Antipsychotic medications have been available via “depot” and other long-acting injectable (LAI) formulations since the 1960s. Such formulations offer alternatives to daily oral dosing and touted advantages in terms of lower side effect burden, convenience, enhanced adherence, and better effectiveness. However, the rate of LAI usage varies widely from 6 to 80% of patients with psychotic disorders around the world, due to differences in the attitudes of both prescribing clinicians and patients across countries and healthcare systems (Manchada et al. 2013). Negative perceptions and stigma about LAIs that impact patient acceptance and refusal commonly include concerns about injection pain, fear of needles, and coercion (Iyer et al. 2013; Das et al., 2014).
In addition to the practical barriers that limit wider use of LAIs, the objective evidence to support their touted advantages over oral antipsychotic medication (OAM) has been mixed. For example, following the release of the first second-generation antipsychotic available in an LAI formulation, several pivotal randomised controlled trials (RCTs) comparing risperidone LAI to oral risperidone found no consistent advantages for LAI in terms of rehospitalisation rates, symptom reduction, incidence of side effects, or rates of medication adherence (Rosenheck et al., 2011; Weiden et al., 2012; Buckley et al., 2015). These somewhat unexpected results led to speculation that RCTs might not offer the best study design to assess LAI superiority, since enrolled subjects in a randomised prospective study with close weekly follow-up and careful monitoring of medication adherence might be more adherent than “real world” patients. In keeping with this methodological account, several meta-analyses have found evidence of LAI superiority relative to OAM in prospective and retrospective cohort studies (Kirson et al. 2013; Kishimoto et al. 2018) and mirror-image studies (Kishimoto et al. 2013), but no difference in effectiveness based on RCTs (Kirson et al. 2013; Kishimoto et al. 2014).
Kishimoto et al. (2021) have now published the largest meta-analysis to date comparing outcomes between LAIs and OAM across all 3 study designs. Will this latest update clarify whether LAIs really offer an advantage over OAM and provide fuel for the argument that LAIs are often under-utilised? Let’s take a look.
In their new meta-analysis, Kishimoto et al. (2021) selected 137 studies including 32 RCTs, 65 cohort studies, and 40 mirror-image or “pre-post” studies comparing LAIs to OAM in adults with schizophrenia and related disorders. These studies were published between 1971 and 2020 and involved both first- and second-generation antipsychotics. For inclusion, the RCT and cohort studies required a minimal duration of 6 months, while the pre-post studies had to last at least 12 months with >6 months on LAI and OAM respectively. Studies that included more than 20% of subjects with non-schizophrenia spectrum disorder were excluded, as were case reports comprising less than 20 patients.
The primary outcome measures for the meta-analysis were hospitalisation or relapse, with preferential use of the former when both outcomes were reported. Secondary outcomes included study discontinuation, hospitalisation days, hospitalisation rates, psychopathology measures, adverse events, as well as other measures of effectiveness and efficacy. Dichotomous outcomes (e.g. hospitalisation or not) were analysed according to pooled risk ratio (RR), while continuous outcomes were expressed as standardised mean difference (SMD) with 95% confidence intervals. For efficacy outcomes, RR < 1 and SMD < 0 were considered to indicate a benefit for LAIs, whereas RR > 1 and SMD > 0 were considered to indicate a benefit for OAM if the effect was significant (p<0.05). For the primary outcome of hospitalisation or relapse, risk was computed as the number of patients who had at least one hospitalisation or relapse divided by the number of patients at risk. The RR was then calculated as the ratio of risk of LAIs versus OAM.
The findings suggest significantly lower rates of hospitalisation or relapse among those treated with LAIs compared to OAM. The number needed to benefit (NNTB) was 25 for RCTs, 31 for cohort studies, and 4 in pre-post studies. Looking exclusively at hospitalisation, LAIs were associated with lower risk compared to OAM, but when looking exclusively at relapse, risk was lower for LAIs only in pre-post studies. LAIs were also associated with significantly lower rates of all-cause discontinuation in cohort studies, but not RCTs.
For most secondary efficacy outcomes related to symptom improvement, LAIs were non-superior to OAM. Based on a small number of studies that reported them, LAIs were only superior for the following:
- Medication adherence (defined by taking medication >75% of the days in the treatment period)
- Hospitalisation rate
- Emergency room visit rate
The authors also tallied secondary outcomes based on the total number of comparisons. Using this approach, LAIs were no different than OAM across all outcomes in 77% of comparisons, advantageous in 18%, and less beneficial in 5%. For effectiveness-related secondary outcomes only, LAIs were superior to OAM in 70% of comparisons and not different in the remaining 30%. For efficacy-related secondary outcomes, LAIs were not different than OAM in 89% of comparisons and superior in the remaining 11%. Finally, for adverse events, LAIs and OAM were equivalent in 82% of comparisons, advantageous in 10%, and worse in 8%.
For adverse events, LAIs were non-superior to OAM for extrapyramidal side effects. As expected, OAM were superior to LAIs in terms of injection site pain based on a single RCT.
This largest and most inclusive meta-analysis of studies to date comparing LAIs to OAM found that LAIs were significantly superior in terms of preventing hospitalisation and relapse among patients with schizophrenia and related disorders. Unlike previous studies, this superiority held true regardless of study design, including RCTs. However, consistent with previous research, the effect size for this superiority was greatest in mirror-image studies and smaller in RCTs.
Strengths and limitations
Comparing the results of this study to existing literature, there is consistent evidence that LAIs are superior to OAM in terms of relapse prevention. However, conclusions of superiority based on this study design must be interpreted cautiously since the results could be biased by lack of blinding or might represent an artefact of continuous, longitudinal treatment. The authors acknowledge that expectation bias and regression to the mean might also preferentially advantage LAIs.
The directional biases of cohort studies are somewhat less clear. Due to lack of randomisation and the potential for more severely ill patients to be given the medication, such studies could disadvantage LAIs based on study design. However, it is also possible that such studies, reflecting real-world clinical practice, could magnify an advantage of them among those who might benefit most from enhanced medication adherence.
What is least clear from this new meta-analysis is why an advantage was found for LAIs compared to OAM in RCTs given previous studies finding no such advantage. As noted, 11 additional RCTs were added since the previous “negative” meta-analysis by the same research group in 2014 (Kishimoto et al. 2014). Two of the new RCTs involved subjects with first-episode schizophrenia, suggesting a possible LAI advantage in this treatment-naïve population. It is also possible that the results of the 11 additional RCTs were biased by pharmaceutical company sponsorship (9 of 11 (81%) were pharma-sponsored compared to 62% in the previous meta-analysis by the same research group) and lack of double-blinding (only 2 of 11 (19%) were double-blinded compared to 43% in the previous meta-analysis). This preponderance of pharmaceutical-sponsored studies relying on open-label or rater-masked study designs might have skewed the reported outcomes to more “positive” results.
Implications for practice
While RCTs usually offer the optimal study design to distinguish medication effects from control interventions in a clinical trial, that may not be the case here. If the clinical advantages of LAIs are mainly related to enhanced treatment adherence, then RCTs might fail to capture the “real-world” benefits in less adherent patients the way that naturalistic studies would. The results of previous research and the differential effect sizes in the current meta-analysis support this interpretation, as does its finding that all-cause discontinuation was greater for OAM compared to LAI in cohort studies, but not RCTs. These conclusions suggest—perhaps unsurprisingly—that LAIs are probably most useful in treatment settings and populations where treatment adherence is compromised during oral dosing. However, OAM remains an important option for patients concerned about autonomy, coercion, and injection pain.
Another key finding of this meta-analysis is that the real-world advantage of LAIs has less to do with efficacy or symptom reduction and much more to do with effectiveness as defined by relapse prevention. This is likely because what LAIs really offer beyond convenience among patients who have difficulty remembering to take OAM on a daily basis is near 100% certainty on the part of clinicians regarding whether or not a patient has taken, or has missed, a dose of medication. This is a significant advantage compared to the notoriously unreliable methods of monitoring OAM adherence that can go a long way towards interventions aimed at preventing relapse and hospitalisation in clinical practice.
It is possible that LAIs administered once or twice a month might offer an advantage due to more frequent contact and monitoring by clinicians compared to OAM. While outpatient visit frequency is likely to be standardised in an RCT, there may be substantial disparity in visit frequency when using LAI vs. OAM in real-world settings and non-RCT clinical trial designs. Accordingly, it will be important to determine if LAIs that are now available to be given every 3-months, can sustain the benefit of LAIs compared to OAM (or more frequently dosed LAIs), either in future clinical research or “real-world” practice.
Statement of interests
Kishimoto T, Hagi K, Kurokawa S, Kane JM, Correll CU. (2021) Long-acting injectable versus oral medication for the maintenance treatment of schizophrenia: a systematic review and comparative meta-analysis of randomised, cohort, and pre-post studies. Lancet Psychiatry 2021. https://doi.org/10.1016/S2215-0366(21)00039-0
Buckley PF, Schooler NR, Goff DC, Hsiao J, Kopelowicz A, Lauriello J, Manschreck T, Mendelowitz AJ, Miller DD, Severe JB, Wilson DR, Ames D, Bustillo J, Mintz J, Kane JM, the PROACTIVE Study. Comparison of SGA oral medications and a long-acting injectable SGA: The PROACTIVE study. Schizophrenia Bulletin, 2015; 42(2):449-459.
Das AK, Malik A, Haddad PM. A qualitative study of the attitudes of patients in an early intervention service towards antipsychotic long-acting injections. Therapeutic Advances in Psychopharmacology, 4(5):179-185.
Iyer S, Banks N, Roy M-A, Tibbo P, Williams R, Manchada R, Chue P, Malla A. A qualitative study of experiences with and perceptions regarding long-acting injectable antipsychotics: Part 1—patient perspectives. Canadian Journal of Psychiatry, 58(5 Suppl 1):14S-22S.
Kirson NY, Weiden PJ, Yermakov S, Huang W, Samuelson T, Offord SJ, Greenberg PE, Wong BJO. Efficacy and effectiveness of depot versus oral antipsychotics in schizophrenia: synthesizing results across different research designs. Journal of Clinical Psychiatry 2013; 74:568-575.
Kishimoto T, Nitta M, Borenstein M, Kane JM, Correll CU. Long-acting injectable versus oral antipsychotics in schizophrenia: a systematic review and meta-analysis of mirror-image studies. Journal of Clinical Psychiatry 2013, 74:957-965.
Kishimoto T, Robenzadeh A, Leucht C, Leucht S, Watanabe K, Mimura M, Borenstein M, Kane JM, Correll CU. Long-acting injectable vs oral antipsychotics for relapse prevention in schizophrenia: a meta-analysis of randomized trials. Schizophrenia Bulletin 2014; 40:192-213.
Kishimoto T, Hagi K, Nitta M, Leucht S, Olfson M, Kane JM, Correll CU. Effectiveness of long-acting injectable vs. oral antipsychotics in patients with schizophrenia: a meta-analysis of prospective and retrospective cohort studies. Schizophrenia Bulletin 2018; 44:603-619.
Manchada R, Chue P, Malla A, Tibbo P, Roy M-A, Williams R, Iyer S, Lutgens D, Banks N. Long-acting injectable antipsychotics: evidence of effectiveness and use. Canadian Journal of Psychiatry, 58(5 suppl 1):5S-13S.
Rosenheck RA, Krystal JH, Lew R, Barnett PG, Fiore L, Valley D, Thwin SS, Vertrees JE, Liang MH, for the CSP555 Research Group. Long-acting risperidone and oral antipsychotics in unstable schizophrenia. New England Journal of Medicine, 2011; 364(9):842-851.
Weiden PJ, Schooler NR, Weedon JC, Elmouchtari A, Sunakawa-McMillan A. Maintenance treatment with long-acting injectable risperidone in first-episode schizophrenia: a randomized effectiveness study. Journal of Clinical Psychiatry 2012; 73(9):1224-1233.