDAR signaling and iclepertin has shown efficacy in a phase 2 CIAS trial (Fleischhacker et al., 2021), while bitopertin and AMG 747 have shown inconsistent efficacy in treating symptoms and cognitive deficits of schizophrenia (Bugarski-Kirola et al., 2017;Dunayevich et al., 2017;Umbricht et al., 2014).D-serine is proposed to be the primary NMDAR coagonist in the prefrontal cortex and hippocampal subfields and has a stronger affinity for the glycine binding site than glycine (Labrie and Roder, 2010;Matsui et al., 1995;Shleper et al., 2005).Evidence suggests that D-serine may act predominantly at synaptic NMDARs, whereas glycine acts at pro-apoptotic extrasynaptic NMDARs (Hardingham and Bading, 2010;Papouin et al., 2012).During development, D-serine is thought to preferentially regulate hippocampal synaptic NMDAR populations, regions that may be associated with the symptoms and pathophysiology of schizophrenia and cognitive deficits in the early phase of schizophrenia, to promote the maturation of glutamatergic synapses (Le Bail et al., 2015;Perez et al., 2021;Pittman-Polletta et al., 2018;Xiu et al., 2021).
A meta-analysis of 40 randomized controlled trials (RCTs) involving 4937 individuals with schizophrenia concluded that NMDAR modulators, including D-serine and glycine, significantly improved schizophrenia symptoms, particularly negative symptoms, as an adjunctive treatment to antipsychotics (with the exception of clozapine) (Goh et al., 2021).To date, most RCTs have investigated low-dose D-serine (30 mg/ kg), with improvements in positive and negative symptoms reported in two meta-analyses (Cho et al., 2016;Goh et al., 2021).However, neither a large study conducted in 195 individuals, nor an international study conducted in India and in the USA showed any significant efficacy with low-dose D-serine (Goh et al., 2021).An open-label dose-finding study found that dosages of 60 and 120 mg/kg/day were clinically superior to 30 mg/kg/day for treatment of CIAS, while all three doses improved positive and negative symptoms (Kantrowitz et al., 2010).Furthermore, pharmacodynamic studies found that higher D-serine serum levels were associated with cognitive improvements in individuals with schizophrenia (Hons et al., 2021;Panizzutti et al., 2019).
In rodents, D-serine has also shown efficacy at high doses (equivalent human dose of 60-120 mg/kg) (Meftah et al., 2021).There is a risk of nephrotoxicity in rats when treated with high-dose D-serine, which has not been observed in other species including humans, except for a single individual who had abnormal renal values when treated with D-serine 120 mg/kg (Kantrowitz et al., 2010;Meftah et al., 2021).Given this single adverse event, the double-blind RCTs of D-serine utilized a dosage of 60 mg/kg, which showed significant efficacy in treating total and negative symptoms (Kantrowitz et al., 2018).Together, these data suggest that high-dose D-serine may be required for effective treatment of schizophrenia symptoms in humans.
An alternative approach to increase D-serine levels in the human brain is by inhibiting D-amino acid oxidase (DAAO), the enzyme responsible for converting D-serine into L-serine which is highly expressed in the cerebellum and brainstem (Jagannath et al., 2017).Luvadaxistat (TAK-831, NBI-1065844) is a highly selective and potent DAAO inhibitor with the potential to influence hippocampal-and cerebellar-based learning, and improve negative and cognitive symptoms, based on animal data (Fradley et al., 2019;Serrats et al., 2021).In rodents, luvadaxistat dose-dependently increased D-serine levels in plasma, the cerebellum, and cerebrospinal fluid (CSF) (Fradley et al., 2019;Serrats et al., 2021).Hippocampal long-term potentiation (LTP) was induced by luvadaxistat at doses of 0.001 or 0.01 mg/kg, indicating increased synaptic plasticity, but higher doses (0.1 or 10 mg/kg) decreased LTP (Fradley et al., 2019;Serrats et al., 2021).Luvadaxistat also improved social interaction deficits in a negative symptom mouse model and performance on cognitive behavioral assays (eye-blink conditioning, novel object recognition, attentional set shifting) (Fradley et al., 2019).
In humans, luvadaxistat elevated D-serine CSF and plasma levels after single (10-1200 mg) and multiple (15-1200 mg) daily dosing in phase 1 studies (Xu et al., 2018) and showed almost full enzyme occupancy at 100-500 mg in a positron emission tomography study (NCT02716987).Pharmacokinetic (PK) and pharmacodynamic (PD) modeling showed that, at steady state, luvadaxistat 50 mg has a predicted enzyme occupancy of 50-90 % in the first 12 h after dosing (Berkhout et al., 2019).In a phase 2a study of a small number of patients with schizophrenia (O'Donnell et al., 2023), improvements in mismatch negativity, an NMDA-dependent biomarker associated with cognitive dysfunction in schizophrenia (Light and Braff, 2005), were observed with luvadaxistat 50 mg but not 500 mg (O'Donnell et al., 2023).However, there were no treatment effects of luvadaxistat on eyeblink conditioning, a cerebellardependent learning measure (O'Donnell et al., 2023).Similar improvements in mismatch negativity have previously been observed with D- serine (Kantrowitz et al., 2018).
Based on the clinical results for D-serine and luvadaxistat to date, it is hypothesized that small molecule DAAO inhibitors could be an effective therapeutic approach for negative symptoms and CIAS.This phase 2 study investigated the efficacy, safety, and tolerability of luvadaxistat as an adjunctive therapy in adults with schizophrenia with negative symptoms (ClinicalTrials.gov:NCT03382639).

Ethics
The study was conducted in accordance with the International Council for Harmonisation of Technical Requirements for Pharmaceuticals for Human Use Good Clinical Practice guidelines and the Declaration of Helsinki.All other ethics guidelines, applicable laws, and local regulations were followed.The protocol, amendments and the informed consent form were approved by all relevant local Institutional Review Boards and the Independent Ethics Committee before the study commenced.

Study design
INTERACT was a phase 2, randomized, parallel-group study conducted in Europe (Bulgaria, Czech Republic, Italy, Poland, Russian Federation, Serbia, Spain, and Ukraine) and the USA between January 2018 (first patient recruited) and January 2021 (last patient completed).Before randomization, participants were screened for ≤4 weeks and underwent a single-blind (participant), 2-week placebo run-in period.The randomization was stratified by age at screening (<35 and ≥ 35 years of age), since younger patients may have better treatment responses than older patients, likely as a result of fewer psychotic episodes and less overall time spent with schizophrenia (Naber et al., 2015).Eligible participants were randomized 3:2:2:2 via an interactiveresponse technology system to receive daily oral doses of placebo or luvadaxistat 50, 125, or 500 mg during the 12-week double-blind (participant/investigator) treatment period, respectively (Fig. S1).Changes to the conduct of the study are reported in the Supplementary Methods.

Study participants
Participants were eligible if they had received a diagnosis of schizophrenia (as defined by the Mini-International Neuropsychiatric Interview [MINI] 7.0.2) ≥1 year before screening, were aged 18-60 years, were clinically and symptomatically stable outpatients (no hospitalizations or worsening of symptoms 3 months before screening), and had a baseline Brief Negative Symptom Scale (BNSS) score of ≥28 (12 items, excluding item 4).Participants must have been receiving stable antipsychotic treatment (other than clozapine) at a 2-6 mg daily dose of risperidone equivalents for ≥2 months.Treatment with a second antipsychotic was allowed if it was used at a subtherapeutic dose for insomnia or anxiety.The plasma levels of antipsychotic were confirmed at screening or placebo run-in periods, and participants were excluded if antipsychotic plasma levels did not meet the minimum acceptable concentration.
Other key eligibility criteria included Positive and Negative Syndrome Scale (PANSS) item scores of ≤5 for delusions, hallucinations, excitement, grandiosity, suspiciousness, and unusual thought content, a PANSS item score of ≤4 for conceptual disorganization, and a Calgary Depression Scale for Schizophrenia (CDSS) total score of ≤9 at screening.A reliable adult informant, such as a family member or social worker, was required to provide feedback on functioning and study rating scales during clinic visits.
Exclusion criteria included the following: extrapyramidal symptoms, antipsychotic-induced parkinsonism symptoms measuring a score of ≥6 on the Simpson Angus Scale (excluding item 10, akathisia), lifetime diagnoses of schizoaffective disorder, bipolar disorder, or obsessivecompulsive disorder, and a recent history of panic disorder, depressive episode, or substance use disorder (excluding nicotine dependence).
Demographic information, including race as described by the participant and Hispanic ethnicity (US sites only), were obtained at screening.All participants and informants provided written informed consent after receiving a full explanation of the study.

Sample collection
PK and PD plasma samples were collected at the start of the placebo run-in period, on days 1 (before dosing; baseline), 28, 42, 56, and 84, or at an early termination visit.The PK/PD analysis sets included randomized participants who received ≥1 luvadaxistat doses during the treatment period and had available luvadaxistat plasma concentration data (PK analysis set) or ≥ 1 post-dose plasma D-serine and L-serine measurement (PD analysis set).

Evaluation tools
Trained clinicians administered psychometric scales to participants in the following order: MINI, PANSS, BNSS, CDSS, Brief Assessment of Cognition in Schizophrenia (BACS), Schizophrenia Cognition Rating Scale (SCoRS), and Clinical Global Impression (CGI).The MINI, PANSS, and BNSS ratings were video-recorded and subject to independent review for quality control purposes.Negative symptoms were assessed using PANSS and BNSS scores and cognitive functioning was assessed using the BACS scale and the SCoRS.Further details can be found in the Supplementary Methods and Table S1.

Prespecified efficacy endpoints
The primary endpoint was the 12-week change from baseline in the PANSS-Negative Symptom Factor Score (PANSS NSFS).The changes from baseline in PANSS NSFS at weeks 4 and 8 were used as secondary endpoints.Key secondary endpoints included the 12-week change from baseline in the BNSS total score, BACS composite score, SCoRS and PANSS total score and subscales.. Other secondary endpoints included the CGI-Schizophrenia-Severity (CGI-SCH-S) overall severity and negative symptoms scores, CGI-Schizophrenia-Improvement (CGI-SCH-I) overall severity and negative symptoms scores at week 12, and luvadaxistat plasma concentrations.
The assessment instruments used are described in the Supplementary Methods.The full analysis set included all randomized participants who received ≥1 luvadaxistat doses during the treatment period.

Safety endpoints
The safety analysis included randomized participants who received ≥1 dose of double-blind study drug.Safety data included the incidence of treatment-emergent adverse events (TEAEs), and evaluation of laboratory tests, vital signs, and electrocardiogram assessments.TEAEs were defined as events occurring after the first dose of double-blind study drug and ≤ 30 days after the last dose of double-blind study drug or early termination.Suicidal ideation or behavior was assessed systematically using the Columbia-Suicide Severity Rating Scale (C-SSRS).

Statistical analysis
A sample size of 234 participants was planned for the INTERACT study to provide ≥82 % power to detect ≥1 statistically significant dose at the 0.10 level (one-sided) with multiplicity correction for the multiple doses.This calculation assumed that 85 % of participants completed the week 12 assessment, and the effect size was 0.4 for the 125 and 500 mg doses.Participants were randomized 3:2:2:2 to placebo and luvadaxistat 50, 125, and 500 mg, respectively, to ensure that a substantial proportion of participants were allocated to placebo, mitigating potential nonspecific effects resulting from treatment expectations.Multiplicity was controlled across dose-arms by first testing the 125 mg QD and 500 mg QD dose-arms in parallel using Holm's method to control the overall type I error rate at the 0.10 level (one-sided), then testing the 50 mg QD dose-arm at the 0.10 level if at least 1 of other doses is found to be statistically superior to placebo.Secondary endpoint testing was not controlled for multiplicity and associated p values are nominal.The decision to report one-sided p values was to facilitate comparison with the one-sided significance level without the need for additional calculations.This decision was made before unblinding and was documented.The pre-specified significance level (one-sided, 0.1) was not changed.
The analyses for the change from baseline in PANSS total score, subscales and factors, BNSS 12-item (excluding item 4) total score, SCoRS interview total score, and BACS composite score were performed using the mixed model for repeated measures with baseline scores as a covariate.The pooled site, visit, treatment, and categorical age (randomization factor) were fixed factors; and treatment-by-visit and V. Murthy et al. baseline-by-visit interactions were included in the model.The analyses were performed using observed cases only (based on a Missing at Random Assumption).Mitigations were put in place to minimize the impact of COVID-19.Alternative methods for conducting visits, including video calls, telephone visits, and in-home study visits, were used per approval by the sponsor or designee to minimize the missing data.
Dose selection was based on target occupancy, elevation of D-serine levels in CSF, and efficacy in preclinical cognition assays seen in prior luvadaxistat studies (Berkhout et al., 2019), as well as observations from studies of GlyT1 inhibitors.
The changes from baseline in CGI-SCH-S score at week 12 and CGI-SCH-I score at week 12 were analyzed using a Cochran-Mantel-Haenszel test stratified by age and performed using observed cases only.

Efficacy
The primary endpoint of superiority to placebo in improving negative symptoms was negative.Comparisons of seven prespecified secondary endpoints were performed for all three doses, of which three were significant based on one-sided p values at the 0.025 level and unadjusted for multiple comparisons.Nominally significant findings were observed for cognition (SCoRS and BACS composite score) and CGI-SCH-I overall severity for the 50 mg dose only.
There were no statistically significant changes in overall schizophrenia symptom severity at week 12 for any luvadaxistat treatment group versus placebo, as measured by the change from baseline in PANSS total score (Table S3).No statistically significant changes were observed for CGI-SCH-I negative symptoms, CGI-SCH-S negative symptoms, or CGI-SCH-S overall severity following luvadaxistat treatment (Table S5).For CGI-SCH-I overall severity, a higher proportion of participants taking luvadaxistat 50 mg showed an improvement at week 12 compared with those taking placebo (one-sided p = 0.029) (Table S6); however, no such improvements were observed with luvadaxistat 125 and 500 mg.

CIAS
Significant improvements were observed with luvadaxistat 50 mg compared with placebo in the BACS composite score (one-sided p = 0.031), the SCoRS interviewer total score (one-sided p = 0.011), the SCoRS informant total score (one-sided p = 0.008), and the SCoRS global rating (one-sided p = 0.024), but not with luvadaxistat 125 or 500 mg (Tables S7 and S8).A moderate effect size of 0.4 was observed on these cognitive measures compared with the placebo group.

Positive symptoms
There were no statistically significant changes in positive symptoms at week 12 for any luvadaxistat treatment group versus placebo, as measured by the change from baseline in the PANSS Positive Symptom Subscale (Table S3).

Safety and tolerability
TEAEs were experienced by 76 participants (29.7 %) and 23 participants (9 %) were considered to have experienced a drug-related TEAE by the investigator.Most TEAEs were mild or moderate in intensity; three events were considered to be severe in intensity.Severe events of rhabdomyolysis (placebo, n = 1) and chronic cholecystitis (luvadaxistat 50 mg, n = 1) were not considered to be drug related.One patient taking luvadaxistat 125 mg experienced a severe worsening of schizophrenia that was considered to be a suspected unexpected serious adverse event and potentially drug related, which led to study-drug discontinuation.It was not possible to establish causality owing to recent recreational drug use (psychostimulant) by the patient, and the TEAE resolved after 2 weeks.For the placebo group, four serious TEAEs were reported (COVID-19, psychotic disorder, rhabdomyolysis, tooth abscess), one of which (psychotic disorder) led to study-drug discontinuation.
Other TEAEs that led to study discontinuation included drug-related moderate delusion and auditory hallucination (luvadaxistat 500 mg, n = 1), increased blood creatine phosphokinase levels (placebo, n = 1), increased liver-function test abnormalities (placebo, n = 1), and psychotic disorder and psychotic symptoms (placebo, n = 1) (Table S9).One patient reported an aborted suicide attempt (impulsive paracetamol b For the placebo and luvadaxistat 500 mg groups, data are from 86 and 54 participants, respectively. overdose followed by vomiting) three days after receiving their last dose of luvadaxistat 50 mg.The suicide attempt occurred on the same day as worsening of positive symptoms of schizophrenia.The investigator assessed the suicide attempt and considered it to be nonserious, not related to luvadaxistat and to have resolved on the same day.No clinically meaningful changes were observed in suicidal ideation or suicidal behavior as measured by the C-SSRS.There were no clinically significant findings in the safety analysis set to indicate additional cardiometabolic burden (Table S10).
Overall, TEAEs relating to psychiatric disorders occurred at similar frequencies in those taking luvadaxistat and placebo.Frequent TEAEs (those occurring in ≥5 participants [≥2 %]) were headache, insomnia, and increased weight.All frequent TEAEs were considered nonserious and mild or moderate in intensity (Table S9).No deaths were reported and no clinically significant findings in safety, laboratory tests, vital signs, or electrocardiogram assessments were observed across groups.

PK and PD
Mean plasma concentrations of luvadaxistat were dose-proportional for the three doses studied (Table S11).The placebo-adjusted mean D- serine and L-serine plasma concentrations were not proportional to  luvadaxistat doses, with luvadaxistat 125 mg treatment resulting in higher D-serine and L-serine plasma concentrations than luvadaxistat 500 mg at weeks 4, 8, and 12.

Discussion
Luvadaxistat did not meet the prespecified primary endpoint of superiority to placebo in improving negative symptoms of schizophrenia.Nevertheless, nominally significant effects were observed on several prespecified secondary endpoints.Individuals taking luvadaxistat 50 mg showed nominally significant improvements versus placebo in cognitive test performance and day-to-day cognitive functioning (based on input from the patient and a close contact of the patient).Overall, luvadaxistat was well-tolerated in INTERACT, with safety findings in line with those reported in previous studies in healthy volunteers (Xu et al., 2018).The findings from INTERACT support additional clinical research to confirm the potential efficacy signal observed for CIAS.
Currently, no pharmacological treatments are approved for CIAS.The novel GlyT1 inhibitor iclepertin (BI 425809) was shown to improve cognition in participants with schizophrenia in a phase 2 study (Fleischhacker et al., 2021).Given that both glycine and D-serine act as modulators of NMDAR, the findings of INTERACT support the hypothesis that activation of NMDAR via the allosteric modulatory site may lead to beneficial therapeutic effects on CIAS (Kantrowitz et al., 2010;Lane et al., 2013).The potential efficacy signal observed for CIAS is being investigated further in individuals with schizophrenia with cognition as the primary endpoint (ERUDITE study, ClinicalTrials.gov:NCT05182476).
In INTERACT, an inverted-U dose response was observed, with the strongest effects at the 50 mg dose, which has a predicted enzyme occupancy of between 50 % and 90 % at steady state (Berkhout et al., 2019).The effects decreased with the two higher doses, which have demonstrated almost full target occupancy (NCT02716987) (Berkhout et al., 2019).These findings are consistent with the results of a phase 2a study showing that luvadaxistat 50 mg, but not luvadaxistat 500 mg, improved mismatch negativity (O'Donnell et al., 2023).While underpowered for this endpoint, the mismatch negativity study also explored the effects of luvadaxistat on BACS after 4 weeks of treatment; however, the results were deemed uninterpretable owing to variability and repeated treatment effect (O'Donnell et al., 2023).Because the results of the phase 2a study were not available before the statistical analysis plan was finalized, INTERACT focused on the luvadaxistat 125 and 500 mg doses.Similar effects have also been observed in vitro in rodent models, in which elevated D-serine levels following subchronic dosing with lower doses of luvadaxistat increased LTP in the hippocampus, whereas higher doses of luvadaxistat decreased LTP (Fradley et al., 2019;Serrats et al., 2021).An inverted-U dose response has also been seen in other compounds with this mechanism of action, notably GlyT1 inhibitors and D- serine (Fleischhacker et al., 2021;Sehatpour et al., 2023), and may be linked to clathrin-based internalization of NMDARs at higher target occupancy (Dunayevich et al., 2017;Umbricht et al., 2014).In particular, d-serine and iclepertin have shown inverted U-shaped dose responses in neuroplasticity-based auditory learning and cognitive impairment, respectively (Fleischhacker et al., 2021;Sehatpour et al., 2023).
The negative results for PANSS NSFS and BNSS total score suggest that luvadaxistat may not be effective at treating negative symptoms at the doses or treatment duration studied.Other NMDAR modulators such as D-serine, glycine and sarcosine have shown significant effect sizes for treating negative symptoms based on standardized mean differences of − 0.49, − 1.27and − 0.66 for the PANSS negative symptom subscale, respectively (Goh et al., 2021).However, pro-glutamatergic agents can show mixed results in treating negative symptoms.For example, the GlyT1 inhibitor bitopertin showed no evidence of efficacy as an adjunctive treatment in phase 3 clinical studies, despite findings in a phase 2 study showing a potential U-shaped dose response for efficacy in negative symptoms (Bugarski-Kirola et al., 2017;Umbricht et al., 2014).
TEAEs were mostly mild or moderate in severity, and TEAEs relating to psychiatric disorders occurred at similar frequencies in those receiving luvadaxistat and placebo.Headache and insomnia were the most frequently reported TEAEs in participants receiving luvadaxistat, both of which occurred at rates similar to placebo.
Noteworthy strengths of the INTERACT study include the doubleblind, placebo-controlled design, range of target occupancy evaluated, and the independent review of the negative symptom measures.Furthermore, the placebo run-in period prospectively evaluated negative symptom score stability and screened out participants who were most likely to be nonadherent to the study drug.Study drug adherence was also monitored using a digital medicine application.Participants who responded to treatment during the placebo run-in period (decrease of ≥20 % BNSS total score) or did not meet thresholds for adherence or background antipsychotics (which was also verified via serum sample testing) were excluded before randomization.These findings should be considered in the context of the study population being enriched for individuals with negative symptoms, which was the primary endpoint, rather than cognitive impairment.
There are some limitations of the INTERACT study.The effects of doses <50 mg on negative symptoms or CIAS were not investigated, nor did the INTERACT study investigate long-term effects.The ongoing phase 2b ERUDITE study (ClinicalTrials.gov:NCT05182476) investigating the effects of luvadaxistat on cognitive performance with a 6 or 12-month open-label extension should address these limitations.
The analyses were performed using observed cases only, based on a Missing at Random Assumption.Although the study coincided with the COVID-19 pandemic, mitigation strategies were employed and widespread COVID-19 infection among participants was not observed.Major COVID-19 related protocol deviations (missed study site visits and assessments due to risk of COVID-19 exposure) were identified for two participants, both of whom were randomized to the placebo group.Sensitivity analysis excluding these two participants did not impact the outcome for the primary or secondary endpoints.When excluding the two participants impacted by COVID-19, the PANSS NSFS, BNSS total score (12-item), BACS composite score and SCoRS interviewer total score were largely unaffected, with the BACS composite score and SCoRS interviewer total score still reaching nominal significance at Week 12. Before unblinding, the study team did not observe factors besides those in the statistical model that strongly predict dropout, indicating that the Missing at Random Assumption was not violated.

Conclusion
Luvadaxistat was well-tolerated, demonstrated no new safety signals and a low rate of study discontinuation, and gave a potential signal of efficacy for CIAS, with adjunctive luvadaxistat 50 mg/day treatment resulting in a nominally significant improvement in BACS and SCoRS.Additional clinical research is warranted to confirm the potential efficacy signal for CIAS observed in INTERACT.

Table 1
Baseline demographics and clinical characteristics (FAS).