Single trajectory treatment response for predominant negative symptoms: Post-hoc analysis of a clinical trial with cariprazine and risperidone

Examining the heterogeneity of negative symptoms of schizophrenia contributes to the identification of available treatment targets. Generally, prior evidence classified three to four symptom treatment response trajectory groups over the course of positive symptoms, yet, no evidence exists regarding the heterogeneity of medium-term response to predominant negative symptoms. The current post-hoc analysis aims to identify the heterogeneity in negative symptom treatment response trajectories among patients with predominant negative symptoms who received either cariprazine or risperidone for 26 weeks. Treatment response was analyzed based on the: the Positive and Negative Syndrome Scale Factor Score for Negative Symptoms (PANSS-FSNS), and the Clinical Global Impression Severity (CGI – S) and Improvement (CGI – I) scales. To identify subgroups of patients with a similar course of treatment response, group-based trajectory modelling was utilized. Results demonstrated that in comparison with competing models, a single trajectory best described the treatment response of patients with predominant negative symptoms. The results indicate that patients with predominant negative symptoms with over ten years of schizophrenia respond rapidly to adequate treatment and follow a course of steady improvement.


Introduction
Negative symptoms are a core feature of schizophrenia, a chronic psychiatric disorder afflicting approximately 1 % of the population (Bleuler, 1950;Galderisi et al., 2018a).Being associated with increased functional impairment (Galderisi et al., 2018a;Milev et al., 2005), even more than positive symptoms, negative symptoms represent a serious unmet medical need (Rabinowitz et al., 2012).Patients with predominant negative symptoms also have poor functional outcomes and worse quality of life (García-Fernández et al., 2022).Therefore, negative symptoms are an important treatment target (Marder and Galderisi, 2017).
Heterogeneity is a key characteristic of negative symptoms and has been examined by various studies using cluster analysis (Fang et al., 2022).Studies have shown that there are distinct subgroups of negative symptom patients with different phenomenological profiles (Strauss et al., 2013).Evidence inidicates for two distinctive negative symptom sub-groups: one group with predominantly Avolition-Apathy symptoms and another with a predominantly Diminished Expression (Strauss et al., 2013).These groups significantly differed on clinically relevant external validators, including measures of functional outcome, premorbid adjustment, clinical course, disorganized symptoms, social cognition, sex, and ethnicity (Strauss et al., 2013).In addition, recent models of negative symptoms present five constructs including anhedonia, asociality, avolition, blunted affect, and alogia (Marder and Galderisi, 2017) and evidince indicates these are also heterogeneous (Paul et al., 2022).Using the Brief Negative Symptom Scale (BNSS) to assess the five negative symptoms domains in a sample of 220 outpatients diagnosed with schizophrenia or schizoaffective disorder, they found that 4 clusters emerge which were classified as: 1. low negative symptoms, 2. severe negative symptoms, 3. predominantly elevated blunted affect and 4. elevated avolition (Paul et al., 2022).These clusters significantly differed on external validators including clinical characteristics, neurocognition, and functional outcome (Paul et al., 2022).
Findings from these studies suggest that schizophrenia heterogeneity can be analyzed by negative symptom subtypes that have distinct clinical and neuropsychological profiles (Paul et al., 2022).Heterogeneity of negative symptoms remains a significant challenge for diagnosis and treatment, given recognised symptom clusters appear to have different underlying mechanisms, suggesting trials of novel agents should focus on specific negative symptom domains (Marder and Galderisi, 2017;Paul et al., 2022;Tandon et al., 2009).
In response to this challenge, schizophrenia studies have examined the extent of heterogeneity in negative symptom severity and treatment response over time.This may assist in deciding which medication to prescribe, when to consider switching or discontinuing medication, providing insight into underlying mechanisms, planning adjunctive pharmacological or psychosocial treatment, and treatment response over time.For instance, studies aimed to identify distinct subgroups based on their different response trajectories throughout the study period (Abdin et al., 2017;Austin et al., 2015;Baandrup et al., 2020;Chan et al., 2020;Chang et al., 2019;Chen et al., 2013;Gee et al., 2016;Levine and Leucht, 2014).As summarised in Table 1, most prior negative symptom heterogeneity research has focused on first-episode psychosis patients with mixed symptoms and reported between three to four groups on average, except one study where a two-symptom trajectory described the course of response (Baandrup et al., 2020).The outcome of most trajectory analyses was based on either the negative subscale of the Positive and Negative Syndrome Scale (PANSS) (Abdin et al., 2017;Baandrup et al., 2020;Chen et al., 2013;Gee et al., 2016) or the Scale for the Assessment of Negative Symptoms (SANS) (Austin et al., 2015;Levine and Leucht, 2014).The follow-up periods were between four weeks (Baandrup et al., 2020) and ten years (Austin et al., 2015;Chan et al., 2020), with the majority of studies focusing on a 1-to 3-year period (Abdin et al., 2017;Chang et al., 2019;Chen et al., 2013;Gee et al., 2016).Only one previous study examined patients with predominant negative symptoms and found the majority of patients to be non-responders (61 %) during a six week clinical trial of amisulpride compared with placebo (Levine and Leucht, 2014).In another study lasting four weeks, a subtle but steady decrease of negative symptom scores was observed over four weeks in first-episode patients (72 %) based on a different amisulpride study (Baandrup et al., 2020).Other long-term studies reported that most patients assumed a symptom trajectory of minimal and sustained negative symptoms amelioration.However, studies also identified patient groups that were characterized by non-response and/or frequent relapses.Hence, collectively, these studies appear to be consistent with the notion that schizophrenia is a heterogeneous disorder with substantial differences between patients regarding how and when they respond to treatment.Given the different findings regardging the number of treatment trajectories, identifying the magnitude of heterogeneity underlying different patient populations, particularly chronic patients with predominant negative symptoms, is still needed.The current post-hoc analysis aims to identify the extent of heterogeneity in negative symptom treatment response trajectories among patients with predominant negative symptoms who received either cariprazine or risperidone for up to 26 weeks.Cariprazine and risperidone were chosen for consideration based on prior studies that have demonstrated treatment response for both risperidone (Carman et al., 1995;Leucht et al., 2009;Mirabzadeh et al., 2014) and cariprazine (Earley et al., 2019;Krause et al., 2018;Németh et al., 2017).Cariprazine has a distinct mechanism of action, being the only antipsychotic with in vivo binding to D3 receptors in the brain (Stahl, 2017).Since D3 receptors are hypothesized to regulate the amelioration of negative symptoms, cariprazine is believed to be a drug candidate with good efficacy.Indeed a study by Németh et al., compared cariprazine and risperidone in patients with persistent, predominant, primary negative symptoms (Németh et al., 2017).The results of that study showed that cariprazine was statistically significantly better compared to risperidone in addressing negative symptoms, although both drugs showed negative symptom improvement based on the Positive and Negative Symptom Scale Factor Score for Negative Symptoms (PANSS-FSNS; (Németh et al., 2017)).Also, both drugs showed reductions in the severity of symptoms based on the Clinical Global Impression -Severity (CGI -S) and change as measured by the Global Clinical Impression -Improvement (CGI-I).However, statistically significant effects were observed in favor of cariprazine over risperidone: Patients with cariprazine were considered less ill from week 3 onwards (based on the CGI-S scores) and improved more than patients on risperidone (Németh et al., 2017).However, not all patients responded equally or at the same rate.Breaking them into subgroups might therefore help to understand the heterogeneous patterns of negative symptom treatment response.

Study design
The post-hoc analysis was based on a phase III, randomized, doubleblind, multicentre, 26-week clinical trial assessing the efficacy and safety of cariprazine compared to risperidone (Németh et al., 2017).The study was sponsored by the owner of a patent for cariprazine.After a 4week lead-in period, during which the patient's current antipsychotic treatment remained unchanged, patients were randomized to receive either cariprazine or risperidone once a day for 26 weeks.During the first two weeks of the trials, previous antipsychotic medication was down-titrated and discontinued on day 14, while cariprazine or risperidone was up-titrated to the target dose of 4.5 mg/day and 4.0 mg/day, respectively.This was a fixed-flexible dose study, meaning although investigators could down-or up-titrate within the dose range of 3-6 mg/ day (for both drugs) once (to decrease side effects or prevent impending deterioration), the investigators were also requested to maintain treatment on the target dose.The trial ended with a two week safety followup.

Participants
The trial participants were adults between ages 18-65 with a diagnosis of schizophrenia according to the Diagnostic and Statistical Manual of Mental Disorders, 4th edition (DSM-IV), with onset occurring at least two years before screening (Németh et al., 2017).Additional trial inclusion criteria were 1) being in a stable condition for at least six months prior to the trial (e.g., no psychiatric hospital admissions), 2) predominant negative symptoms for at least six months (based on medical records/investigator judgment), 3) having a score of at least 24 on the Positive and Negative Syndrome Scale Factor Score for Negative Symptoms (PANSS-FSNS) (Marder et al., 2011) with a score of four or more on at least two negative items such as blunted affect.Participants who had a history of non-response to risperidone and/or had taken risperidone within six weeks of the trial screening were excluded from the study.In addition, a PANSS factor score for positive symptoms exceeding 19 or having a score of four or more on two or more positive items, were participation ineligibility criteria (Fig. A1).The study protocol received approval from 9 central and 37 local independent ethics committees, which were associated with the 66 sites that enrolled at least one patient.The study adhered to the guidelines of good clinical practice and the principles outlined in the International Conference on Harmonisation.Written informed consent was obtained from all participating patients.

Measures
The primary measure of efficacy was change from baseline up to week 26 on the PANSS-FSNS scores with assessments at weeks 1, 2, 3, 4, 6, 10, 14, 18, 22, and 26 (Németh et al., 2017).The Marder factor score was chosen for these analyses a priori because it was the primary outcome measure of the original study.In the study it was chosen because it is one of the most established, well-cited and most often used score to assess negative symptoms on the PANSS (Hopkins et al., 2017).Additionally, the Marder factor score was originally used to describe risperidone's superiority over haloperidol on the negative factor score (Marder et al., 1997), hence for consistency with prior research, we used this score also in the present analyses.Additional measures were the Clinical Global Impression Severity (CGI -S) and Improvement (CGI -I) scales.

Statistical analyses
To identify subgroups of patients with a similar course of treatment response, group-based trajectory modelling was utilized (Levine et al., 2010;Levine and Leucht, 2010;Levine and Rabinowitz, 2010), like in a previous study (Levine and Leucht, 2014).This type of modelling identifies subgroups that are homogeneous on a specific outcome within the group and significantly dissimilar (i.e., heterogeneous) from other subgroups in substantial ways (Haviland and Nagin, 2005;Levine and Leucht, 2014).In the present case, models were fit to trajectory models of three different psychometric evaluation scores (PANSS-FSNS, CGI-I, CGI -S), separately for the two arms (cariprazine and risperidone) and the combined trajectory set.To identify the most parsimonious and hence the appropriate number of subgroups, the Bayes Information Criterion (BIC) was used, where lower values are a better fit (Levine and Leucht, 2014).For each outcome, the model fit was repeated to test 1 to 6 trajectories, accounting for dropout, and the resulting models were compared based on their BIC values; where the lowest BIC represents the most parsimonious trajectory model.For the best-fitting model, to illustrate the course of treatment response each week PANSS-FSNS, CGI-I, and CGI-S percentage improvement scores were plotted for the resultant treatment response trajectories.Improvements were calculated using the following equations: PANSS-FSNS: (100 × [score(0)-score (week)]/score(0)), CGI -S: (score(0)-score(week)), and CGI -I: (4-score (week)).The model covariates were age, sex, and baseline value (except for the CGI-I scale, due to its constant baseline).A treatment arm and time dependent dropout assumption was employed in the model.All models were computed in SAS Proj traj and accounted for dropout.

Characteristics of sample
The baseline characteristics of patients are presented in Table 2.There were 227 patients in the cariprazine (CAR) and 229 in the risperidone (RIS) arm (intention-to-treat group).The mean participant age was 40.1 years in the CAR and 40.8 years in the RIS group, with 54 % (CAR) and 61 % (RIS) of the patients being male.The average time from the initial schizophrenia diagnosis to joining the trial was approximately 11 years in both treatment arms, with mostly under five prior psychotic episodes (CAR: 64 %, RIS: 55 %).The baseline PANSS-FSNS score was 27.7 in the CAR and 27.5 in the RIS arm.Regarding the CGI -S, the mean scores were 4.1 for the CAR and 4.2 for the RIS group, meaning that patients were moderately ill.

Treatment response of patients with predominant negative symptoms
The total PANSS-FSNS percentage reduction as a single trajectory was more parsimonious compared to solutions with two to five trajectories; based on the BIC value (BICs were: 1 = − 18,381, 2 = − 17,106, 3 = − 16,601, 4 = − 16,325, 5 = − 16,350, 6 = − 16,215).The single group trajectory in Fig. 1 assumed initial amelioration of reaching 15 % at week 4 and steady amelioration to week 24.Similar results were acquired when analyzing the CGI-S (Fig. 2) and CGI-I (Fig. 3 3).The observed mean values increased monotonously, with the largest response increment occurring between by week 4.

Discussion
The current study is the first trajectory analysis of a randomized, double-blind clinical trial that compared two antipsychotic medications in patients with predominant negative symptoms.The results indicate that a single trajectory described the treatment response pattern of patients irrespective of the type of scale used for the analysis.Most improvements occurred within the first four weeks of treatment, then a monotonous amelioration of negative symptoms characterized the patients.When comparing cariprazine to risperidone, the trajectories had a similar form, although greater improvement was observed with cariprazine than risperidone, as previously assessed in a separate analysis (Németh et al., 2017).
The current study results do not align with the existing literature regarding the heterogeneity of negative symptoms.Existing studies reported three to four trajectories when analyzing treatment response or symptom severity of negative symptoms in schizophrenia patients (Table 1).This may be due to notable differences between the current and past study populations and methods of analyses.Most research focused on drug-naïve, first-episode patients in the past (Abdin et al., 2017;Austin et al., 2015;Baandrup et al., 2020;Chan et al., 2020;Chang et al., 2019;Gee et al., 2016).In contrast, the stage of the disorder of the participants in the current study was rather chronic and not drug naïve.Moreover, prior study participants had different initial levels of negative symptom severity.In contrast, in the present study, participants were included based on a certain severity of predominant negative symptoms.Importantly, the inclusion criteria of the one other predominant negative symptom study (Levine and Leucht, 2014) was more comparable to the present study: patients were included if having a SANS score of 60, 75 or above and were not first episode patients (indicating moderatly to severely ill patients) (Levine and Leucht, 2013).The SANS is a 25-item, 6-point scale (0 to 5), measuring the five domains of negative symptoms (Andreasen, 1983).With a mean of 79.5 SANS total score in the amisulpride studies (Levine and Leucht, 2014), the patient population was quite similar in terms of negative symptomatology to the present study (moderetly to markedly ill patients - (Németh et al., 2017); (Leucht et al., 2019).In the present analysis  however, patients were much older (average age 40 vs 34) and were investigated for a longer time period (26 weeks) which is more appropriate for analyzing predominant negative symptoms.Furthermore, no wash-put period was included, whereas in the amisulpride study previous antipsychotic was down-titrated 6 weeks prior randomisation.These differences may provide an explanation for the different trajectory numbers acquired.Furthermore, the treatment duration in the current study was shorter compared to the majority of prior trajectory studies where researchers followed patients for either a 1-3-year or even a 10-year period.Generally, evidence and clinical lure suggest that negative symptoms become more dominant than positive symptoms over time, and then they remain persistent (Fountoulakis et al., 2019).Hence, plausibly more trajectories may emerge if one investigates a protracted time period.When looking at the short-term studies, however, prior research identified three treatment trajectories in a 6-week trial (Levine and Leucht, 2014), and two treatment clusters in a 4-week trial (Baandrup et al., 2020).Patients were either antipsychotic naïve first episode patients, or had restricted antipsychotic exposure prior to study inclusion (Baandrup et al., 2020).In addition, the trial was open-label, and the mean baseline PANSS negative subscale score was 19.4,much lower than the average PANSS-FSNS score in the present analysis (note: both the PANSS negative subscale and PANSS-FSNS consisted of 7 items with two items being different).Even given these differences, most patients (72 %) had a subtle but steady decrease in negative symptom scores and only 4 % had an unstable course of negative symptoms (Baandrup et al., 2020), which makes the results somewhat similar to the present findings.
To sum up, the main differences between the two past amisulpride studies and the present trial are: lack of wash-out period, PANSS-FSNS instead of SANS, older age, longer duration, and participanst were chronic schizophrenia patients.
Our analysis is not without limitations.First and foremost, the use of the PANSS scale for negative symptom evaluation is a limitation.The PANSS is not an ideal scale to assess negative symptoms as the items cannot measure the 5 negative symptoms domains as they are currently understood (Kirkpatrick et al., 2006;Marder and Galderisi, 2017).Nonetheless it is important to note that the PANSS is one of the most used scales for schizophrenia symptom assessment in general (Hopkins et al., 2017;Opler et al., 2017) and is preferred by regulatory authorities globally (Food and Drug Administration, European Medicine Agency) even for the assessment of negative symptoms ("Improving the Design of Clinical Trials of Drugs to Treat Schizophrenia," n.d.)Therefore, for the sake of regulatory approval as well as comparability with previous studies (see Table 1) the PANSS was used in the present analysis as well.
There have been multiple attempts to better cluster the individual PANSS items into factor models (Aboraya and Nasrallah, 2016;Marder et al., 1997) and the 5-factor structure (including positive, negative, cognitive, depression/anxiety, excitability/hostility domains), is the basis of most models including the Lindenmayer, Marder, Mohr and the Wallwork models (Lindenmayer et al., 1994;Marder et al., 1997;Mohr et al., 2004;Wallwork et al., 2012).Among these, the Marder factor score is most often used and considered "standard" for the past two decades for assessing efficacy in schizophrenia clinical trials (Hopkins et al., 2017).For negative symptoms, this one-dimensional structure has been questioned with newer models, ranging from 2-to 5-factors.More recent analyses provided evidence for a two-dimensional structure with experiential (avolition, apathy, and lack of energy) and expressive (blunted affect, poverty of speech) deficits (Khan et al., 2017;Liemburg et al., 2013;Purnine et al., 2000), among them the Khan model.Other studies have provided support for a 5 single-item structure, where anhedonia, asociality, avolition, blunted affect, and alogia are all separate factors (Ahmed et al., 2019;Galderisi et al., 2018b;Marder and Galderisi, 2017).A recent network analysis conducted by Demyttenare et al. with the same dataset as this present analysis found that the expressive and experiential factors of the Khan model were not distinct, and negative symptoms were better conceptualized as distinct negative symptom dimensions (Demyttenaere et al., 2022).Given the agents and selection criteria in our study and its primary outcome, we used the Marder subscale.Nonetheless, future research is advised to study the severity of negative symptoms with alternative measures.
Secondly, some study design related criteria may further contribute to the limitations: (1) patients had to meet the inclusion criterion that determined the minimum severity level of negative symptoms.Accordingly the population may have been less heterogeneous than in previous analyses and, therefore could not unfold into more treatment trajectories than one.(2) Regarding study duration, patients were followed-up until 26 weeks which is a relatively modest follow up time compared to prior analyses where patients were tracked for years.However, our study included multiple assessments and measures in that interval.(3) In addition, the current study is based on a clinical trial with selection criteria, hence the extent the results translate to the broader population with schizophrenia requires prospective examination.(4) Hawthrone, and placebo effects cannot be excluded in this study either and could potentially have had an effect on either compound.Moreover, according to a meta-analysis, a pervasive placebo effect is seen in negative symptoms of schizophrenia (Czobor et al., 2022).Hence, heterogeneity may have been reduced by placebo effects.However, these effects would impact both drugs equally, hence they would not impact the overall outcome of study results.Finally, the current analysis is post hoc; hence a prospective study is more appropriate.
Nonetheless, this is the first study to examine the heterogeneity of negative symptom treatment response in patients with predominant negative symptoms in patients with schizophrenia treated with cariprazine and risperidone.The main strengths of the study are the patient population and multiple measures and study duration.The duration of 26 weeks reinforced by multiple outcomes is ideal to consider treatment response, and no other study considered negative symptoms over this duration with multiple outcomes.

Conclusions
Collectively, our results appear to imply that patients with predominant negative symptoms with a history of over ten years of schizophrenia respond rapidly to adequate medication with cariprazine or risperidone and then assume a course of steady negative symptom improvement.In clinical practice, this may translate into psychopharamacological decision-making regarding treatment continuity.Potentially this reveals a profile of a chronic patient with persistent predominant primary negative symptoms not confounded by positive or depressive symptoms whose negative symptoms of schizophrenia are not heterogenous, and rather present a single and homogeneous symptom trajectory when treated with cariprazine or risperidone.

CRediT authorship contribution statement
PS conducted the statistical analyses.ZBD wrote the initial manuscript draft.SL, SZL, and AB revised the final manuscript.

Table 1
Previous studies analyzing negative symptom trajectories.
Note.Abbreviations.HEN, High Royds Evaluation of Negativity Scale; SANS, Scale for the Assessment of Negative Symptoms; PANSS, Positive and Negative Syndrome Scale.

Table 2
Baseline patient characteristics.