Mapping the evidence on pharmacological interventions for non-affective psychosis in humanitarian non-specialised settings: a UNHCR clinical guidance

In people with non-affective psychosis who do not improve after treatment with a FGA, is switching to another FGA effective and safe?

Haloperidol and chlorpromazine were used as the reference standard, as these medications are the first-line recommended treatments for patients with non-affective psychosis in the mhGAP guidelines [29, 36].

We did not find studies of SGA intervention in individuals who had failed to improve after treatment with one FGA. Therefore, we included studies conducted in the general population of people suffering from non-affective psychosis. This evidence was rated as indirect (Additional file 1).

According to available evidence, no difference in treatment response was identified between haloperidol and FGAs as a class, and between haloperidol and the following medications considered individually: chlorpromazine, perphenazine, pimozide, fluphenazine and trifluoperazine. Haloperidol showed a slightly better overall acceptability with respect to chlorpromazine, but caused more movement disorders. Perphenazine and FGAs as a class did not show benefits compared with haloperidol in terms of efficacy, acceptability and tolerability. Pimozide, fluphenazine and trifluoperazine did not show benefits over haloperidol in terms of efficacy, while data on acceptability and tolerability were not available.

With few exceptions, the GRADE tables showed ‘low’ or ‘very low’ quality for the vast majority of outcomes, mostly due to the indirectness of the evidence, as no data were collected in low-resource settings, and individuals were not included on the basis of being not responsive to haloperidol as first-line treatment. Moreover, many of the included studies had small sample sizes and high attrition rates.

In conclusion, the clinical implication was that it is not possible to identify individual FGAs to recommend when a first-line treatment with haloperidol proves to be ineffective. Common clinical practice would suggest switching to the other most commonly available treatment or chlorpromazine. However, there is still uncertainty and further studies in low-resources/humanitarian settings may provide relevant insights on this issue.

The therapeutic dose of haloperidol may vary between 2 and 12 mg/day, and a maintenance dose of 4 to 6 mg/day is usually required [35, 37, 38]. For chlorpromazine, the dose range is from 25 to 1000 mg/day, with a maintenance dose of 75 to 300 mg/day [29, 35].

Are SGAs effective and safe in people with non-affective psychosis who do not improve with FGAs used as first-line treatment?

Each SGA was compared initially to haloperidol, as the reference standard of first-line FGAs. SGAs that proved to be more effective than haloperidol were then compared to each other (head-to-head) in order to identify possible advantages of one medication over another. We found no data specifically referring to individuals who had already failed to improve after treatment with one FGA. Therefore, studies conducted in the general population of people suffering from non-affective psychosis were used (indirect). Compared to haloperidol, all SGAs considered individually (with the exception of paliperidone, for which no data were available) appeared to cause less sedation and motor symptoms, and had a better overall acceptability profile. Among these medications, only risperidone and olanzapine showed a more favourable efficacy profile in comparison with haloperidol. After having compared these two medications head-to-head, risperidone showed better acceptability and was associated with less weight gain than olanzapine, while olanzapine caused less motor symptoms and prolactin increase.

Indirectness, high attrition rates and imprecise results (due to small sample sizes) contributed to set the quality of the evidence to ‘low’ or ‘very low’ for all of the outcomes of interest. In conclusion, the clinical implication was that risperidone and olanzapine may be chosen as second-line treatments as they are both associated with benefits in comparison with haloperidol in terms of efficacy and overall acceptability in the medium- and long-term. The level of confidence in this clinical implication was judged uncertain. The panel pointed out that the choice between risperidone and olanzapine should be based on specific patient characteristics and the antipsychotic profile, considering that, in general, risperidone may have a slightly better overall tolerability and cause less weight gain compared to olanzapine, while olanzapine is associated with less motor symptoms and prolactin increase compared to risperidone. Further, the choice should take into account availability, affordability and sustainability of provision in the long term, in relation to the setting of care. The therapeutic dose of risperidone may vary between 2 and 10 mg/day, and a maintenance dose of 4 to 6 mg/day is usually required [35, 39], while the dose of olanzapine may vary between 5 and 20 mg/day, and the maintenance dose required is usually 10 mg/day [35].

Which individual antipsychotic is effective and safe in people with a diagnosis of treatment-resistant non-affective psychosis?

For the purposes of this review, individuals were defined as treatment-resistant when at least two adequate trials with different antipsychotics, one of which is a SGA, proved ineffective. This is an adaptation of the definition provided by Suzuki et al. [40]. However, considering the lack of a widely shared consensus on definitions of treatment resistance [41], we included studies of participants with treatment-resistant psychosis even where this definition varied. Compared to FGAs, clozapine appeared to be more effective in terms of clinical improvement, similarly effective in terms of relapse rates, and similarly acceptable and more tolerable in terms of motor symptoms, while blood problems and weight gain were more frequent in patients taking clozapine. In patients with treatment-resistant psychosis, risperidone and olanzapine appeared to be similarly effective and acceptable in comparison with clozapine. Risperidone was associated with less weight gain and sedation, but more motor symptoms, when compared to olanzapine.

Studies comparing clozapine and olanzapine, as well as studies comparing clozapine and risperidone, provided efficacy outcomes of ‘moderate’ quality. By contrast, for most of the other outcomes the quality was ‘low’ or ‘very low’ due to indirectness and high attrition rates.

As a clinical implication, it was concluded that, in patients not improving after at least two antipsychotics (one of which is an SGA) administered at adequate dose and duration, a switch to risperidone or olanzapine (in people with no previous ineffective exposure to these medications) or clozapine may be considered. Almost no evidence exists on other SGAs. The choice to use clozapine must take into account context-related issues, particularly in relation to safety, given that routine clinical and laboratory monitoring (for the risk of life-threatening agranulocytosis, but also for other potentially severe adverse events such as seizures and myocarditis [42]) and supervision by a specialist should be regarded as a fundamental prerequisite.

The therapeutic dose of clozapine may vary between 25 and 900 mg/day, and a maintenance dose of 300 to 600 mg/day is usually required [35, 40, 43].

Are antidepressant-antipsychotic combinations effective and safe in people with non-affective psychosis who develop depressive, cognitive and negative symptoms?

Augmentation strategies of antipsychotic treatment with antidepressants are often considered in the treatment of depressive, cognitive and negative symptoms in people with non-affective psychosis. These symptom dimensions are often overlooked and may be associated with unfavourable outcomes, such as chronic functional impairment and higher suicide risk [44–46]. We therefore retrieved and analysed all available data on augmentation strategies of antipsychotic treatment with antidepressants.

Adding antidepressants to antipsychotic treatment appeared to be similarly acceptable in comparison with antipsychotics alone, and associated with a statistically relevant benefit on depressive, cognitive and negative symptoms. A ‘low’ and ‘very low’ quality rating was given for all outcomes of interest, due to indirectness, high attrition rates, very small sample sizes and few events for the majority of included studies, which led to imprecise results (Additional file 1).

In conclusion, the clinical implication was that adding one antidepressant to antipsychotic treatment may be considered in the case of clinically relevant depressive, negative or cognitive symptoms. Preference should be given to selective serotonin reuptake inhibitors, including fluoxetine (available widely in generic formulations and included in the WHO essential list of medicines), considering their favourable balance between efficacy and tolerability.

Quality shortcomings and implications for research

Translating scientific data on antipsychotics into pragmatic suggestions to be implemented in humanitarian settings carries methodological limitations. For this particular setting, indirectness represented the most pressing quality issue. First, none of the studies included in selected reviews were performed in low-resource settings or in humanitarian contexts (Additional file 1). Data were collected in people from stable, high-income, Western countries, which are often considerably different from humanitarian settings with regards to distribution of risks and mediating factors such as medical conditions (e.g. dehydration, malnutrition, infectious illnesses), exposure to potentially traumatic events, stability of family and social support, and access to complementary healthcare resources (e.g. psychosocial support, rehabilitation and a safe medical environment in case of acute symptoms). Second, studies from high-income, Western countries may not capture culturally specific concepts of distress that are relevant to other countries. In general, it is unclear whether the efficacy of treatments may differ in contexts characterised by on-going, chronic adversities [47–49]. Third, even when supported by sound scientific evidence, some interventions may not be feasible in low-resource settings due to the need for expensive and/or time-consuming practices (e.g. laboratory investigations and specialist oversight needed for clozapine). Fourth, the setting may strongly affect the burden and impact of side effects and adverse events. For example, prolactin increase can be managed with relatively sophisticated interventions in high-income countries, including laboratory monitoring of blood prolactin levels, addition of low doses of aripiprazole to the current antipsychotic medication, or the addition of bromocriptine or cabergoline under specialist supervision [50]. None of these options is likely to be feasible and suitable in humanitarian settings. Finally, focusing on the aim of pragmatically translating available data into a clinical guidance, we included reviews referring to populations for which some degree of heterogeneity cannot be excluded, as in the case of treatment-resistant psychosis.

In general, the overall quality of evidence was low for the majority of outcomes considered. This was due not only to the setting of care (as in the case of indirectness), but also resulted from the internal quality of the included studies. Most studies included small samples and had high attrition rates and a short follow-up period, resulting in both inconsistency of estimates across studies and imprecise estimates (even after the aggregation of results from single studies).

For all these reasons, we urgently need to broaden the evidence base around antipsychotic treatments and include direct evidence from populations in low- and middle-income countries and, where possible, from humanitarian settings, instead of merely extrapolating results from studies in high-income settings [51, 52]. Within the scope of this paper, we see a need for pragmatic research to establish the cost-effectiveness of various SGAs and clozapine.

Given the specific challenges related to conducting research in humanitarian settings, such studies should ideally have the form of randomised pragmatic trials focused on issues such as feasibility and cost-effectiveness [53, 54]. This review suggests that the introduction of clozapine as a third step in a treatment protocol could yield significant benefits. However, there is considerable uncertainty about the routine use of clozapine in low- and middle-income countries, mostly because of the risk of serious, and sometimes life-threatening, side effects [55, 56].

Implications for practice and policy

In recent years, strong emphasis has been placed on addressing mental health issues in low-resource and humanitarian contexts in a timely manner, not only to improve the quality of life for people suffering from mental illness, but also as a necessary step to achieve global health and development goals [57, 58]. The growing number of humanitarian and emergency settings, many of which develop into protracted crises that take years if not decades to resolve, should prompt global health researchers to explicitly take these settings into consideration. The proposed algorithm (Fig. 1) is aimed at optimising not only the quality of treatments, but also the timeliness of care for people with non-affective psychosis, considering that early intervention is widely recognised as an essential precondition for achieving higher response rates and better functioning outcomes [59]. This algorithm shows relevant differences if compared with current guidelines developed for general settings of care in high-income countries [60–63], which confirms how, starting from the same evidence base, clinical decisions can radically differ in relation to context-related preferences, values, feasibility and cost-effectiveness considerations.

Among possible obstacles preventing this guidance from having an effective impact, it should be acknowledged that accurate psychiatric assessment and differential diagnosis might be particularly challenging for non-specialised health workers, especially in emergency and humanitarian contexts. This is generally true for mental health, but particularly relevant for the area of psychosis. For example, the onset of non-affective psychosis is often preceded by non-specific symptoms, such as perplexity, obsessive-compulsive manifestations and sub-threshold mood alterations, whose recognition is challenging even for trained psychiatrists [64]. In addition, in contexts with high levels of disorders related to extreme stress, pre-psychotic stages can be misdiagnosed as mood disorders. At the same time, severe mood or post-traumatic disorders with psychotic features can be mistaken for prodromal psychosis. This is particularly relevant if we consider that culturally related manifestations often include mood episodes with psychotic features [65–67]. Furthermore, the most appropriate options for an effective implementation of mental health recommendations into clinical practice remain unclear [68], and this issue results particularly challenging for humanitarian, low-resources settings [69, 70]. Therefore, although guidance on psychopharmacology can notably improve the effectiveness and timeliness of interventions in humanitarian settings, this tool alone cannot be considered as exhaustive. Sustainable improvements in the quality of treatments for people with severe mental illness need to be supported by a broader cultural and structural change in health systems on multiple levels [71–73].