Introduction

Expert groups differ in their recommendations for early-stage schizophrenia treatment. Some expert groups, including American Psychiatric Association (APA), are non-directive and do not specify preferred agents for patients with first-episode psychosis1,2,3. This approach, while patient-centric, may challenge early practitioners and learners. Others offer algorithmic guidelines, but existing algorithms only partially agree on “first-line” treatments (see the section “Summary of first-line treatment options for patients without concurrent violence” for definition), reflecting the complexity of risk-benefit analysis4,5,6,7,8,9. In addition, updates to guidelines are essential to incorporate the latest research.

Based on up-to-date evidence (as of January 2024), we present a rationale for the selection of “first-line treatments” for patients with early-stage schizophrenia and challenges surrounding the selection of these agents. We present a general rationale, such that when idiographic factors of individuals dictate a different approach, the considerations discussed here should defer to individualized plans, and patients and practitioners should engage in shared decision-making at every step.

Initial patient stratification

After diagnosis of schizophrenia spectrum disorder is made and need for treatment established, patients in this algorithm are stratified into two groups (see the section “Treatment of patients with violence” for rationale) based on comorbid violence (in research studies typically defined as high score on the Positive and Negative Syndrome Scale (PANSS) hostility item or Modified Overt Aggression Scale10,11; henceforward “violence” will be used for simplicity). Figure 1 summarizes selection of antipsychotics proposed in this manuscript.

Fig. 1: Selection of antipsychotics in the treatment of patients with schizophrenia spectrum disorders.
figure 1

The figure summarizes the guiding factors and sequential options involved in the decision-making process for selecting antipsychotic medications. *In alphabetical order.

Full size image

Treatment of patients without violence

Initial treatment of patients without concurrent violence

Clinicians face challenges in balancing efficacy and side effects when prescribing antipsychotics to treatment-naive patients. Existing algorithms approach this issue by assigning high significance to a few side effects, such as weight gain and/or tardive dyskinesia, commonly leading to exclusion of agents such as olanzapine and first-generation antipsychotics (FGAs), respectively, as first-line treatments4,5,9. The approach here deviates from such a rationale. In our opinion, selection of first-line treatments should be guided by three key overlapping factors:

  1. (1)

    Overall efficacy: Randomized controlled trials (RCTs) comparing antipsychotics with long-term follow-up should be preferentially considered. Most RCTs are only several weeks long12. Since schizophrenia is a disorder requiring treatment far longer in duration, shorter duration RCTs are less relevant (though still important) and have less external validity compared to studies with longer outcomes3.

  2. (2)

    All-cause discontinuation: Discontinuation is typically influenced by perceived (in)efficacy and tolerability13. Instead of weighing risks of particular side effects against effectiveness, we use all-cause discontinuation rates as a surrogate measure of side effect burden versus effectiveness. All-cause discontinuation is ideally based on RCTs with long-term follow-up.

  3. (3)

    Mortality: Schizophrenia has one of the highest mortality risks of all psychiatric disorders14. While complex and difficult to study, mortality is a crucial outcome deserving attention for antipsychotic selection3,15. There is an emerging consensus that untreated psychosis has more adverse health effects compared to risks posed by antipsychotics16,17, but there is little guidance on how to weigh risk of discontinuation (and associated risks of untreated psychosis) versus long-term side effects. One could imagine a scenario of a weight-gaining antipsychotic with higher efficacy and lower overall discontinuation rate initially, but higher long-term mortality because of cardiovascular problems associated with weight gain over a longer period. Increased mortality may exclude this medication from first-line therapies. Conversely, if a weight-gaining antipsychotic has higher efficacy, lower discontinuation rates, and similar or lower long-term mortality relative to alternative treatments, there is no convincing reason to exclude it from first-line treatments. Thus, mortality data combines the above factors to allow determination of the “safest” first-line treatment choices.

Regarding criteria 1 & 2 above we discuss selection of key published literature here. With regard to criterion 3, RCTs generally do not offer long-term mortality data, and we discuss relevant mortality literature for each medication below.

There is only one large-scale network meta-analysis focused directly on comparing RCTs of antipsychotics with at least 6-month follow-up of acutely ill patients12. Amongst all non-clozapine antipsychotics available in the U.S., five antipsychotics—lurasidone, olanzapine, perphenazine, risperidone, and aripiprazole—rank highest for overall efficacy and lowest overall discontinuation12.

Aripiprazole and risperidone are among first-line treatment options in most existing algorithms, and we find their place here to be non-controversial4,5,7,9. Sections below will contrast the remaining three antipsychotics to aripiprazole and risperidone to demonstrate how our rationale diverges from existing protocols.

Olanzapine

In RCTs, olanzapine consistently demonstrates superior efficacy compared to risperidone or aripiprazole; evidence is of moderate confidence and small effect size12. In the Clinical Antipsychotic Trials of Intervention Effectiveness (CATIE) trial, the number need to treat for olanzapine versus risperidone was ~10; for olanzapine versus perphenazine it was 9; and for olanzapine versus quetiapine it was 5.518. While equivalent dose of olanzapine was higher relative to other agents in CATIE19, other trials with comparable doses confirmed favorable efficacy of olanzapine over risperidone20,21. In addition, olanzapine has a significantly lower discontinuation rate, indicating favorable efficacy-to-tolerability ratio12. In a Finnish cohort study of patients discharged from first schizophrenia-related hospitalization, those discharged on oral olanzapine had lower risk of rehospitalization versus oral risperidone22.

On the other hand, olanzapine is associated with significantly more weight gain; higher total cholesterol and triglycerides (compared to aripiprazole, lurasidone, and risperidone)12,23; and is more strongly associated with incident diabetes relative to risperidone24. Since cardiovascular disease affects patients with schizophrenia at high rates, such side effects deserve serious attention25. Studies in adults with mean duration of 42 weeks show that olanzapine leads to 3 kg more of weight gain compared to aripiprazole and about 2 kg more compared to risperidone12; younger antipsychotic-naive patients may show more pronounced differences26. In the CATIE trial, olanzapine relative to risperidone was associated with an increase of 1% in the 10-year coronary heart disease risk, which was statistically significant27. However, consistent with its favorable efficacy profile and the hypothesis that improved psychiatric status leads to healthier lifestyles, Solmi and colleagues found that olanzapine ranked best among non-clozapine antipsychotics with regard to patients’ adherence to cardiometabolic drugs28. While olanzapine is clearly more metabolically unfavorable than other potential first-line agents, the key question is whether olanzapine’s advantages outweigh its disadvantages. We will attempt to answer this question quantitatively based on most recent data on antipsychotic-associated mortality.

In a 1-year, open label, randomized trial, ziprasidone and olanzapine had the same rate of non-suicidal mortality in patients with schizophrenia29. A prospective Finnish register study with 5-year follow-up of patients with first-episode schizophrenia found decreased overall mortality with olanzapine relative to non-users of antipsychotics; this effect was not observed for risperidone30. An 11-year follow-up study in patients with schizophrenia from Finland compared several antipsychotics with perphenazine serving as a reference. Overall mortality in users of olanzapine was not significantly different compared to perphenazine, but mortality was significantly higher in patients on risperidone31. Risk of death from ischemic heart disease was similar between risperidone and olanzapine, though trended higher in risperidone31. Cumulative exposure to olanzapine was similarly associated with a trend toward lower mortality relative to risperidone31. A study with up to 20 years of follow-up (FIN-20) found that monotherapy with olanzapine, risperidone, or aripiprazole was associated with the same risk of somatic hospitalization and same risk of cardiovascular hospitalization32. Overall mortality in patients on monotherapy with olanzapine was not different from aripiprazole and lower for olanzapine compared to risperidone; cardiovascular mortality was similar in patients on olanzapine and aripiprazole and trended towards lower mortality in olanzapine versus risperidone32. A number of large-scale studies not limited to schizophrenia describe similar findings33,34.

To summarize, the higher metabolic burden of olanzapine is a serious problem for some patients. However, decisions about olanzapine’s use as a first-line agent cannot simply be reduced to reflexive exclusion based on metabolic risk. Based on the three key criteria and findings described above, we do not see a legitimate reason to include risperidone among first-line treatment options while excluding olanzapine. Rather, we believe there are sufficient and legitimate reasons to consider olanzapine as a valid first-line treatment option (see the section “Summary of first-line treatment options for patients without concurrent violence”) while appropriately monitoring for metabolic effects. This conclusion is contrary to several existing algorithms and recommendations5,6,9,35 but in agreement with the Texas Medication Algorithm Project7. As discussed in the introduction, individual patient preferences are crucial, side effects influence adherence36, and thus for patients who list weight gain as a particularly undesirable outcome olanzapine should be reserved for later trials.

Perphenazine

Perphenazine is excluded from first-line treatment options due to recognized risk of tardive dyskinesia (TD) and other extrapyramidal side effects (EPS) associated with FGAs4,5,6,7,9. However, there is considerable variability in TD risk within FGA and second-generation antipsychotic (SGA) classes, a nuance frequently ignored by other algorithms. Perphenazine has one of the lowest known risks of TD among FGAs37. It also may have a lower risk of TD relative to some SGAs (e.g., lurasidone)37. Perphenazine users have higher antiparkinsonian medication use than aripiprazole users, but similar to risperidone users; akathisia risk is also comparable between perphenazine and risperidone12. Perphenazine trends toward less weight gain compared to aripiprazole and risperidone, and demonstrates significance versus olanzapine [12]. It is linked to lower prolactin elevation than risperidone12. Finally, in CATIE, perphenazine outperformed olanzapine and risperidone in exploratory analysis of neurocognitive performance at 18 months but not in the primary outcome of neurocognitive scores at 2 months38.

In the context of overall mortality, perphenazine outperformed risperidone in the 11-year follow-up study (as discussed in the section “Olanzapine”)31. In FIN-20, oral perphenazine trended toward lower risk of somatic hospitalization relative to risperidone32.

There is paucity of studies on perphenazine in first-episode psychosis (see the “Summary” section). Aside from this limitation, we see no other compelling rationale to include risperidone in first-line treatment options while excluding perphenazine. This contrasts with other available algorithms4,5,6,7,9.

Lurasidone

Lurasidone performed well in the recent network meta-analysis with regard to efficacy; however, evidence was low level12. Thus, true comparative efficacy remains uncertain, and mortality data is lacking, so lurasidone’s place amongst first-line treatment options remains unknown. Including mortality as a key selection factor risks excluding more novel antipsychotics which by definition will not have 10- to 20-year mortality data available. Hence, lurasidone’s known advantages and disadvantages need to be carefully weighed against aripiprazole, risperidone, olanzapine, and perphenazine.

Lurasidone has a favorable metabolic profile, similar to aripiprazole, but the latter has advantages of lower risk of EPS, easier administration (due to no caloric restrictions), and availability of long-acting formulation12,39,40. A second possible advantage of lurasidone is an antidepressant effect, but evidence from long-term trials is limited12. Short-term trials on depressive symptoms in patients with schizophrenia show that lurasidone did not outperform aripiprazole, olanzapine, or risperidone41. In summary, the currently available literature fails to reveal compelling advantages of lurasidone over aripiprazole or perphenazine (see the section “Perphenazine“). Consequently, based on the three core criteria and a fruitless analysis of other potential advantages, lurasidone does not currently merit inclusion among first-line treatment options, a conclusion that deviates from other algorithms4,5.

Summary of first-line treatment options for patients without concurrent violence

It is important to clarify the notion of “first-line treatment option” as used in this text. Given high number of Food and Drug Administration (FDA) approved antipsychotics in the U.S., it is impossible to comprehensively discuss advantages and side effects of all available medications within the confines of a clinical encounter. In our interpretation, “first-line options” pertain to a limited range of medications that can feasibly be discussed with a patient during a clinical encounter. Based on discussion above, we posit that it is reasonable to discuss risks and benefits of aripiprazole, risperidone, olanzapine, and perphenazine in this context.

Treatment after first unsuccessful trial

This algorithm does not meaningfully deviate from other algorithms and recommendations; hence we only provide a brief summary1,4,5,7. If initial trial does not lead to clinically satisfactory response despite sufficient duration42,43 and dose44, we recommend switching45 to monotherapy with an alternative antipsychotic with a different mechanism of action or metabolism. If initial failed trial was due to side effects, alternative antipsychotic with low propensity for that specific side effect should be selected12. Clozapine should be offered after two failed trials with appropriate monitoring1,8,46,47,48.

Treatment of patients with violence

Initial treatment of patients with concurrent violence

Approximately 9% of first-episode patients present with at least moderate hostility49. Rates of homicide in patients with psychosis were found to be elevated before treatment and decrease post-treatment50. It is crucial not to view violence solely as a psychopharmacological issue, and it is essential to understand the context within which it occurs and establish a therapeutic alliance. However, violence can disrupt development of therapeutic alliance, limit access to health care, and impair functioning in the community. Hostility is also associated with higher all-cause antipsychotic discontinuation51. Therefore, pharmacological means of decreasing violence is an important consideration, and differential effects of antipsychotics on violence merit attention.

A Swedish registry study found that use of clozapine, risperidone, or olanzapine was associated with fewer violent crime arrests in patients with psychotic disorders, compared to aripiprazole, haloperidol, or quetiapine52. Two randomized, double-blinded trials demonstrated that clozapine is superior to olanzapine which, in turn, was superior to haloperidol in reducing the number and severity of physical assaults and aggressive events in patients with schizophrenia11. In an open, randomized, European First Episode Schizophrenia Trial (EUFEST), olanzapine was superior to haloperidol, quetiapine, and amisulpride in decreasing hostility scores in the first 3 months10. In CATIE, olanzapine decreased hostility scores significantly more than perphenazine, quetiapine, risperidone, or ziprasidone53, and superiority was maintained beyond 9 months.

Given these outcomes, we recommend olanzapine as the preferred first-line treatment for patients with schizophrenia and significant violence concerns. In our opinion, the potential benefit of mitigating violence outweighs the concern for metabolic side effects. Olanzapine’s lower overall discontinuation rate also supports its preference in this population12,52. Finally, olanzapine offers the advantage of both oral, short-acting intramuscular, and long-acting (though latter is rarely used due to monitoring barriers and risk of post-injection delirium-sedation syndrome) formulations. If adherence with oral medications is a concern, preference for long-acting injectable (LAI) medications can be considered54. In summary, unless other individual characteristics of a patient dictate otherwise, we believe that patients with schizophrenia and comorbid violence should preferentially be offered olanzapine as a first-line treatment.

Treatment after first unsuccessful trial

If violence does not meaningfully improve with olanzapine, offering a non-clozapine antipsychotic as the next step could be considered suboptimal11,52,55,56. While clozapine is typically recommended after two unsuccessful trials in patients without violence, we propose modifying this strategy in patients with significant violence. After trialing olanzapine, if laboratory monitoring and medication compliance are feasible, clozapine should be offered as second-line treatment without the requirement of two failed trials. This aligns with APA’s recommendation to offer clozapine in patients with aggression (though APA does not recommend use of clozapine after one failed trial), as well as with Texas Medication Algorithm Project recommending early trial of clozapine in patients with violence1,7.

Summary

We propose a stepwise treatment rationale for patients with early-stage schizophrenia that diverges in small but important ways from established algorithms. The first point of departure is early patient stratification based on the presence or absence of comorbid violence. For patients with comorbid violence, we advocate for initiation of olanzapine as a preferred first-line treatment option, then (when feasible) clozapine if olanzapine fails. We also diverge in our selection of “first-line” (see the section “Summary of first-line treatment options for patients without concurrent violence”) treatment options for the general patient population, endorsing aripiprazole and risperidone4,5,6,7,9 but also recommending olanzapine and perphenazine as first-line options. Conversely, we demur regarding lurasidone, opining that its inclusion as a first-line agent may not (yet) be warranted.

Our proposed rationale has several limitations. First, this manuscript focuses on five antipsychotics which were selected based on long-term follow-up of acutely ill patients, but not first-episode patients12. Another reasonable strategy to select relevant literature for first-line treatments in drug naive patients would be to focus solely on trials with first-episode patients. However, there are fewer long-term studies available on this topic, and the available network meta-analyses and systematic reviews found no convincing difference in treatment response between first-episode patients compared to the general patient population—with the exception of quetiapine and olanzapine outperforming haloperidol, and olanzapine outperforming risperidone for negative symptoms57,58,59,60. These findings align with our recommendations. On a similar theme, it is often mentioned that antipsychotic-naive patients and those with shorter illness duration respond better to antipsychotics61, and hence focusing on overall efficacy may be less relevant in early stages of illness. However, medications that have traditionally been considered to be less efficacious, such as aripiprazole62, have nonetheless made it into our treatment recommendations based on the three key criteria, which we find reassuring.

Second, selection of antipsychotics for this Perspective was based on meta-analytic evidence of studies with longer follow-up durations12. This decision has at least two downstream consequences: it can exclude reasonable medication options that have not yet been tested in long-term trials, and the number of studies supporting our recommendations is lower relative to shorter-term trials. For example, perphenazine inclusion in our recommendations is based mostly on the CATIE trial; however, we note that studies with other designs, nonetheless support efficacy of perphenazine41,63.

Third, one of our core criteria for antipsychotic selection is association between specific antipsychotics and mortality. Mortality is inherently difficult to study, especially because randomized trials tend to be brief and epidemiological studies are prone to confounding. Despite these challenges, we believe that sufficient data has accumulated to give relevant guidance. Mortality is also a problematic criterion for novel antipsychotics due to absence of available long-term mortality data. As with lurasidone, serious consideration of novel antipsychotics as first-line agents will require more expansive analyses and consideration of novel advantages. At the time of this publication, trace amine-associated receptor 1 agents failed phase 3 trials, and xanomeline-trospium combination and emraclidine are not FDA-approved64. Current short-term trials have not demonstrated advantages of cariprazine or brexpiprazole, with perhaps the exception of cariprazine for negative symptoms41,65,66. Should any of the newer (e.g., brexpiprazole, cariprazine, lumateperone) or novel antipsychotics demonstrate superiority in efficacy and tolerability, our rationale does not prohibit their adoption as first-line treatments. Similarly, we did not deliberate on paliperidone, a newer but pharmacologically similar agent to risperidone, that has shown similar efficacy and discontinuation rates relative to risperidone in available meta-analyses12,41. This omission is based on non-pharmacological complexities associated with paliperidone use; in particular Medicaid in some states does not cover oral paliperidone, and its long-acting formulation is presently non-generic and may be expensive based on patient pharmacy benefits, leading to potential insurance/financial reasons for discontinuation in clinical practice in the United States. All authors of this Perspective clinically practice within the United States and have focused on agents available in the region. Based on the available literature, amisulpride is promising with regard to all three criteria12,67,68. However, since this agent is not available in the United States for treatment of schizophrenia, we have no firsthand experience with this medication, and we welcome other authors’ opinions on this matter.

Some of the recommended “first-line” antipsychotics are available as LAI formulations. The data on the clinical benefits (e.g., reduced hospitalization, decreased mortality) of LAIs is mixed14,69, and success likely depends on more than just pharmacology, such as the supports in place within LAI clinics (e.g., effort invested in reaching out and arranging follow-ups for patients who miss LAI appointments). We are of the opinion that LAIs are a reasonable option for patients in early stages of illness, that they should be offered to patients early and should not be reserved only for patients with documented adherence difficulties. In addition to potential clinical benefits, other relevant discussion points may include reduction of pill burden, convenience, extended-interval dosing, variable formulations, and decreased peak and trough effects. Stigma, costs, access to administration, and fear of needles may be barriers to LAIs.

Lastly, for all four “first-line” options listed here (olanzapine, perphenazine, aripiprazole, risperidone) and clozapine, women have been found to reach higher plasma concentrations in relation to dose, increasing risk of overmedication70,71,72,73. However, many other factors (e.g., smoking, race, co-prescribed medications, age, body weight, etc.) also influence concentration-to-dose ratio; hence sex is only one factor to consider in appropriate dosing of antipsychotics70,71,72,73. In addition to the limitations listed above, our rationale will require continued refinement as new data on older antipsychotics emerges, recognizing the dynamic nature of psychopharmacology.