Comparative risk of Parkinsonism associated with olanzapine, risperidone and quetiapine in older adults‐a propensity score matched cohort study

The purpose of this study was to examine the incidence of Parkinsonism in new users of second‐generation antipsychotics (SGAs) in older adults (≥65 years). In the secondary analyses, we examined the risk of Parkinsonism by type and dose of SGA and conducted age‐sex interactions.


| INTRODUCTION
Second-generation antipsychotics (SGAs) are widely prescribed in older adults for the management of behavioural and psychological symptoms of dementia (BPSD). 1 SGAs are associated with cardiovascular, metabolic and neurological adverse effects in older adults. [2][3][4] One of the most debilitating neurological adverse effects of SGAs in older adults is drug-induced Parkinsonism. 5,6 Evidence from clinical trials and observational studies show an increase risk of Parkinsonism associated with SGAs. 7 Selection criteria for clinical trials are very stringent and often exclude older people with several comorbidities and are underpowered to detect small, but very important, differences in the rates of uncommon adverse effects. 8,9 This is particularly relevant to Parkinsonism associated with SGA use in older adults. In a real-world setting, older adults have more severe BPSD, higher comorbidity, and are frailer and hence have a higher baseline risk of harms from drug exposures than patients recruited in a clinical trial. 10 Hence, extrapolation of evidence derived from clinical trials to real-world patients is barely accurate.
Studies have shown that SGAs may have different extrapyramidal symptoms (EPS) adverse effect profile. 11 Hence epidemiological studies are needed to examine multiple antipsychotic exposures (type and dose) together with social and clinical risk factors for a longer period of follow up time from large and representative sample of real-world population of older adults. 12 However, observational studies are likely to provide biased estimates of risks due to confounding, because background characteristics of individuals may favour the use of certain treatments, which results in imbalance in observed characteristics between treatment groups. 13 A large cohort study (N = 25 769) conducted among older adults with dementia in Ontario found a 30% increased risk of Parkinsonism associated with first-generation antipsychotic use relative to SGA use 14 with a hazard ratio of 1.69 (95%CI 1.04-1.58) reported. Despite careful selection of covariates, the study was potentially vulnerable to confounding due to imbalance in the observed baseline characteristics in the exposed and non-exposed groups. Similarly, a population-based retrospective cohort study involving older adults in the Canadian province of Manitoba found a lower risk of Parkinsonism associated with incident use of risperidone relative to first-generation antipsychotic use, at 30 days following exposure to risperidone the hazard ratio was 0.38 (95% CI 0.22-0.67). 15 This study did not investigate the effect of dose on incidence of Parkinsonism in older adults. In order to understand and quantify the risk of new-onset Parkinsonism posed by SGAs we need reliable population-level evidence with appropriate methods to control for confounding. 16 Currently, to our knowledge, research estimating the causal effect of new-onset Parkinsonism associated with SGAs under the context of confounding due to variations in background characteristics and comorbidity of individuals between treatment groups is limited.
In this study, we examined the association of Parkinsonism associated with SGAs using population-level data with adequate confounding control. We hypothesize that the association of SGAs on Parkinsonism is a class effect and all SGAs will have similar effect sizes concerning new-onset Parkinsonism.

| Study population
Eligible patients were older adults aged 65 years of age at entry into the cohort on 1 January 2007. We defined the cohort entry as the date of first prescription for any atypical antipsychotic (olanzapine, risperidone or quetiapine). We defined incident use as a new prescription for olanzapine or risperidone or quetiapine with no previous prescription claims for those antipsychotics during the 12 months before cohort entry. We excluded older adults with a history of Parkinson's disease and those dispensed drugs known to cause secondary Parkinsonism.
We also excluded individuals treated with drugs for the management of

| Exposures and covariates
We used dispensing claims data to determine SGA treatment exposure. We obtained de-identified dispensing claims data for individuals prochlorperazine, fluphenazine and haloperidol. 18,19 The SGAs (ranked low to high risk of inducing Parkinsonism) included clozapine, olanzapine (long acting) aripiprazole, amisulpride 6 and others included amiodarone, lithium, methyldopa, and tranylcypromine.
Covariates of interests included socio-demographic characteristics, exposures to any above-mentioned effect-modifiers, and comorbidity.
We used the medicines comorbidity index (MCI) to derive a comorbidity score. The MCI is an appropriate tool for measuring comorbidity, validated on the NMDS, and is an appropriate index for adjusting comorbidity in pharmacoepidemiological studies. 20 In brief the MCI includes

F I G U R E 1 Cohort study design
To calculate the exposure length, we counted the total number of days the individual was dispensed the SGA of interest between the cohort entry date and the end of the last prescription before censoring, and accounting for the possibility that prescriptions can overlap.

| Outcomes
The In the secondary analyses, we used the generalized propensity score (GPS) method 23 to quantify the change in risk of Parkinsonism in response to one unit increase in the average daily dose of the SGA.
GPS assumes that the multiple treatments are continuous variables.

| Sensitivity analyses
We conducted additional sensitivity analyses. To reduce potential bias from differential follow-up times between SGA users, we limited the maximal follow-up to 100 days.
All analyses were performed with the use of R software, version 3.2.1.5. 24 3 | RESULTS

| Study participants
There were 2220 incident SGA users with 922 622 person-days of follow up ( Table 1). The total population were weighted by inverse probability treatment weighted to mitigate confounding. PSB calculations indicate that all confounders were balanced within the three groups after IPTW weighting (PSB < 0.2, Figure 2). There were 487 of new-onset Parkinsonism among the 2220 incident SGA users.

| Outcomes
Survival analysis confirms that olanzapine and risperidone are associated with "higher" risk of Parkinsonism compared to quetiapine ( Figure 3). Despite olanzapine being associated with a "higher" risk of Parkinsonism compared to quetiapine, users of olanzapine are less likely to discontinue treatment or switch to another SGA ( Figure 4).
However, users who initiate risperidone or olanzapine are more likely to switch to quetiapine (Figure 4). Schoenfeld residuals analysis permits the calculation of hazard ratios of new-onset Parkinsonism due to SGA exposures by COXPH regression. Compared to quetiapine users, hazard ratio calculations indicate that the risks of Parkinsonism onset in individuals exposed to risperidone and olanzapine are 31% and 76% higher respectively ( Figure 5).

| Secondary analyses
Dose-response risk of Parkinsonism revealed a higher risk of newonset Parkinsonism with olanzapine followed by risperidone and quetiapine ( Figure 6). Age-sex interactions revealed risk of Parkinsonism in 65 to 74-year-old was higher with risperidone compared to olanzapine for both sexes, but the risk increased significantly high for both sexes with olanzapine in the 85+ age group (Figure 7).

| DISCUSSION
A propensity score analyses conducted on a population of older adults showed that SGAs increased the risk of new-onset Parkinsonism. Our study showed that the olanzapine followed by risperidone had the highest risk compared to quetiapine in older adults. Our study findings are partly in contrast to that reported by Marras et al who found that the risk of Parkinsonism in older adults with dementia was similar among quetiapine, olanzapine or risperidone users. 7 We did not restrict our analyses to adults with dementia and that the difference in the study populations may be the primary reason for this discordant finding.
The safety profiles of SGAs vary because of their dissimilarities to dopamine-2 (D2) receptor, 5-hydroxytryptamine receptor 2A (5-HT2A) affinity and anticholinergic activity. 25 Olanzapine, but not risperidone or quetiapine has central anticholinergic activity, and both olanzapine and risperidone have higher 5-HT2A affinity relative to quetiapine. It is postulated that both antiserotonergic and anticholinergic actions may help to counter the impact of the D2 blockade of these SGAs on the basal ganglia. 26 Our findings are plausible with these biological mechanisms.
Our study demonstrated a dose-response risk of Parkinsonism associated with SGAs. This finding is consistent with a retrospective cohort study conducted in older adults in Ontario. However, their study did not stratify the cohort by type of SGAs and dose. 14  higher risk of EPS than olanzapine, but the trial included risperidone doses (4-12 mg day) which has much higher than the average daily dose in our study. 27 Interestingly, our study found differences in age and sex interactions by type of SGA. A higher risk of Parkinsonism is associated with olanzapine in both sexes across all ages compared to risperidone.
While there is a paucity of literature to compare age and sex interactions with specific type of SGAs, a large population-based study found that the incidence of drug-induced Parkinsonism increased with older age and was higher in women at all ages. 28 Quetiapine may be a suitable alternative following an unsuccessful trial of non-pharmacological interventions for the management of BPSD because of its lesser propensity to cause Parkinsonism than other SGAs and has a relatively safer metabolic and cardiovascular profile than olanzapine or risperidone.

| Strengths
This study has several strengths, including its large size, nation-

| Limitations
We extracted the exposures and the outcomes from the administrative data sources used in these analyses. We did not ascertain if F I G U R E 5 Hazard ratios for newonset Parkinsonism with secondgeneration antipsychotics F I G U R E 6 Dose-response risk of new-onset Parkinsonism for secondgeneration antipsychotics individuals prescribed SGAs have taken them as this could potentially lead to misclassification of antipsychotic exposure. We excluded clozapine in our study due to the small numbers of clozapine users. We only included variables available in the prescription and hospital discharge data for computing propensity scores and did not validate the ICD-10-AM codes for Parkinsonism to confirm a diagnosis. Literature has also identified that there is a potential for under-reporting and under recognizing Parkinsonism in hospital patients and the elderly. 29,30 The NMDS data only allowed us to identify individuals with new-onset Parkinsonism in patients admitted to hospitals. Hence replication of this study in a more general population with a clinical assessment of Parkinsonism is needed. The retrospective nature of our study design is prone to bias and residual confounding despite controlling our analyses for age, sex, ethnicity and comorbidity using propensity scores.

| CONCLUSION
The study found that the risk of new-onset Parkinsonism in older adults is 31% and 76% higher with risperidone and olanzapine respectively compared to quetiapine. A higher risk of Parkinsonism is associated with olanzapine in both sexes across all ages compared to risperidone.

ETHICS STATEMENT
The Ethical Implications of Research Activity Form (EIRA1-2960) to conduct this study was approved on 8 April 2019 by the University of Bath.
F I G U R E 7 Risk of new-onset Parkinsonism for second-generation antipsychotics stratified by sex and age groups