Real-World Utilization Patterns of Long-Acting Injectable Antipsychotics in Canada: A Retrospective Study

Abstract

Objective: The objective of this study was to analyze the real-world prevalence of long-acting injectable (LAI) antipsychotic use and determine when LAIs are being used in sequencing of antipsychotic medications among Canadian patients with schizophrenia. Methods: This was a retrospective, longitudinal cohort study using Canadian pharmacy prescription data between August 2005 and June 2017. Patients with inferred schizophrenia spectrum disorder were indexed on the date of their first antipsychotic prescription and analyzed for minimum 12 months to track lines of antipsychotic therapy and LAI utilization. Results: A total of 16,300 patients were identified for analysis. 48.2% and 46.0% of index antipsychotic prescriptions were prescribed by a general practitioner/family medicine doctor and psychiatrist, respectively. 1,062 (6.5%) patients used an LAI during the study period. Of those patients, 789 used an LAI within two years of index (74.3% of LAI users; 4.8% of all patients). The majority of LAI use (62.0%) occurred in the third line of therapy or later. 65.0% of patients had tried at least two therapy lines, and most patients reported gaps of six months to one year between treatment lines. Conclusion: Despite their potential to reduce relapse in schizophrenia by improving treatment adherence, this study shows LAIs continue to be under-utilized in Canada. When used, LAIs are positioned late in sequencing of antipsychotic medications, often not initiated until years after diagnosis. Continued preference for oral APs with poor adherence may be negatively impacting prognosis and exacerbating burden of schizophrenia. Efforts should be invested to understand barriers to LAI uptake and advocate for earlier, widespread use of LAIs.

Introduction

Schizophrenia is a common, complex mental illness characterized by positive symptoms (e.g., delusions and hallucinations), negative symptoms (e.g., social withdrawal and emotional flattening), and impaired cognitive ability.^1,2 It has a prevalence of 1% in Canada, with onset of symptoms typically occurring in late adolescence or early adulthood.^1,3 Schizophrenia can go undiagnosed for several years, during which symptoms can be severe. Following diagnosis and proper treatment, symptoms tend to stabilize or improve. ⁴ Uncontrolled schizophrenia can have lasting consequences on health outcomes, including reduced life expectancy and high rates of comorbid illness.^3,5 It is associated with significant personal and economic burden, accounting for almost 2% of national healthcare expenditures. ⁶

The psychotic symptoms associated with schizophrenia are highly treatable with antipsychotic (AP) therapy, ⁷ and most patients achieve remission within their first year of treatment. ⁸ With continuous treatment and minimal relapse, the prognosis for schizophrenia would likely be much improved. However, adherence to medication is critical to treatment success, and roughly 70% of patients using oral APs are non-adherent.^9,10 Treatment discontinuation increases the risk of relapse almost five-fold,^11–13 and even brief periods of non-adherence or partial non-adherence can lead to relapse. ¹⁴ Indeed, over 80% of patients with schizophrenia relapse within five years of their first psychotic episode, and many experience subsequent relapses in the next few years.^11,15 Relapse can negatively impact psychosocial and functional outcomes, increase risk of harm to self or others, exacerbate social isolation, and compromise education or employment status.^12,16,17 Relapse episodes can also produce an attenuated or delayed response to subsequent AP treatment, further interfering with the ability to achieve remission ¹⁸ and thus highlighting the importance of rapid treatment initiation and proactive measures to improve treatment adherence.^5,19

Long-acting injectable (LAI) AP formulations may present a means of improving treatment adherence. In comparison to oral APs that require daily dosing, LAIs are given as intramuscular injections every two weeks to three months and have shown superior adherence versus oral AP formulations. ²⁰ LAIs also allow physicians to directly observe treatment adherence at patients’ monthly injection appointments, removing the biases associated with patient self-report. ⁵ There is growing evidence demonstrating the superiority of LAIs in reducing risk of medication discontinuation, relapse, and hospitalization.^21–25 However, LAIs have typically been reserved for patients with greater illness severity or greater risk of non-adherence. This heterogeneity in patient populations likely biases unadjusted comparisons of relapse prevention between oral APs and LAIs, and may partly explain inconsistent results in the literature regarding the superiority of LAIs.^21,22

Despite the advantages of LAIs, prevalence of their use is low compared to oral APs, and most prescribers do not consider them as a potential first-line treatment despite increasing advocacy for early LAI initiation.^26–28 Global estimates show LAIs are consistently under-used relative to oral APs, with the proportion of schizophrenia patients using LAIs estimated at just 6.3% in Canada. ²⁹ The continued reliance on oral APs and under-utilization of LAIs in Canada may be perpetuating a vicious cycle of non-adherence, relapse, and hospitalization. Increased and earlier uptake of LAIs may present an opportunity to interrupt this cycle and improve the prognosis for schizophrenia. To this end, it is important to understand real-world utilization of LAIs and their positioning in the context of current Canadian treatment algorithms. The objective of this study was to analyze the real-world prevalence of LAI use and conduct a line of therapy analysis to determine when LAIs are being used in the sequencing of AP medications among Canadian patients with schizophrenia.

Methods

This was a retrospective, longitudinal cohort study using Canadian retail pharmacy prescription data between August 2005 and June 2017. Patients were selected between August 2006 and March 2016 and indexed at their first observed AP prescription during this selection period. These patients were then followed in the database for minimum 12 months post-index to study lines of AP therapy and LAI utilization. A one-year look-back period and a three-month look-forward period were used to ensure patients were previously naïve to their index medication and confirm continued patient activity in the database (Supplementary Figure 1).

Data Source

Longitudinal prescription data of de-identified patients were extracted from the IQVIA Canadian Longitudinal Prescription (LRx) database. These data track prescription purchases of individuals at retail pharmacies across the Canadian provinces, with a consistent capture of publicly and privately reimbursed prescriptions and cash transactions. It does not include some specialty pharmacies, or hospital transactions. The database does not have the ability to track patients between different retail pharmacies, thus individuals visiting multiple pharmacies within the database appear as a unique patient at each store. To mitigate the potential bias this could introduce, an active enrolment criterion was incorporated into the study eligibility criteria to screen out patients who have switched pharmacies during the study period. At the time of this study, the LRx database captured approximately 71% of national retail pharmacy prescriptions in Canada. The remaining 29% of retail pharmacy prescriptions not captured occurred at pharmacies that do not contribute to the LRx database, based on administrative considerations.

Study Population

All patients needed at least one prescription for an AP during the selection period and at least two years of prescription history to be included. LRx data do not include information on patients’ diagnosis or drug indication; therefore, schizophrenia and psychotic disorder patients were identified using an indication algorithm that inferred diagnosis based on the patient's age and prescription history (Supplementary Figure 2). Rules for the algorithm were primarily derived from current guidelines for the pharmacological treatment of patients with schizophrenia spectrum disorder,^30–32 as well as the broader body of literature to identify therapies most common among patients with schizophrenia, and age groups at highest risk of schizophrenia. Face validation of the algorithm was conducted through review of the rules by a clinical expert in the field (Dr. Ofer Agid). This algorithm was applied to the dataset to infer three cohorts: (1) schizophrenia spectrum disorders; (2) bipolar and schizoaffective disorder; and (3) sedation, anxiety, and age-related AP indications. Cohort 1, “schizophrenia spectrum disorders,” were the focus of this study. Once patients were inferred with schizophrenia spectrum disorder, additional study eligibility criteria were applied:

Inclusion:

• First AP prescription between August 2006 and April 2016

• Age 16–35 years at index

• Meets active enrolment criteria: Active in database at least 12 months pre- and 15 months post-index

Exclusion:

• Indexed on low-dose quetiapine in Line 1 and no products other than quetiapine in Lines 1–5

Although inclusion in the cohort for schizophrenia spectrum disorders was restricted to ages 16–65, the additional inclusion criteria restricting the study to patients aged 16 to 35 at index were added to capture patients early in the course of disease. Patients receiving only low-dose APs were assumed to be using APs for reasons other than psychosis and thus were excluded.

Study Variables

Baseline data were assessed at index and included index year, sex, age, province, payer, and prescriber specialty. Index AP therapy was considered the first therapy line, and subsequent therapy changes were studied for minimum 12 months post-index. A new therapy line was started whenever a patient made one of the following changes:

• Add-on: Added a new therapy to the previous therapy

• Reduce: Stopped one of the 2 + therapies from the previous line therapy

• Switch: Stopped the previous line therapy and started a new therapy

This analysis focused on changes between AP molecules, with the exception of molecules with both oral and injectable formulations, which were analyzed separately (Supplementary Table 1). A 90-day grace period between prescriptions (in addition to the days’ supply of the prescription) was used to determine whether a patient was still using the same treatment while allowing for a baseline level of non-compliance. A gap within or between therapy lines occurred if no product was claimed for a period of days’ supply plus 90 days. Gap periods were summed to determine the number of medication-free days that occurred within each respective therapy line. Time on therapy line included gap periods and was calculated from the first claim that marks the start of the line to the start of the subsequent line (Supplementary Figure 3).

Data Analysis

No statistical hypothesis testing was performed; all analyses were purely descriptive. Categorical variables were summarized using counts and proportions (%). Continuous variables were summarized using means, standard deviations (SD), medians, and interquartile ranges (IQRs). All analyses were performed using SAS 9.4 (SAS Institute Inc., Cary, NC).

Results

Baseline Demographics

The indication algorithm captured 195,316 de-identified patients with inferred schizophrenia spectrum disorders in the LRx database who had their first AP prescription between August 2006 and March 2016. 16,300 patients were included in this study after applying the pre-defined eligibility criteria (Figure 1), and the average length of prescription history available for analysis post-index was 3.9 years (IQR 2.5–6.0). The number of patients indexed increased each year, with an approximately equal split of male (50.2%) and female (49.8%) patients. 66.7% of patients were aged 26–35 at index, and nearly half of patients filled their index prescriptions in Ontario (47.1%). 39.3% of patients had coverage by public insurance at index, and the majority of index AP prescriptions with prescriber specialty data were prescribed by a psychiatrist (46.0%) or a general practitioner/family doctor (GP/FM; 48.2%). For patients with an LAI index AP, 77.3% were prescribed by a psychiatrist and 19.5% by a GP/FM (Table 1).

Figure 1.

Patient selection flowchart for eligibility criteria following identification of schizophrenia spectrum by indication algorithm.

Table 1.

Baseline Demographics.

Total patients	16,300
Follow-up time (years), median (IQR)	3.9 (2.5–6.0)
Sex, n (%)a
Male	8,177 (50.2)
Female	8,123 (49.8)
Age, n (%)a
16–25	5,432 (33.3)
26–35	10,868 (66.7)
Province, n (%)a
Ontario	7,683 (47.1)
Quebec	2,092 (12.8)
Alberta	1,994 (12.2)
British Columbia	1,619 (9.9)
Manitoba	733 (4.5)
New Brunswick	701 (4.3)
Saskatchewan	672 (4.1)
Nova Scotia	588 (3.6)
Newfoundland	145 (0.9)
Prince Edward Island	73 (0.4)
Index year, n (%)a
2006	284 (1.7)
2007	411 (2.5)
2008	582 (3.6)
2009	739 (4.5)
2010	1,094 (6.7)
2011	1,328 (8.1)
2012	1,806 (11.1)
2013	2,331 (14.3)
2014	2,884 (17.7)
2015	3,633 (22.3)
2016	1,208 (7.4)
Payer, n (%)ab
Public	6,413 (39.3)
Private	5,852 (35.9)
Cash	4,035 (24.8)
Prescriber specialty, n (%)ac
Total with prescriber data	13,044
GP/FM	6,282 (48.2)
Psychiatrist	5,999 (46.0)
Other	763 (5.8)
First line LAI prescriber specialty, n (%)ac
Total with prescriber data	185
GP/FM	36 (19.5)
Psychiatrist	143 (77.3)
Other	6 (3.2)

^a Reported at index, or for index AP prescription.

^b Where an index prescription was covered by multiple payers, the payer covering the largest proportion of the cost was considered the sole payer.

^c Prescriber data was not available for all patients. Prescriber data is obtained through mapping of prescription to a database of physician IDs containing physician specialty. These data are limited by provincial variations in privacy regulations, variability in data supplied by contributing pharmacies, and data entry errors. Percentages reflect the proportions of those with prescriber data available.

Abbreviations: GP/FM, general practitioner/family medicine; LAI, long-acting injectable.

LAI Utilization

Overall, 1,062 (6.5%) patients used an LAI at some point during the follow-up period, and this proportion remained relatively consistent across index years (Figure 2). 789 (4.8%) patients used an LAI within two years of index, although this proportion increased from the beginning of the selection period (3.9% in 2006) to the end (5.6% in 2016) (Supplementary Figure 4). When LAIs were used, it was rarely as a first-line therapy, with only 13.3% of LAI initiations occurring in Line 1, and the majority (62.0%) occurring in Line 3 and beyond (Figure 3).

Figure 2.

Proportion of patients that used an LAI at any point in treatment journey, by index year.

Figure 3.

Distribution of LAI use across lines of therapy, as a proportion of all uses of an LA.

Lines of Therapy

65.0% of patients tried at least two therapy lines, 37.4% tried at least three, 21.0% tried at least four, and 12.1% tried at least five (Table 2). 75.6% of patients that moved to a second line reported an average gap of 355 days between initiation of Line 1 and initiation of Line 2 with no medication (i.e., 355 cumulative medication-free days, but not necessarily consecutive). Medication-free gaps within or between each therapy line were observed for most patients, but the duration of this gap decreased with each line (185 days by Line 5 to Line 6) (Table 2). Of all antipsychotic therapies, oral or injectable, quetiapine was the most common first line therapy (34.0%), followed by oral risperidone (24.7%) and olanzapine (17.0%). Oral aripiprazole became increasingly common as a first-line therapy, growing from less than 2.0% of index products in 2010 to 15.1% in 2016 (Supplementary Figure 5).

Table 2.

Number of Therapy Lines and Gaps between Therapy Lines.

Number of therapy lines	1 +	2 +	3 +	4 +	5 +
Total patients, n (%)	16,300 (100.0)	10,596 (65.0)	6,103 (37.4)	3,415 (21.0)	1,970 (12.1)
Line gaps	Line 1-2	Line 2-3	Line 3-4	Line 4-5	Line 5-6
Total patients with gap, n (%)a	8,012 (75.6)	4,411 (72.3)	2,356 (69.0)	1,308 (66.4)	643 ( − )
Average duration of gapb (days), M (SD)	355 (474)	255 (363)	229 (324)	202 (290)	185 (292)

^a Percentages are expressed as the proportion of patients that had a gap in treatment out of those that moved on to the subsequent therapy line.

^b Duration of gap refers to the cumulative number of gap-free days between the start of one line, and the start of the subsequent line.

Discussion

For many patients with schizophrenia, the most significant obstacle to remission is relapse related to treatment discontinuation or poor adherence.^11–13 Relapse can trigger significant decline in a patient's condition ¹⁶ and confers otherwise preventable economic burden. ³³ APs are the cornerstone of treatment for schizophrenia, and if properly adhered to, can have robust effects on symptom management and relapse prevention. Despite LAIs having the potential to improve adherence, the uptake of LAIs remains low in Canada. ²⁹

Although global adoption of LAIs has been slow, uptake in Canada has been particularly poor relative to global counterparts. Estimates of LAI use from countries such as Austria, Belgium, Sweden, Australia, and New Zealand range from 15–50%. ²⁹ In the U.S. and U.K., LAI use has been estimated between 18–28% and 29–80%, respectively, with a trend toward improved uptake over time. ²⁹ Comparatively, this analysis showed less than 7% of Canadian patients with schizophrenia used an LAI during treatment, indicating a pronounced delay in LAI uptake in Canada that cannot be attributed solely to methodological differences in estimating LAI use. Further, when LAIs were used, they were typically positioned later in the course of treatment, in Lines 3 or beyond. This is problematic considering the frequency and speed with which patients tend to relapse, particularly during periods of partial- or non-adherence. ¹⁶ The vast majority of patients experience symptom exacerbation or relapse within the first few years of treatment, with some reporting symptom recurrence within just days or weeks of treatment discontinuation, even following extended periods of consistent AP treatment.^16,34 Given the six-month to one-year medication-free gaps during therapy lines observed in this study, it could be several years after diagnosis before a patient is prescribed an LAI, during which multiple relapse episodes have likely already occurred. Given these results are primarily driven by the over-representation of oral AP use in this data, these medication-free periods likely reflect the poor adherence to oral AP treatment that has been documented many times over, and only further emphasizes the importance of increasing LAI adoption.

Although this study showed an upward trend in LAI use from study start to end, this increase appeared very modest and did not indicate meaningful improvement over previous estimates of LAI use in Canada. ²⁹ The proportion of patients who used an LAI within two years of index was just 2.4% for patients indexed in 2008, but consistently grew to 5.6% for patients indexed in 2016 (Supplementary Figure 4). The proportion of patients who used an LAI at any time stayed roughly the same across index years, but given the shorter analysis time for patients indexed in later years, this indicates a trend toward earlier LAI initiation. In absolute terms, however, these results reflect changes among only a few hundred patients. There were only 211 patients in 2015 who had ever used an LAI, 179 more than was reported in 2006. Similarly, only 199 more patients used an LAI within two years of index in 2015 than in 2006. In contrast, thousands of patients reported no use of LAIs at all.

There are a multitude of reasons why LAIs are not widely used or offered earlier, as described in previous reviews of barriers to LAI adoption. ³⁵ However, it should be noted that many, if not most of these barriers stem from the prescriber/physician rather than the patient. Previous studies of patient attitudes toward LAIs have found the majority of patients are willing to try or would prefer less frequent injections over daily oral medications but are not being offered LAIs or involved in treatment decision-making with physicians.^36–38 The results of this analysis suggest the reluctance to prescribe LAI APs may be particularly significant among GP/FM doctors, who prescribed 48.2% of total index AP prescriptions, but only 19.5% of index LAI prescriptions (Table 1). This suggests that reluctance to adopt LAIs as a first-line AP therapy may be largely driven by primary care physicians, a speculation that is further supported by a recent real-world evidence study in Quebec, where GPs prescribed less than 20% of LAIs compared to psychiatrists who prescribed over 80%. ³⁹ Going forward, attention must be paid to how GPs, who are the primary point of access into the healthcare system, influence treatment trajectory for patients with schizophrenia, and what opportunities exist to improve knowledge, experience, and comfort with LAIs.

Despite overwhelming evidence of poor adherence to oral APs, and evidence of LAIs superiority in relapse-prevention, the results of this study and previous research suggest Canadian prescribers remain either unwilling or greatly hesitant to prescribe LAI APs to patients with schizophrenia, particularly as a first-line therapy. These prescribing behaviors may be directly impacting the prognosis of schizophrenia patients, and by extension the associated burden related to persistent psychotic symptoms and hospitalizations.^6,33,39,40 In Canada, direct healthcare and non-healthcare costs of schizophrenia have been estimated at $2.02 billion with hospitalizations accounting for over $1.23 billion, ⁶ and the excess cost of relapse has been estimated at $6,033 to $32,753 USD for the 12-month period post-relapse. ⁴¹ Previous studies estimate LAIs, relative to oral APs, could reduce relapse by upward of 30%^24,28 and significantly reduce hospitalizations, ER visits, and resource utilization. ⁴² With the ability to reduce the rate of relapse and associated health expenditures, widespread adoption of LAIs as a first-line treatment for schizophrenia could translate to hundreds of millions in annual savings for the Canadian healthcare system. Less quantifiable, but equally important, is the potential to improve the overall well-being of the patient, as well as their family, caregivers, and treating physicians, for whom frequent relapse and symptom recurrence can dramatically reduce quality of life.^39,40

This is the first study to use real-world data to study usage of LAIs and other AP therapies among Canadian patients with schizophrenia. The use of routinely collected prescription data allowed for a substantial sample size that is broadly representative of the population, not restricted by geography, prescriber, or insurance. In addition, the use of real-world data captured patient behaviors in a naturalistic manner that did not rely on patient or prescriber recall or self-report and may provide a more accurate picture of patient treatment behaviors. However, use of real-world data has some limitations.

First, the prescription database did not include information on diagnoses, and so indication was inferred using an algorithm applied to available data. This algorithm was developed from literature and expert opinion and was not formally validated in a separate study. Thus, this analysis was only able to analyze “suspected” schizophrenia cases rather than confirmed diagnoses, and it is possible the sample included diagnoses other than schizophrenia and missed true schizophrenia cases. However, the algorithm was purposefully designed to rule out patients with the potential to be mislabeled schizophrenia. For example, an exclusion rule for prescription history of mood stabilizers was used to distinguish bipolar and schizoaffective patients that may be otherwise similar to schizophrenia patients in age and medication history, and patients with exclusive use of low-dose APs were excluded to minimize the inclusion of non-schizophrenia patients using APs for sedation and anxiety. The relative capture of bipolar versus schizophrenia patients also aligned with Canadian prevalence estimates, increasing confidence that the algorithm had not mistakenly captured a substantial number of non-schizophrenia diagnoses. Future studies performing formal validation of indication algorithms such as the one used here would be valuable in retrospectively assessing the validity of findings.

Second, lines of therapy were inferred based on changes in patients’ prescription histories and may not perfectly capture clinical intention. For example, what presents in the data as combination therapy of two treatments may instead be a period of cross-tapering as the patient switches from one product to another. In this regard, it is possible that the methodology for defining lines of therapy in this analysis may have overestimated total lines.

Third, the LRx database is limited to a capture of 71% of retail pharmacy prescriptions in Canada, which represents most of the major pharmacy franchises and stores across the country. The remaining 29% of prescriptions not captured come primarily from smaller, independent stores, for which it would not be feasible to manage consistent data-sharing. Although this results in a capture rate of less than 100%, this is attributable primarily to administrative factors, and the 71% capture should be broadly representative of the majority of the population.

Fourth, the prescription database only captures transactional data reported by retail pharmacies, from which it is not possible to confirm patients are taking medication as prescribed. This further highlights the value of LAIs, though, where treatment adherence is often directly observed by the physician, as opposed to oral APs, which requires significant assumptions around if and how patients are using their medication. In addition, the database tracks patients’ prescriptions at the store level and does not have the ability to track patients between different pharmacies. This could result in double-counting of patients or a lack of visibility of the true patient course for those obtaining medication from multiple pharmacies over the study period. To minimize the impact of such patients, a rule was applied requiring patients to have filled a prescription at the same pharmacy at least 12-months pre- and 15-months post-index. This rule would have excluded patients who may erroneously appear to have discontinued therapy when in fact they have simply traveled to a new pharmacy.

Lastly, clinical and health economic outcomes were not included as they were beyond the scope of this study. Further research is warranted to better understand the real-world impact of LAIs on treatment patterns, rates of relapse and remission, overall patient health, and economic burden of schizophrenia. Nevertheless, these results establish the overwhelming under-utilization of LAIs in Canada and provide ample justification for continued efforts to better understand and resolve barriers to LAI adoption.

Data Access

Further information on the LRx dataset can be obtained from IQVIA Canada.