Cross-sectional comparison of first-generation antipsychotic long-acting injections vs risperidone long-acting injection: patient-rated attitudes, satisfaction and tolerability

Abstract

Objectives:

The objective of this study was to compare patients’ attitudes and satisfaction with medication and patient-rated tolerability between those prescribed a first-generation antipsychotic long-acting injection (FGA-LAI) and those prescribed risperidone long-acting injection (RLAI).

Method:

A cross-sectional study of a representative sample of outpatients prescribed an FGA-LAI or RLAI for a minimum of 6 months and attending a depot clinic. Attitudes to medication were assessed by the Drug Attitude Inventory (DAI-30), tolerability was measured by the Liverpool University Neuroleptic Side Effect Rating Scale (LUNSERS) and satisfaction with antipsychotic medication was assessed by the Satisfaction with Antipsychotic Medication (SWAM) scale.

Results:

The RLAI (n = 28) and FGA-LAI (n = 39) groups did not differ in terms of mean age, sex, diagnosis and ethnicity. All individual LAIs were prescribed within British National Formulary limits. The most commonly prescribed FGA-LAI was flupentixol decanoate (n = 22). There was no significant difference between the RLAI and FGA-LAI groups in terms of mean total scores on the DAI-30, LUNSERS and SWAM or the tolerability subscales of the LUNSERS or the two subscales (treatment acceptability and medication insight) of the SWAM. In both LAI groups there was a low level of side effects (LUNSERS) and a generally positive attitude (DAI-30) and reasonable satisfaction (SWAM) with medication.

Conclusions:

Patients treated with FGA-LAI and RLAI for at least 6 months did not differ in terms of patient-rated tolerability, attitudes and satisfaction with medication. The current design cannot determine whether differences would have been evident earlier on during treatment. These results should be regarded as preliminary and are subject to prescribing bias. Randomized studies avoid prescribing bias and are a superior way to compare specific LAIs. Ideally randomized studies should include patient-rated outcome measures including medication tolerability; assessment of side effects, efficacy and quality of life made by blinded raters; and additional objective side-effect data including changes in weight and key blood parameters.

Introduction

The course of schizophrenia is variable but for most patients it is a chronic relapsing disorder and as such maintenance antipsychotic medication is an important part of treatment [NICE, 2014]. In meta-analyses, maintenance antipsychotic treatment reduces the risk of relapse compared with placebo [Leucht et al. 2003]. However, in clinical practice the effectiveness of maintenance antipsychotic treatment is often reduced by poor adherence [Haddad et al. 2014]. This is a common problem affecting at least one-third of patients with schizophrenia in any given year [Valenstein et al. 2006]. Observational studies have repeatedly shown that poor adherence to antipsychotic medication is associated with worse outcomes in schizophrenia, including a higher risk of hospitalization and deliberate self-harm and a lower risk of remission [Novick et al. 2010].

Antipsychotic depots or long-acting injections (LAIs) were introduced in the 1960s as a means of improving adherence and thereby clinical outcomes [Johnson, 2009; Simpson, 1984]. Studies that have compared the effectiveness of oral antipsychotics and LAIs have produced contradictory results [Haddad et al. 2009, 2015]. Trial methodology, including variation in the characteristics of the patients studied, appears a key factor in explaining these differences. The most recent meta-analysis of randomized controlled trials (RCTs) found that LAIs and oral antipsychotics did not differ in terms of rates of relapse or hospitalization [Kishimoto et al. 2013b]. In contrast, a nationwide cohort study found a decreased risk of rehospitalization with LAIs compared with the same antipsychotics in oral form [Tiihonen et al. 2011] and a meta-analysis of mirror image studies showed that LAI treatment decreased the number of hospitalizations compared with prior treatment with oral antipsychotics [Kishimoto et al. 2013a]. These differences have led to suggestions that the traditional RCT, often with numerous exclusion criteria, may not be the gold standard method to determine the comparative effectiveness of oral antipsychotics and LAIs [Haddad et al. 2015]. Most treatment guidelines for schizophrenia recommend LAIs as a potential strategy to improve medication adherence [NICE, 2014; Lehman et al. 2004; Canadian Psychiatric Association, 2005].

Clinicians currently have access to several first-generation antipsychotic LAIs (FGA-LAIs) and second-generation antipsychotic LAIs (SGA-LAIs). Risperidone long-acting injection (RLAI) was the first SGA-LAI to be approved and entered clinical practice in the United States in 2003. Since then, three other SGA-LAIs, paliperidone LAI, olanzapine LAI and aripiprazole, have become available in various countries [Citrome, 2013].

A key question is how FGA-LAIs and SGA-LAIs compare in terms of efficacy and tolerability. SGA-LAIs have higher acquisition costs than FGA-LAIs. The current economic climate, with tighter controls on drug budgets in many countries, means that comparative data on the effectiveness and tolerability of FGA-LAIs and SGA-LAIs are particularly relevant. RCTs are the gold standard method to determine this but only two RCTs comparing an FGA-LAI and a SGA-LAI have been published [Rubio et al. 2006; McEvoy et al. 2014]. Rubio and colleagues reported greater improvement in schizophrenia symptoms and extrapyramidal symptoms (EPS) with RLAI versus zuclopenthixol decanoate in patients with schizophrenia and comorbid substance misuse [Rubio et al. 2006]. McEvoy and colleagues compared paliperidone LAI (PLAI) and haloperidol LAI in patients with schizophrenia and showed equivalent efficacy but greater weight gain with PLAI and greater akathisia with haloperidol LAI [McEvoy et al. 2014].

Various studies, including subanalyses of larger studies, have assessed the outcome of switching to RLAI from a range of prior antipsychotic medications. A weakness with switching designs is the absence of a control group, which means the design cannot prove that longitudinal changes are the result of switching medication. Another weakness is lack of blinding which can introduce bias. Two studies have assessed patients switching from a FGA-LAI to RLAI and both reported a reduction in EPS and Positive and Negative Syndrome Scale (PANSS) total scores following the switch [Lasser et al. 2004; Lai et al. 2009]. Van Os and colleagues reported on patients switching from oral FGAs to RLAI and found a decrease in EPS and PANSS total [Van Os et al. 2004]. Marinis and colleagues reported on patients who switched from either an oral FGA or depot FGA to RLAI [Marinis et al. 2007]. Following the switch, improvements were seen in PANSS total scores, tolerability, quality of life and patient satisfaction. RLAI switch studies that have recruited patients prescribed a wider range of prior antipsychotics (i.e. SGAs and FGAs; orals and LAIs) are less informative when attempting to answer the question of how RLAI compares with FGAs. Studies in this category include the primary analysis from the StoRMi Study [Möller et al. 2005] and a study by Lindemayer and colleagues [Lindemayer et al. 2004]. Nevertheless both studies showed a reduction in symptom severity and EPS. Lindenmayer and colleagues found that the reduction in EPS was more marked when prior treatment was oral haloperidol, rather than oral olanzapine or quetiapine, but that switching to RLAI was accompanied by an increase in mean serum prolactin levels [Lindenmayer et al. 2004]. Mirror image studies, a specific type of switch study, have assessed the effects of switching to various LAIs, including RLAI and aripiprazole LAI, from prior oral antipsychotics. Such studies nearly always show a reduction in bed occupancy for the LAI versus ‘prior’ treatment [Kishimoto et al. 2013a; Kane et al. 2015]. However, mirror image designs have methodological weaknesses [Haddad et al. 2015] and furthermore these studies do not usually provide subanalyses by category (FGA or SGA) of prior antipsychotic.

The relative merits of oral FGAs and oral SGAs has been debated for many years. When oral SGAs were introduced it was believed that they would offer better tolerability and efficacy compared with oral FGAs. However, large-scale independent trials, including the CATIE (Clinical Antipsychotic Trials of Intervention Effectiveness) study [Lieberman et al. 2005] and the CUtLASS (Cost Utility of The Latest Antipsychotics in Severe Schizophrenia) study [Jones et al. 2006] showed little difference between FGAs and SGAs, other than clozapine, in overall effectiveness. Quality of life, the primary outcome measure in CUtLASS, did not differ between the FGA and SGA groups [Jones et al. 2006]. Total discontinuation rates, the primary outcome measure in CATIE, did not differ between perphenazine, the FGA comparator, and individual SGAs other than olanzapine which was associated with a lower discontinuation rate [Lieberman et al. 2005]. Some differences in side effects were apparent in CATIE, including that olanzapine was associated with more discontinuations for weight gain or metabolic effects and perphenazine was associated with more discontinuations for EPS. Neither CUtLASS [Jones et al. 2006] nor CATIE showed a difference in the incidence of EPS between the FGA comparators and SGAs. The choice of FGA in CUtLASS was at the discretion of the clinician, with the most widely prescribed FGA being sulpiride. The findings of the CATIE and CuTLASS studies cannot be extrapolated to all FGAs; FGAs and SGAs are not homogeneous groups in terms of side-effect profiles. It is more clinically meaningful to make comparisons between individual FGAs and SGAs [Haddad and Sharma, 2007]. A second randomized phase in both CATIE and CUtLASS showed that clozapine was superior to other antipsychotics in treatment-resistant schizophrenia [McEvoy et al. 2006; Lewis et al. 2006].

Systematic reviews of patient attitudes reveal that most patients who are currently prescribed LAIs prefer this formulation to tablets, although attitudes towards LAIs among patients currently prescribed oral antipsychotics are less positive [Walburn et al. 2001; Waddell and Taylor, 2009]. Most patients in these studies have chronic schizophrenia and the attitude of patients early in the course of schizophrenia to LAIs remains neglected [Kirschner et al. 2013]. We are not aware of any research that has compared patient attitudes, tolerability and satisfaction between SGA-LAIs and FGA-LAIs in routine clinical practice. We undertook a cross-sectional survey of patients attending depot clinics at the Lancashire Care NHS Trust to investigate these variables.

Methodology

This study was approved by the Lancashire and Cumbria Research and Ethics committee under reference number 09/H1016/113. Patients currently prescribed an LAI within the general adult psychiatry service run by the Trust were identified from records maintained by all community mental health teams (CMHTs) and the pharmacy at the Parkwood Psychiatric Unit in Blackpool.

Through liaison with clinical staff at CMHTs, patients who met the inclusion and exclusion criteria were identified and invited to participate. Inclusion criteria required the patient to currently be prescribed an LAI and to have been prescribed that same LAI for at least the preceding 6 months. The following patient groups were excluded from participation in the study: inpatients, patients detained under the Mental Health Act, patients lacking capacity to consent, and those who could not understand or speak English. The study was restricted to patients receiving treatment in a depot clinic as this facilitated identification and recruitment of patients. A total of 160 patients were eligible for inclusion in the study. They were informed about the study through patient information leaflets and face-to-face explanations. They were advised to take up to 2 weeks to decide if they wanted to participate in the project. Out of the 160 eligible patients, 80 consented to participate, of which 67 completed the survey. Data collection began in November 2009 and was completed in January 2010.

Patients who consented to participate were given the study questionnaires to complete in private and return by post. Questionnaires were anonymous but were coded to enable the research team to send out a reminder to those who did not return their questionnaires. We stopped sending reminders soon after we obtained the minimum number required for analysis based on our power calculation.

Instruments

Participants completed a demographic and clinical history form and three self-completion questionnaires: the Drug Attitude Inventory (DAI-30), the Liverpool University Neuroleptic Side Effect Rating Scale (LUNSERS) and the Satisfaction with Antipsychotic Medication (SWAM) scale. All these instruments are simple to complete. The DAI-30 [Hogan et al. 1983] is a 30-item scale with an equal distribution of positive and negative items to reduce acquiescence bias. It assesses patients’ views about medication with a positive total score indicating a positive subjective response and a negative total score indicating a negative subjective response. The scale has been used to predict medication adherence [Hogan et al. 1983].

The SWAM scale [Rofail et al. 2005] is a 23-item scale that comprises two subscales: treatment acceptability (14 items) and medication insight (9 items). Higher scores on the treatment acceptability subscale indicate higher satisfaction. Higher scores on the medication insight indicate poorer insight into antipsychotic medication, which encompasses an understanding and awareness of the need for medication and its potential side effects.

The LUNSERS scale is a self-rating scale for rating neuroleptic side effects [Day et al. 1995]. This scale has been widely used both in research and in clinical practice, and has been found to be easy to use by patients on antipsychotic medication. The scale includes 41 questions plus 10 ‘red herring’ items to test the robustness of the results. A wide range of side effects are covered, including extrapyramidal, autonomic, anticholinergic, sexual and other side effects. Higher scores on the LUNSERS scale indicate a poor tolerability and various subscale scores, including EPS, can be calculated.

Power calculation

A power calculation was carried out for the DAI−30 score which was the primary outcome measure. It was estimated that if the standard deviation was 12, a clinically significant difference on the DAI-30 would be 10 points [Hogan and Awad, 1992]. To achieve a power of 86% at a significance of 0.05, the minimum number in each arm in the study would be 28. This meant that there was an 86% chance of detecting a 10-point difference on the DAI-30 at a significance level of 0.05.

Statistical analysis

Data were entered into SPSS/PASW version 18 for analysis. Scores on the DAI-30, SWAM and LUNSERS scales were compared using the t test. Ages of the patients prescribed FGA-LAIs versus those prescribed RLAI were compared. The Kruskal–Wallis test was used to analyse differences in duration of illness and other nonparametric data. Median doses were calculated for all LAI antipsychotics.

Results

Within the service, 160 patients met the inclusion and exclusion criteria. Of these, 80 consented to the study and 67 patients returned completed forms (84% response rate). The results that follow are based on data obtained from these 67 individuals. Table 1 summarizes the sociodemographic and clinical characteristics of the sample.

Table 1.

Sociodemographic and clinical characteristics.

	FGA-LAI	RLAI	Statistical significance
	n = 39 (58.2%)	n = 28 (41.8%)
Age in years, mean (SD)	45.97 (10.54)	42.32 (8.92)	0.173
Men, number (%)	22 (56%)	16 (57%)	0.576
Diagnosis, number (%)			NA
Schizophrenia	36 (92%)	24 (86%)
Bipolar	1 (2%)	1 (4%)
Borderline PD	2 (5%)	1 (4%)
Schizoaffective	0	2 (7%)
Ethnicity: white, number (%)	39 (100%)	28 (100%)	NA

NA, not applicable; PD, personality disorder.

Of the 67 participants, 28 were prescribed RLAI while 39 were prescribed an FGA-LAI. All the patients were white and most were diagnosed with schizophrenia. The breakdown of current LAI within the FGA-LAI group was flupentixol decanoate (n = 22), zuclopenthixol decanoate (n = 13), fluphenazine decanoate (n = 3) and haloperidol decanoate (n = 1). Median doses for all the FGA-LAIs and RLAI were calculated and compared with the British National Formulary (BNF) maximum licensed dose (Table 2). All individual patient doses were within current BNF limits. The median duration of treatment for the FGA-LAI group was significantly longer than that for the RLAI group (33 months versus 84 months) (p < 0.001) (Kruskal–Wallis test).

Table 2.

Details of doses for LAIs (note: doses for all individual patients were within BNF limits).

LAI	Number of patients (% of total sample)	Median dose	Median dose as % of BNF maximum licensed dose
RLAI	28 (41.8%)	37.5 mg every 2 weeks	75%
Flupentixol decanoate	22 (32.8%)	100 mg every 2 weeks	12.5%
Zuclopenthixol decanoate	13 (19.4%)	200 mg ever 2 weeks	16.7%
Fluphenazine decanoate	3 (4.5%)	50 mg every 2 weeks	50%
Haloperidol decanoate	1 (1.5%)	100 mg every 4 weeks	33.3%

BNF, British National Formulary; LAI, long-acting injection; RLAI, risperidone long-acting injection.

The mean scores for the DAI-30, SWAM and LUNSERS scales are given in Table 3. No significant difference in total or subscale scores between the FGA-LAI and RLAI groups was observed.

Table 3.

Rating scale scores (independent sample t tests).

	FGA-LAI (n = 39)	RLAI (n = 28)	p value
	Mean score	Mean score
DAI-30 total*	16.18	14.43	0.491
LUNSERS total$	30.03	30.89	0.893
LUNSERS EPS	4.86	5.23	0.753
LUNSERS autonomic	3.29	2.28	0.158
LUNSERS anticholinergic	3.14	3.23	0.924
LUNSERS others	19.61	19.28	0.937
SWAM total‡	36.46	35.71	0.825
SWAM treatment acceptability	55.93	57.03	0.752
SWAM medication insight	20.21	20.56	0.822

^*DAI-30 scores range from +30 to −30 with negative scores being associated with nonadherence and positive scores being associated with adherence.

^$LUNSERS total ranges from 0 to 156 (men) and 0 to 164 (women) with higher scores indicating greater side-effect burden. The LUNSERS items can be split to score groups of side effects separately, including extrapyramidal, autonomic, anticholinergic, sexual and other side effects.

^‡SWAM total scores range from 23 to 115. There are two subscales: treatment acceptability (scores 14–70) and medication insight (scores 9–45).

DAI-30, Drug Attitude Inventory; FGA-LAI, first-generation antipsychotic long-acting injection; LUNSERS, Liverpool University Neuroleptic Side Effect Rating Scale; RLAI, risperidone long-acting injection; SWAM, Satisfaction with Antipsychotic Medication.

Discussion

Main finding

No difference in subjective tolerability (LUNSERS), attitude towards medication (DAI-30) or satisfaction with medication (SWAM) was observed between patients prescribed an FGA-LAI versus RLAI. For both treatment groups the total LUNSERS score indicated a low level of side effects, the DAI-30 score a generally positive attitude to medication and the SWAM subscale scores showed a reasonable level of treatment acceptability and medication insight. As far as the authors are aware this is the first study to compare the attitudes and subjective tolerability of patients prescribed RLAI and FGA-LAIs in an observational design. The failure to find differences between RLAI and FGA-LAIs is broadly consistent with the CATIE [Lieberman et al. 2005] and CUtLASS [Jones et al. 2006] studies, both of which showed no difference between oral FGAs and oral SGAs in the incidence of EPS and little or no difference in overall effectiveness between the drugs studied (other than clozapine).

Conversely our finding of equivalent tolerability between RLAI and FGA-LAIs is at odds with several studies, including a RCT that reported fewer EPS with zuclopenthixol decanoate versus RLAI [Rubio et al. 2006]. Paliperidone is the 7-hydroxy metabolite of risperidone and a recent RCT comparing paliperidone LAI and haloperidol LAI reported greater akathisia but less weight gain with haloperidol versus paliperidone [McEvoy et al. 2014]. A reduction in EPS was seen in two studies in which patients were switched from a FGA-LAI to RLAI [Lasser et al. 2004; Lai et al. 2009]. Finally, one would expect to see differences in the side-effect profiles of different LAIs based on extrapolation of side-effect data from RCTs of oral antipsychotics [Leucht et al. 2013].

The differences between these studies and our own in terms of tolerability may reflect the different time periods; we assessed patients who had been on LAI treatment for at least 6 months whereas the other studies assessed patients from the time of initiation of treatment. In addition we assessed patients’ subjective views whereas the other studies used objective or researcher rated measures of side effects. Subjective assessment may be less likely to detect differences in side effects than objective measures. For example, patients are usually unaware of tardive dyskinesia and metabolic disturbances such as hyperglycaemia will not be evident without blood screening. Furthermore, modest weight change may not be apparent without weighing. A full assessment of side effects requires objective assessments in addition to patients’ perceptions of side effects. Finally, differences in outcomes between studies may reflect the individual drugs comprising the FGA-LAI group in each study and differences in the doses of the LAIs, particularly within the FGA-LAI group. Haloperidol is the FGA with the greatest propensity to cause EPS but was only prescribed to one patient in our study.

The risk of EPS is dose related for both RLAI [Kane et al. 2003] and FGA-LAIs [Donnelly et al. 2013; McCreadie et al. 1979]. However, the licensed dose range of RLAI is narrower than for most FGA-LAIs. Dosing studies of FGA-LAIs do not show an efficacy advantage for high over medium doses [Adams et al. 2001]. Consequently judicious use of low and medium doses of FGA-LAIs may allow these drugs to be effective yet with a lower EPS risk than was the case in the pre-SGA era when the doses of FGA-LAIs used in clinical practice tended to be higher. The median dose of the two most commonly prescribed FGA-LAIs in this study (flupentixol decanoate and zuclopenthixol decanoate), which accounted for nearly 90% of patients in the FGA-LAI group, were each less than 20% of the maximum BNF licenced dose. Median doses of the other two FGA-LAIs were 33.3% and 50% of the maximum licensed dose (see Table 2). This modest dosing is likely to have contributed to the relatively good tolerability that we found.

Sample

All patients in the study were white. This reflects the catchment area of the Trust where the study was conducted in which ethnic minorities form a small proportion of the population [Office for National Statistics, 2001]. There was no significant difference in terms of mean age or sex distribution between the FGA-LAI and RLAI groups. The duration of treatment was significantly higher for patients who were prescribed an FGA-LAI compared with an RLAI (84 versus 33 months, p < 0.001). This is consistent with RLAI being introduced into clinical practice in the UK in 2003 whereas FGA-LAIs have been in use for several decades. We are unable to comment on how many patients initiated treatment with FGA-LAIs and RLAI, and stopped treatment before 6 months to leave the groups that we assessed in this study. Nevertheless the mean duration of treatment is relatively long in both groups, indicating that some patients find a LAI an acceptable maintenance treatment. This implies that LAIs, when used appropriately, can aid adherence [Hong et al. 2013].

Strengths and limitations

A strength of this study is that it was conducted independently without industry funding or support. A power calculation was conducted to determine the minimum number of participants to be recruited into each group for statistical significance on the primary outcome. This recruitment target was exceeded. Validated and reliable scales to measure patient attitudes (DAI-30), tolerability (LUNSERS) and satisfaction (SWAM) were used. Sixty-seven (84%) of the 80 consenting patients completed the study, which is a relatively high completion rate.

We used a cross-sectional and observational design which leads to some limitations. The most important is that since this was an observational study, prescribing bias may affect the results. For example, it might be that patients who have already experienced EPS or who were regarded as high risk of developing EPS were prescribed RLAI rather than an FGA-LAI. If one assumes that FGA-LAIs are associated with a greater risk of EPS than RLAI, then the prescribing bias would tend to reduce differences between the two groups in the prevalence of EPS. A randomized design would overcome this and is the most reliable way to determine differences between two or more treatments in terms of side effects, quality of life and other outcome measures. However, randomized studies are more complex to conduct than observational studies and are often associated with a higher refusal rate, which may lead to problems with generalizability.

With regard to generalizability, 80 out of 160 eligible patients consented to enter this study. This could have led to bias but we cannot comment on this further as we have no data on the characteristics of those who declined to enter the study. To assist identification and recruitment we only considered community patients who received their LAI from a ‘depot clinic’ and those who received their LAI at home were excluded. This is unlikely to be a major bias as in the CMHTs that we recruited from, most LAIs were given in ‘depot clinics’ rather than at patients’ homes (verbal reports from key clinicians in the CMHTs). The exclusion of inpatients is unlikely to be a significant bias as the proportion of patients treated with a LAI for 6 months or more (an inclusion criterion) who are inpatients is likely to be very low.

The power calculation was related to the primary outcome measure, the DAI-30. No power calculation was conducted for differences in the LUNSERS and SWAM scales. The sample size was small and represented only one coastal region in the county of Lancashire. Hence, it might not be possible to generalize the results to other parts of the UK. Demographic details and diagnosis were based on patient report rather than the medical notes. We did not study the effect of additionally prescribed medications, such as additional oral antipsychotics or anticholinergic medications.

We included patients prescribed an LAI for at least 6 months. Therefore, it could be argued that we preselected patients who were relatively satisfied with their LAI and who were unlikely to be experiencing significant side effects, otherwise they would have been switched to another treatment before reaching 6 months of treatment and being eligible to enter the study. As such, the study deign could minimize differences between the two treatment groups. However, there were good reasons behind our decision to restrict the study to those completing 6 months or more of treatment, namely:

1. To ensure that carryover effects from the preceding antipsychotic did not affect ratings. In particular, a proportion of those prescribed RLAI are likely to have been switched from a FGA-LAI due to tolerability problems and it can take several months for an FGA-LAI to be eliminated from the body [Nyberg et al. 1997].

2. To ensure that assessments were conducted when steady-state antipsychotic plasma levels had been achieved. It can take several months to reach steady-state levels with a LAI [Taylor, 2009]. As a result, if patients who had only recently started LAI medication had been included, then potential adverse effects associated with the drug might not have been fully manifest at the time of assessment.

3. LAIs are primarily used as maintenance treatments, hence we wanted to assess the patients’ views during the maintenance phase of treatment rather than soon after initiation.

In summary, there were good methodological reasons for recruiting patients who had been prescribed an LAI for 6 months or more. However, this decision means that the study cannot determine whether differences in attitudes, tolerability or satisfaction were present early during treatment.

The LUNSERS scale does not assess injection-related side effects such as pain or injection site nodules. Whether FGA-LAIs and RLAI differ in terms of the risk and severity of injection site side effects is unknown. It is conceivable that differences might exist given that FGA-LAIs are oil based and can be administered at intervals of up to 5 weeks [BNF, 2015] depending on the specific LAI, whereas RLAI is water based and has a fixed 2-week injection interval. Water-based LAIs may be associated with a lower risk of injection-related side effects versus oil-based LAIs due to irritation associated with the ‘carrier oil’, though as far as we know this possibility has not been investigated in a RCT.

No FGA-LAIs are licensed for deltoid administration. RLAI was initially licensed for gluteal administration but was subsequently approved for deltoid administration. In this study we did not record the current site of administration of RLAI but it is likely that most injections were gluteal as the deltoid site had only recently been approved when the study was conducted. It is likely that awareness of the choice of injection site among patients treated with RLAI will increase over time. It is unclear whether or not this will affect tolerability, attitudes or satisfaction.

Conclusion

We did not find a difference in subjective side effects, patients’ attitudes or satisfaction between patients who had been prescribed RLAI versus FGA-LAIs for at least 6 months. However, methodological weaknesses and a small sample size mean that our findings should be regarded as preliminary. Further research, ideally from RCTs, is required to compare different LAIs. This is particularly relevant given that LAIs are prescribed to between a quarter and a third of patients with schizophrenia in the UK depending on the clinical setting [Barnes et al. 2009].

SGA-LAIs have a higher acquisition cost than FGA-LAIs. However, antipsychotic selection needs to be made on an individual patient basis [NICE, 2014], considering many factors, not just acquisition cost, and patients should be fully involved. There are few head-to-head studies that compare different LAIs but a meta-analysis of RCTs that compare different oral antipsychotics indicates that differences in efficacy, clozapine apart, are minimal but that the risk of side effects can vary significantly [Leucht et al. 2013]. Thus side-effect profiles are likely to be a major factor in influencing drug choice. In addition, individuals can differ in their response to specific antipsychotics, even when drugs show little or no difference in effectiveness or tolerability at a group level in RCTs. Consequently, it is important that clinicians and patients have access to a range of LAIs and oral antipsychotics. Having said this, if all other factors are equal when considering prescribing for a specific patient, then it makes sense to prescribe the cheapest drug. It will assist clinical practice and audit if clinicians document their reasons for prescribing a specific antipsychotic, oral or LAI, in the patient’s notes.