Remission in schizophrenia spectrum disorders: A randomized trial of amisulpride, aripiprazole and olanzapine

one-year remission rate was calculated for 59 patients: 29 % (17/59). Antipsychotic drug naivety and low negative symptom load at baseline correlated highly with belonging to the remission group. Use of amisulpride was more probable to lead to remission than that of aripiprazole, but it was not more probable than the use of olanzapine (in per-protocol analyses). Negative symptoms showed the largest resistance to treatment. The lack of remission for the majority of the participants in this closely monitored antipsychotic drug trial is alarming and could act as a reminder that novel treatment principles are needed, especially targeted towards the negative symptoms in schizophrenia.


Introduction
Schizophrenia is a disorder that affects the individual, family and society in general (McCutcheon et al., 2020;Tajima-Pozo et al., 2015).The Global Burden of Disease study 2019 ranked schizophrenia as the 22nd among all causes of disability for the age group 25-49 years (GBD 2019 Diseases andInjuries Collaborators, 2020).The course of illness is highly variable and it is important to identify outcome predictors to optimize treatment effects.
Antipsychotic drug treatment is the main effective approach to treat the clinical symptoms of schizophrenia, with second-generation antipsychotics as the most widely distributed drugs (at least in Western countries), and with clozapine as the most efficacious agent (Huhn et al., 2019;McCutcheon et al., 2020).Corresponding to the treatment mechanism of blocking dopaminergic D2/3 receptors, patients with schizophrenia, as a group, may display dysfunctions in the brain dopamine systems.Striatal hyperdopaminergia is connected to delusions and hallucinations, and reduced dopamine transmission in cortical areas is associated with negative symptoms and cognitive dysfunction (McCutcheon et al., 2020).However, the blocking of striatal D2 receptors does not normalize brain dopaminergic function and D2 blocking drugs also cause side-effects limiting drug adherence.New treatment principles have been suggested including also non-dopaminergic receptor systems (Kaar et al., 2020).
Meanwhile, it is important to determine and optimize the relative effectiveness of current antipsychotic treatment, and several concepts are used to describe the different grades and areas of improvement in schizophrenia treatment.While "response" refers to the improvement of psychotic symptoms, "recovery" is the ultimate goal in schizophrenia treatment.It embraces not only the amelioration of psychotic symptoms, but also the functional improvement in important areas, such as the ability to work, live independently and have adequate social contacts.However, there is still no consensus in the definition of recovery.The third important term used to describe improvement is "remission" which is a general medical term implying partial or complete absence of symptoms.The National Cancer Institute defines for example partial and complete remission as the reduction or disappearance of signs and symptoms of cancer (National Cancer Institute).In the definitions of remission in psychiatric disorders, there is no differentiation between partial and complete remission.
Remission in schizophrenia is operationalized as the six-month sustained improvement of key psychotic symptoms and additional symptoms associated with schizophrenia, defined in 2005 by the Remission in Schizophrenia Working Group (Andreasen et al., 2005) and here referred to as the "consensus remission criteria".When using the Positive and Negative Syndrome Scale (PANSS) (Kay et al., 1987) to assess remission, three positive (P), three negative (N), and two general (G) symptoms are monitored: P1 (delusions), P2 (conceptual disorganization), P3 (hallucinatory behaviour), N1 (blunted affect), N4 (passive/ apathetic social withdrawal), N6 (lack of spontaneity), G5 (mannerisms and posturing), and G9 (unusual thought content).PANSS was developed upon the proposal of the distinction between the positive syndrome (composed of symptoms, which are superimposed on the mental status) and the negative syndrome (characterized by deficits in cognitive, affective, and social functions) in schizophrenia.The General psychopathology component was added in PANSS as an important supplement to the positive-negative assessment (Kay et al., 1987).This scale serves as a separate and parallel measure of the severity of schizophrenia and comprises a group of non-specific symptoms against which the severity of positive and negative manifestations is judged.
The consensus remission criteria were rapidly disseminated and utilized, and by 2010 (5 years after the proposal of the consensus remission criteria) >30 publications reported remission rates, 17 in multiple episode and 15 in first-episode psychosis/schizophrenia (Lambert et al., 2010).However, the results vary largely, depending on differences in selection of samples (acute vs stable patients, first-episode schizophrenia (FES) vs non-FES, inpatients vs outpatients, diagnoses other than schizophrenia, comorbid disorders, substance users, etc.), the duration and frequency of follow-up, the drop-out rate and the use of antipsychotic drugs (AlAqeel and Margolese, 2012;Lambert et al., 2010).An example is a systematic review from 2012 including 27 studies, which showed one-year remission rates for FES ranging from 17 % to 78 % (AlAqeel and Margolese, 2012).A more recent meta-analysis comprising results from 60 studies of FEP with more than one-year follow-up, found a pooled remission rate of 58 % (Lally et al., 2017).The mean follow-up was 5.5 years, and the duration criteria of remission were variable.An interesting finding of this study was the stability of the remission rate regarding the use of stringent criteria (57 % remission rate using the consensus remission criteria versus 59 % with the use of broader remission criteria).Also, remission rates without using the time criterion of six months have been published, giving higher remission rates (61 % vs 47 %) (Lambert et al., 2010).In a recent naturalistic study, the one-year remission rate was 18.7 % for patients with FES and 22.8 % in non-FES (Fountoulakis et al., 2020).In an earlier study from our group, we found a one-year remission rate of 32 % for patients with FES (Drosos et al., 2020).
In order to optimize treatment outcome, it is of high importance to identify predictors of non-remission.Some of them, like medication nonadherence and duration of untreated psychosis (DUP), can be modified and should be a focus in clinical practice.An overview study on remission in schizophrenia identified 12 studies that assessed predictors of remission using multivariate regression models (Lambert et al., 2010) and the six most relevant predictors were: (i) shorter DUP; (ii) better premorbid adjustment; (iii) lower psychopathology or illness severity scores at baseline; (iv) better functioning level at baseline; (v) early improvement in symptoms or functioning; and (vi) medication adherence during treatment.These predictors were found relevant even for long-term outcome studies in first-and multiple episode studies, indicating relevance for the overall outcome in schizophrenia.
This study is a substudy of BeSt InTro, a randomized controlled trial (RCT) of three atypical antipsychotic drugs in patients with schizophrenia spectrum disorders, excluding F23 diagnosis.Our main aim was to study the treatment effectiveness in terms of induced remission.Clinical predictors of remission and other factors correlating with remission were explored.A further aim was to study how antipsychotic treatment influenced the different dimensions of the defined remission conceptthe positive, negative and general symptoms domains, and which treatment goals the drugs did not fulfill.

Design and duration
The BeSt InTro-study is a 12-month rater-blind, head-to-head RCT of oral treatment with amisulpride, aripiprazole and olanzapine (Johnsen et al., 2020).Study medications were oral tablets and the doses prescribed ranged from 50 to 1200 mg per day for amisulpride, for aripiprazole 5-30 mg per day, and 2,5-20 mg per day for olanzapine.More information about the procedures regarding antipsychotic medication used in the study can be found in the overview article (Johnsen et al., 2020).
The assessment points (visits) were at baseline, after one week, three weeks, six weeks, 12 weeks, 26 weeks, 39 weeks and 52 weeks.
The inclusion took place between Oct 20, 2011, and Dec 30, 2016.Patients were included from four academic centers of psychiatry: Bergen, Trondheim and Stavanger in Norway, and Innsbruck in Austria.The institutions the patients were recruited from were both inpatient and outpatient units.Non-pharmacological, psychosocial interventions were offered to the patients from their treating clinicians according to the clinicians' evaluation of each case, but this information was not registered in our study.
Adverse events were assessed at all visits by means of the UKU Side Effects Rating Scale, patient administered version (UKU-SERS-Pat) (Lindstrom et al., 2001); open questions from the assessors about any adverse events since last visit; laboratory assessments including ECG, blood pressure and weight.Furthermore, the clinical monitor screened the medical records for any serious adverse events.
For the analyses in this substudy, pooled data from all treatment arms were used.

Study population
The full inclusion and exclusion criteria for our study can be found in the BeSt InTro primary outcome article (Johnsen et al., 2020).We included patients with a diagnosis within the schizophrenia spectrum according to the ICD-10 diagnoses F20-29, but excluded those with diagnosis F23 Brief psychotic disorder, which is characterized by symptoms of schizophrenia that resolve within a month.Eligible patients for the BeSt InTro study had symptoms of ongoing psychosis, as determined by a score of four or more on at least one of the following PANSS items: P1 (delusions), P3 (hallucinatory behaviour), P5 (grandiosity), P6 (suspiciousness/persecution) or G9 (unusual thought content).Thus, as the BeST InTro inclusion criteria (P1, P3, P5, P6, G9) did not fully overlap with the consensus remission criteria (P1, P2, P3, N1, N4, N6, G5, G9), patients theoretically could have been remitted at inclusion as also a score of four or higher on P5 or P6 (which are not included in the consensus remission criteria) was sufficient for inclusion.However, this was not the case and none of the included patients in this substudy was in remission at baseline.Moreover, two of our patients had a diagnosis of F21 schizotypal disorder, which is not characterized by definite and characteristic abnormalities at any stage.They were included based on their PANSS scores corresponding to ongoing psychosis.Lastly, patients were recruited in various stages of their illness (both first-episode and multiple-episode patients), and with various previous exposure to antipsychotics (both antipsychotic-naïve patients and previous users of antipsychotics).
The exclusion criteria were as follows: inability to understand the native language (Norwegian or German), pregnancy or breastfeeding, hypersensitivity to the active substance or to any of the excipients of the study drugs, prolactin-dependent tumors, phaeochromocytoma, concomitant use of medications which could induce torsade de pointes, use of levodopa, and known risk of narrow-angle glaucoma.Patients with suicidal ideation or substance misuse were not excluded from the study.

Randomization
Patients were randomly selected to receive one of the studied oral antipsychotics (1:1:1).The starting dose was decided by the attending physician and the treatment allocation was open for the patient and the members of the clinical staff, but not for the research personnel who performed the assessments in the study.The patients were randomized to a sequence of the studied drugs, where they were listed in a random sequence.These sequences were put in sealed envelopes and when a new patient was included a new envelope was opened by the attending physician.The patient was offered the first drug of the sequence and if this could not be used because of intolerability or previous inefficacy issues, the treating physician offered the next drug in the sequence.The same principle was followed if the second drug also could not be used.The first drug in the sequence defined the randomization group and was the basis of the Intention-to-treat (ITT) analyses.The drug which was given to the patient was the basis for the per-protocol (PP) analyses.More details about the randomization and masking can be found in the overview article (Johnsen et al., 2020).

Outcome measures
The outcome measure was the remission status at one-year follow-up and patients were categorized as "remitted" and "non-remitted", according to the consensus remission criteria.Additionally, the remission status was assessed at all assessment points, both with and without the time criterion, where this was applicable.The time criterion in the consensus remission criteria is six months (Andreasen et al., 2005).
Lastly, we wanted to examine the degree to which the scores of the different PANSS items included in the remission criteria improved during the study period.The lack of improvement contributed to nonremission, and we wanted to find differences among both the individual items and the three clusters of symptoms (positive, negative and general symptoms).

Data analyses and statistical methods
Because of the nature of the definition of remission, different ways of calculating the remission rate have to be used depending on the inclusion or not of the time criterion of six months.In addition, for the consensus remission criteria, different ways are used for defining remission than for non-remission.
To define a patient as remitted at a follow-up point according to the consensus remission criteria, scores at all available PANSS measurements in a 6-month period before that point had to be under four.It was also mandatory to have a PANSS measurement at the beginning and the end of this period, however, missing PANSS values at visits inside the 6months period were allowed.For non-remission, one PANSS measurement over four during the examined 6-month period was sufficient, which makes non-remission less complicated to calculate than remission.Thus, even if a PANSS score is missing for a certain time point, a patient can be classified as non-remitted at this point, if there is a PANSS score over four at a previous follow-up point.This is the reason for calculating the remission rate for 114 patients at six months with the consensus remission criteria versus 58 patients without using the time criterion of six months.
For the definition of remission status without the time criterion, we based our calculations on the highest PANSS value of all eight PANSS items included in the consensus remission criteria, for each follow-up point.
The patients who dropped out of the study were not taken into consideration when we calculated the remission rate without using the time criterion of six months.When we used the consensus remission criteria, patients who were not in remission at drop-out point were per definition not in remission for the six months after their drop-out, affecting by this way their remission status at later follow-points, even if they were then missing PANSS measurements.On the contrary, for patients who were remitted at drop-out, we did not define their remission status at later follow-up points, as we could not be sure if they would still be in remission or if they would have a relapse between these points.As an example, there were 12 patients who dropped out during the first six months of the follow-up and were remitted at drop-out point; these were not included in the analyses for the 6-month remission rate.
We accepted that patients were categorized as remitted at 26 weeks, even though they were psychotic at baseline, if they were in symptomatic remission at one week and until 26 weeks.
Since this is a substudy of a randomized trial of three antipsychotic drugs, it is of highly interest to compare these medicaments further regarding their possible predictive value for remission status.A logistic regression model was fitted to our data to investigate if medication could be used to predict the probability of remission.We included age, gender and medication as explanatory variables, and used remission status at one year as outcome variable.The data were analyzed with both a medication variable that divided the patients into groups according to the medication they were randomized to ̶ ITT analysis ̶ ̶ and with a medication variable that divided patients into groups according to which medication they received ̶ PP analysis.
The remission groups were compared at baseline by analyzing categorical and continuous variables with the use of chi-square tests and one-way ANOVAs in the IBM SPSS Statistics (version 24) and R (R core team, 2021).
To examine the impact of each PANSS item on remission status, we counted the number of patients with a score of four or higher for each of the eight PANSS items at each follow-up point and the percentage of patients that obtained a score lower than four at one year.We did the same for the group of positive, negative and general symptoms.
We also calculated how many participants had their first score as remitted at each assessment point using the PP dataset and excluding assessments obtained after the participant had stopped using the study medication.

Ethics
The study was approved by the regional ethical committees and competence authorities in Norway and Austria and was performed according to the Declaration of Helsinki.

Subjects
In the BeSt InTro study, 359 patients were assessed for eligibility and 144 of them were included and randomized to one of the study drugs (102 recruited in Bergen and 13 in Stavanger, five in Trondheim and 24 in Innsbruck).
P. Drosos et al.In this substudy, patients (n = 18) with the diagnosis F23 Brief psychotic disorder were excluded, and 126 patients at baseline and 49 patients with remission data at one-year follow-up were included in the analyses.The drop-out rate of the study was 61.1 % (77/126).From the 126 patients, in Bergen 86 were recruited, in Stavanger 13, five in Trondheim and 22 in Innsbruck.The mean age was 32.3 years, and 63 % of the patients were men.The percentages of the diagnoses at baseline were as follows: schizophrenia (F20) 67 %, delusional disorder (F22) 17 %, schizo-affective disorder (F25) 8 %, unspecified nonorganic psychotic disorder (F29) 6 %, schizotypal disorder (F21) 2 %, and other nonorganic psychotic disorder (F28) 1 %.
The demographic and clinical characteristics for each remission group and for the whole sample of this substudy are presented in Table 1.A flowchart with the treatment allocation at each site at baseline and at one year, including the drop-out rate for each site, is presented in Fig. 1.The drop-out rates were as follows: 56 % in Bergen, 38 % in Stavanger, 40 % in Trondheim, and 75 % in Innsbruck.

Remission using the consensus remission criteria
At one year, remission status could be defined for 59 patients using the consensus remission criteria and we found that 17 patients (28.8 %) were in remission (Table 2).The difference from the number of patients we could calculate the one-year remission rate without using the time criterion of six months (N = 49), occurred because non-remission (with time criterion) could be defined for some patients at one year without having data from their one-year follow-up.In these cases, our calculations were based on their symptom status at 26 and 39 weeks.

Remission without using the time criterion of six months
When we did not use the time criterion of six months, we could calculate the one-year remission rate for 49 patients, and it was 55.1 % (27/49) (Table 2).Using this method, the remission rates increased by time and seemed to stabilize between 26 and 52 weeks.The number of remitted patients without the time criterion remained stable from three to 52 weeks, while the number of non-remitted patients gradually decreased (from 126 at baseline to 22 at one year).When we calculated the number of new participants reaching remission at every follow-up point, it remained stable from week 1 to week 26, and then dropped significantly at 39 weeks (Fig. 3).

Drop-out
Of the 77 patients who dropped out, 51 (66.2 %) were not in remission at drop-out point.By adding the number of patients being in remission at one year (n = 27) with the number of them being in remission at previous drop-out points during the one-year period (n = 26), we find the total number of patients reaching remission at any time during the one-year follow-up (without using the time criterion of six months).The cumulative remission rate is 42 % (53/126).
The remission status at all follow-up points, the drop-out point and the status of medication with antipsychotics (medicated vs nonmedicated) are illustrated in Fig. 2. The proportion of non-medicated patients in the group with non-remission at last observation before drop-out point was 27.4 % (14/51) and the respective number for the patients who were in remission at last observation before drop-out point was 23.1 % (6/26).

Predictors of remission
In the logistic regression model, the three study antipsychotic drugs were compared with amisulpride as the reference drug.When the ITT model was used, the probability of remission for aripiprazole patients was smaller compared to amisulpride patients, but did not reach statistical significance (p-value: 0.070).For olanzapine patients there was also a smaller remission probability (p-value: 0.186), but no statistical significance.When using the PP model, the probability of remission for aripiprazole patients was smaller compared to amisulpride patients and did now reach statistical significance (p-value: 0.038).For olanzapine patients there was a smaller remission probability but still no statistical significance (p-value: 0.339).None of the demographic parameters at baseline showed group differences for remission at one year (Table 1).Lower level of symptoms as measured by CGI-S (p = 0.068), and better function at baseline as measured by GAF (p = 0.081) were more frequent at the remission group, however only reaching trend level for significance.It was more probable to belong to the remission group for patients who had not tried any antipsychotic drugs previously (p = 0.003) and for those with a low negative PANSS subscore at baseline (p = 0.025).

The development in the PANSS items scored to assess remission rate; portions scoring below threshold of 4
Positive symptoms as a group and especially delusions (PANSS P1) were the most frequent at baseline and conceptual disorganization (PANSS P2) was the one with the largest reduction after one year (100 % with a score below 4) (Table 3).The item with the least improvement after one year was passive/apathetic social withdrawal (PANSS N4).All  negative symptoms displayed less improvement than the positive items, with the largest reduction being in lack of spontaneity (PANSS N6): 53.3 %.None of the patients was non-remitted at baseline exclusively because of negative symptoms (all positive items <4 and at least one negative ≥4), but this was the case for eight patients at one year.Sixtyeight patients were defined as non-remitted at baseline because of at least one positive PANSS item ≥4 combined with all negative items <4, while this was the case for only nine patients at one year.No patients were defined as non-remitted solely because of having a general PANSS item ≥4 at baseline or at one year.

Doses of the study antipsychotics
Doses of the study drugs were generally within Summary of Product Characteristics (SPC)-approved ranges, and the defined daily doses (DDD) were about one.The serum levels reflected the doses used in a great extent, which may indicate satisfactory compliance (Johnsen et al., 2020).
We calculated the average prescribed dose for each antipsychotic drug at baseline (PP analysis), both for the remitted and the nonremitted group (Table 1).No statistically significant difference was found between the remitted and non-remitted group for each drug.
The mean prescribed dose for each antipsychotic drug, both for the remitted and non-remitted group, was calculated for the whole year of follow-up (Table 4).A statistically significant difference was found between the remitted and the non-remitted patients who used aripiprazole (p = 0.0027).The remitted patients used an average dose of aripiprazole, which was almost half of that used by the non-remitted patients.

Side effects burden of the study antipsychotics
According to UKU-SERS-Pat scores, side effects were generally mild, with few drug-drug differences.We found a substantial and comparable weight gain in all three drugs, as well as negative impact on serum lipid profiles, with very few drug-drug differences.These findings are of clinical concern and somewhat surprising given the more benign metabolic profile of aripiprazole as documented in RCTs and metaanalyses (Smith et al., 2019).More information about the side effects burden of the studied antipsychotics can be found in the appendix of the overview article (Johnsen et al., 2020).

Discussion
Our main finding is the one-year remission rate of 29 % according to the consensus remission criteria, which increased to 55 % when calculated without the time criterion of six months.It is rather alarming that the majority of patients do not reach adequate levels of sustained symptomatic remission during a closely monitored one-year treatment study with first-line antipsychotic drugs.This underlines that remission in schizophrenia is difficult to achieve with first-line treatment, and that also patients with fairly good response to antipsychotic treatment, as evaluated through the PANSS total score, can be non-remitted as evaluated with the consensus remission criteria.To achieve remission, the individual must not only show improvement in key positive psychotic symptoms, but also negative and general symptoms, and over a relatively long period.Moreover, the proportion of remission among patients who dropped out from our study (33.8 %, without the time criterion), was lower than that for the completers group (55.1 %, without the time criterion).This explains the fact that the cumulative rate of patients reaching remission at any time during the one-year follow-up (without the time criterion of six months) was lower than that at one year when not using the time criterion (42 % vs 55 %).As we did not record the reason for participants dropping out in this study, we can only speculate about the causes, and one of them may be nonremission.However, it is likely that the lower remission level in the dropouts is caused by the shorter duration of antipsychotic treatment, while other causes are also possible.When not using the time criterion, we found that the number of patients in remission remains almost the same at each point from the 3-weeks follow-up until the end of the year, while the proportion of remitted patients increases.This can be explained by the fact that most of the patients who drop out from the study are not in remission.We chose to calculate the remission rate (without the time criterion) by only taking into account the available PANSS data at each point and not the patients who have dropped out from the study.This is the reason for not using N = 126 as a denominator in our remission analyses.By using N = 126 as a denominator, we would Table 3 Impact of PANSS items included in remission criteria on patients' remission status.Number of patients with score of four or higher in the PANSS items included in consensus remission criteria and percentage reduction from baseline to one year.9.1 %) 3 (5.9%) 88.9 % n = Number of patients with score of four or higher.% = percentage of patients with score of four or higher in the PANSS item in the total sample of patients with data at this follow-up point.

Table 4
Total mean antipsychotic dose used during the first year (with PP method)."decide" a priori that the entirety of the patients who dropped out from the study would not be in remission at one year or at a certain follow-up point.The group of drop-out patients is highly heterogeneous regarding their drop-out point, their remission status at drop-out (66 % were nonremitted) and the medication status at last observation before the dropout point.When using the time criterion, the remission rate is increasing over time substantially, but the value of this finding is questionable.Given the needed time to sustain symptomatic remission of six months, it is not strange that very few patients could reach remission during the first six months of the follow-up.The fact that most of the patients who dropped out were non-remitted, contributes to the increasing remission rate over time.Another factor is that the period of six months is a sizeable fraction of the follow-up time in this study (50 % of the 12months follow-up), which may contribute to the overestimation of the remission rate.Another important finding is the high variation in the individual remission trajectories as illustrated in Fig. 2, which also shows that a portion of participants seem to rest around the remission threshold having varying scores in consecutive visits with or without using antipsychotic drugs.
The one-year remission rate of 29 % found in our study is low compared to most of the studies operating with the consensus remission criteria, although we have to show caution when interpreting these findings because of the heterogeneity of the studies (Leucht et al., 2013).At baseline, all participants had a score of four or higher in at least one of the PANSS positive items of the consensus remission criteria, and the respective numbers were only 44.4 % for negative symptoms and 53.2 % for the general symptoms included in the consensus remission criteria.This gap was eliminated at one year, where the numbers for the positive and negative symptoms at the 4-point PANSS threshold or above were similar (29.4 % vs 27.5 %).While positive psychotic symptoms resolve in the majority of patients by the use of second-generation antipsychotic drugs acting on dopamine receptors, the negative symptoms show a high resistance to this treatment also in our study.The percentage change from baseline of the number of participants showing remission in the three positive psychotic symptoms at one-year follow-up, was almost double than the respective number for the negative symptoms (70.6 % vs 38.1 %).When we examined separately the course of the psychotic symptoms included in the consensus remission criteria, we found that the most frequent symptom at baseline was delusions (PANSS P1), followed by hallucinatory behaviour (PANSS P3).The negative symptoms were not so prominent at baseline (range from 21 % to 28.2 %) but showed persistence until the one-year follow-up (from 9.8 % to 17.6 %).In a meta-analysis, the negative symptoms showed some improvement when antipsychotics were compared to placebo (Leucht et al., 2017), however other studies indicate these changes to be small and not clinically significant (Fusar-Poli et al., 2015).
Aripiprazole was found less probable to lead to remission than amisulpride in the PP analyses in our study, but not in the ITT analyses.This comes in line with previous findings from the BeSt InTro study, indicating the favorable use of amisulpride when response was used as primary outcome, though amisulpride was found more superior than both aripiprazole and olanzapine in that study (Drosos et al., 2022).Our finding is also indicative of the sensitivity of the PP analyses to show differences among medications, compared to the ITT analyses.Further, none of the demographic parameters could predict the remission status at one year in our study.Of the clinical variables, antipsychotic drug naivety and low negative symptoms load at baseline predicted belonging to the remission group.These two predictors are not included in those mentioned as the six most relevant predictors of symptomatic remission in a review of 12 studies using the consensus remission criteria (Lambert et al., 2010).However, drug naivety in this study most likely should be seen as a marker for FES.Studies have shown that these patients with FES have better treatment response than chronic patients (Zhu et al., 2017).
Another unresolved question in antipsychotic treatment is how long to treat before considering a switch.Our results show that the remission rate using the time criterion increased substantially from 26 weeks to 52 weeks, from 2.6 % to 28.8 %.This reflects the time needed to achieve full remission in antipsychotic drug treatment.On the other hand, about one third of the patients achieved symptomatic remission already after three weeks, and this remained stable until 12 weeks and increased even more thereafter, reaching 55 % at the end of follow-up.Internationally, and reflected in several guidelines for the treatment of schizophrenia, an antipsychotic switch is suggested after maximum 6-8 weeks on an optimum dosage when there is no or partial effect/response, but these outcomes are not further defined (Helsedirektoratet, 2013;NICE, 2014).On the other hand, the OPTiMiSE study, a multicenter three-phase switching study in FES, found no significant improvements in clinical outcomes when switching antipsychotics and the authors suggested a single antipsychotic trial for up to 10 weeks followed by the use of clozapine in individuals who do not reach remission (Kahn et al., 2018).The results from our study show that new patients reaching remission are being added at roughly the same level until the 26 weeks, indicating that continuing antipsychotic drug treatment on the same substance after 6-8 weeks can be useful.Though, the usefulness seems to be restricted to a minority of the sample after 26 weeks.A relative factor is the clozapine use, which is recommended for patients with schizophrenia not responding to two trials with non-clozapine antipsychotic drugs.This is incorporated in national and international treatment guidelines for schizophrenia.In our study, five patients switched to clozapine.The design of the study allowed the clinicians to switch from the randomized study antipsychotic either to another one of the three study drugs or to a different antipsychotic of their choice.Of the five patients who received clozapine, three completed the one-year followup, one dropped out at 26 weeks and the last one at 39 weeks.None of these patients were remitted at completion or drop-out point, so that the use of clozapine did not offer any benefits regarding the achievement of remission, though a longer follow-up of these patients is needed to extract any conclusions.
Taken together, the proportion of 29-55 % of participants reaching remission (depending on the inclusion of the time criterion) in this study illustrates clearly the limited efficacy of three representative atypical antipsychotic drugs, specifically in the domain of negative psychotic symptoms, but also that the effect on positive symptoms is less than perfect.Thus, there is an urgent need to develop new pharmacological strategies for the treatment of schizophrenia.Several new therapeutic agents have been studied during the last years and are currently under development, but their clinical usefulness remains to be determined.Some examples of new therapeutic agents under investigation are new immune regulative agents as add-on antipsychotic treatment (Khandaker et al., 2015), and muscarinic agonists such as KarXT (xanomelinetrospium) (Karuna Therapeutics; Shekhar et al., 2008).
Among the strengths of this study is the one-year follow-up and the relatively large number of follow-up points, especially in the first weeks of treatment with antipsychotic drugs, where most changes occur.Given the heterogeneity of the remission studies in schizophrenia, our study is in our knowledge one of the few examining the remission rate in several points during one-year follow-up, and also both with and without the time criterion of six months.This gives the opportunity to watch the trajectory of the remission rate over time and the impact of the time criterion.Many studies calculate remission in schizophrenia without a specified period of maintenance (Lambert et al., 2010).The clinical usefulness of such results is unclear, as shown from the comparison of our results from using and not using the time criterion of six months, and this discrepancy may be misleading.More studies on remission in schizophrenia should be performed using the full consensus remission criteria, to increase the clinical utility and generalizability of the studies.Our study was financially independent from pharmaceutical industries and was conducted in hospitals with high levels of expertise in both clinical practice and research in the schizophrenia field.The inclusion of four different hospitals from two countries is another strength of our study.
Our study had a high drop-out rate (about 61 %), which represents a major limitation for the interpretation of our results.Around two thirds of the patients who dropped out were not in remission at drop-out point.Non-remission can indicate increased symptom burden, which possibly affects negatively the ability and willingness of the patient to comply with the requirements of a drug trial over a prolonged period.The larger proportion of non-remitted patients among those who dropped out may contribute to the increasing remission rate over time (when not using the time criterion), which limits its interpretability.In addition, the percentage of non-remitted patients among those who dropped out is larger than the respective percentage among those who completed the oneyear follow-up (around two thirds versus 55 %).The above mentioned findings may represent attrition bias, which should be taken into consideration when interpreting our results.Attrition during follow-up in BeSt InTro was larger than other trials (Kahn et al., 2008), but not very different from that found in antipsychotic drug trials (Leucht et al., 2008).Attrition has not caused underpowering of BeSt InTro, as the study was powered to this rate of attrition (Johnsen et al., 2020).
Moreover, our study may have selection bias.In our study, like in many other drug trials, a proportion of patients are excluded, as they are not capable of providing informed consent.This is a common selection bias in studies of schizophrenia (Leucht et al., 2008).Attrition analyses in BeSt InTro indicate random missingness, and the sample at 52 weeks was representative of the sample at baseline.We believe that these findings represent the actual status at certain follow-up points, since we chose to present results only for patients for whom we had clinical data at these points (when not using the time criterion).
Another limitation of our study is the relatively small sample of patients with data on remission at one year, which can be explained by the high drop-out rate.Generally, small sample size in studies can lead to type II errors (i.e., false negative results).On the other hand, it is worth mentioning that the researchers consumed a considerable amount of time (more than five years) for the inclusion of eligible participants for the BeSt InTro study, and personnel from four psychiatric hospitals in two countries were recruited for this cause.The nature of the disorder challenges the inclusion of patients with schizophrenia in drug trials.In our study, standardized protocols were used (including PANSS to measure the symptom burden) and all PANSS raters were trained by the PANSS Institute until satisfactory interrater reliability and agreement with a "gold standard" was achieved.These advantages may compensate for the relatively small sample size of our study.
The fact that remission is more difficult to achieve than response, as well as the relatively small sample of patients in our study, may contribute to equaling out any possible differences among the studied drugs.Moreover, results for predicting remission when using amisulpride only reached statistical significance using PP and not ITT analyses.This finding may indicate that PP analyses are more sensitive than ITT analyses in showing differences among the drugs as it reflects the actual drugs chosen, but PP analyses are vulnerable to selection bias as the strength of randomization does no longer adhere.Larger studies are needed for the replication and validation of our results.Another limitation is the inclusion of both first-episode and multiple-episode schizophrenia, and both antipsychotic naive patients and patients who had used antipsychotic drugs prior to their inclusion in the study.Moreover, participants were also allowed to switch to another antipsychotic drug during the one-year follow-up, either one of the study drugs or a different one.On the other hand, these parameters add to the pragmatic quality of our study, as it is closer to everyday clinical practice.Lastly, more men than women participated in the study putting some limitations in interpreting the results with regard to sex differences.

Fig. 1 .
Fig.1.Flowchart with the study population at each site at baseline and at one year, including the drop-out rate for each site.

Fig. 2 .
Fig. 2.Course of the remission status, the use of AP drugs, and the drop-out rate at all visits during the study period.

Table 1
Baseline demographic and clinical characteristics in remission analyses.

Table 2
Remission status according to the consensus remission criteria and without the time criterion used.
(Andreasen et al., 2005)sample; n = number with characteristics.Remission criteria as defined by the Remission in Schizophrenia Working Group(Andreasen et al., 2005).P.Drosos et al.