Antipsychotic Medication and Risk of Metabolic Disorders in People With Schizophrenia: A Longitudinal Study Using the UK Clinical Practice Research Datalink

Abstract

Background and Hypothesis

Antipsychotics are first-line drug treatments for schizophrenia. When antipsychotic monotherapy is ineffective, combining two antipsychotic drugs is common although treatment guidelines warn of possible increases in side effects. Risks of metabolic side effects with antipsychotic polypharmacy have not been fully investigated. This study examined associations between antipsychotic polypharmacy and risk of developing diabetes, hypertension, or hyperlipidemia in adults with schizophrenia, and impact of co-prescription of first- and second-generation antipsychotics.

Study Design

A population-based prospective cohort study was conducted in the United Kingdom using linked primary care, secondary care, mental health, and social deprivation datasets. Cox proportional hazards models with stabilizing weights were used to estimate risk of metabolic disorders among adults with schizophrenia, comparing patients on antipsychotic monotherapy vs polypharmacy, adjusting for demographic and clinical characteristics, and antipsychotic dose.

Study Results

Median follow-up time across the three cohorts was approximately 14 months. 6.6% developed hypertension in the cohort assembled for this outcome, with polypharmacy conferring an increased risk compared to monotherapy, (adjusted Hazard Ratio = 3.16; P = .021). Patients exposed to exclusive first-generation antipsychotic polypharmacy had greater risk of hypertension compared to those exposed to combined first- and second-generation polypharmacy (adjusted HR 0.29, P = .039). No associations between polypharmacy and risk of diabetes or hyperlipidemia were found.

Conclusions

Antipsychotic polypharmacy, particularly polypharmacy solely comprised of first-generation antipsychotics, increased the risk of hypertension. Future research employing larger samples, follow-up longer than the current median of 14 months, and more complex methodologies may further elucidate the association reported in this study.

Introduction

Schizophrenia is a serious long-term mental health disorder with a prevalence of approximately 1% worldwide.¹ Long-term prognosis is typically poor for at least a third of patients, with many experiencing residual psychotic symptoms,² poor social functioning, unemployment and a reduction in quality of life.³ People with schizophrenia die on average 20 years before the healthy population and this gap may be widening.⁴ Part of the reason for the life expectancy gap is an increased risk of preventable metabolic disorders such as diabetes, obesity, hypertension, and hyperlipidemia.⁵

Antipsychotic medication is the first-line treatment for schizophrenia.⁶ There is strong evidence that a side-effect of some antipsychotic drugs is an increased risk of diabetes, hypertension, and hyperlipidemia.⁷ This is particularly true of second-generation antipsychotics.⁷ Since the early 1990s, second-generation (atypical) antipsychotics have commanded an ever-increasing share of antipsychotic drug prescriptions in Western countries, supplanting first-generation (conventional/typical) drugs in response to concerns about their extra-pyramidal (EPSs) and anticholinergic side-effects.^8,9 Second-generation antipsychotics are heterogenous in their pharmacology with some, such as risperidone, having similar affinity for dopamine D2 receptors to prototypical first-generation drugs like chlorpromazine and haloperidol, while others such as olanzapine, quetiapine, and the original atypical drug, clozapine, first introduced in the early 1970s, have far lower dopamine D2 receptor affinity.¹⁰ Another sub-group, which includes aripiprazole, are Dopamine D2 receptor, partial agonist.¹¹ Some consider aripiprazole and related drugs to constitute a distinct “third generation” of antipsychotics^10,12 but for this study, we consider them as second-generation drugs. Atypical¹³ antipsychotics have a reduced rate of EPS, but are associated with other side-effects, such as weight gain.^14,15

Antipsychotic Polypharmacy (APP) is the prescription of more than one antipsychotic medication concurrently. Evidence from a review of systematic reviews showed that most studies report some associations between APP and aspects of metabolic syndrome, but data are inconsistent and of low quality.¹⁵ Evidence-based guidelines either advise against co-prescription of antipsychotics,^6,16 or point to the lack of evidence of efficacy from large randomized controlled trials¹⁷ and advise that any episodes of combination treatment are time-limited with patients carefully monitored.^18–20 However, given the limited number of mechanistically distinct pharmacological treatment options for schizophrenia, APP is frequently used internationally,^13,21,22 nationally,²³ and locally.²⁴ Polypharmacy is controversial not only because of the lack of evidence for enhanced treatment efficacy compared with monotherapy, but also for the potential for increased risk of side effects, perhaps through increased likelihood of exposure to higher cumulative antipsychotic doses or arising from pharmacodynamic interactions between specific drug combinations. There is cross-sectional evidence that APP is associated with an increased risk of metabolic disorder^14,21,25 in secondary care. However, although most patients with well-controlled schizophrenia are managed in primary care, there is little evidence about the side effects of polypharmacy in this setting.

This study used the UK Clinical Practice Research Datalink (CPRD) to investigate the association between exposure to APP and risk of developing diabetes, hypertension, or hyperlipidemia for schizophrenia patients in primary care. As a secondary aim, the study compared the impact of polypharmacy based on the generation of co-prescribed drugs, ie, exclusively first-generation drug combinations vs. mixed first- and second-generation and exclusively second-generation drug combinations.

Methods

Data Sources

Primary care data were obtained from the UK CPRD GOLD database of pseudonymized electronic patient records from over 700 general practices in the United Kingdom.²⁶ Records contain symptoms or diagnoses coded using the READ classification system, clinical observations, immunizations, prescriptions, and demographic details to August 30, 2018. CPRD records were linked to hospital episode statistics records, provided by NHS Digital, which contained information on all hospital admissions and causes of each episode of inpatient care coded using ICD-10 classification system to the December 31, 2017.²⁷ These were linked to Mental Health Services Dataset (MHSDS) records, also provided by NHS Digital, containing information on psychiatric admissions and causes of each episode of care coded using ICD-10 classification system to the November 30, 2015,²⁷ and patient level index of multiple deprivations (IMD) scores at Lower layer super output area in 2010.²⁷

Study Population

We conducted a cohort study using CPRD records for patients registered with participating practices with an ICD-10 code and/or a READ code for schizophrenia between July 1, 1994 and August 30, 2018. CPRD data only include primary care records, and therefore, exclude more severely ill patients treated in secondary and tertiary care. The long study period was chosen to ensure we had adequate power for planned analyses and to account for changes over time in prescribing behaviors and patterns. Cohort entry was the date of the first prescription of an antipsychotic medication within 6 months of an incident diagnosis of schizophrenia occurring on either side of the prescription date. Patients were followed up for up to 5 years in the study time period, though most (due to reaching the outcome or censoring) had no more than 2 years of follow-up, with a median follow-up of 14 months. We only included patients whose records were acceptable for research purposes (a data quality indicator provided by CPRD) and with at least 2 complete years of CPRD up-to-standard clinical data prior to the date of cohort entry.

Patients were followed up until they had: (1) a prescription and/or diagnostic code for diabetes, hypertension, or hyperlipidemia, (2) died, (3) transferred out of the GP surgery, (4) GP surgery no longer deemed up-to-standard, (5) last recorded contact with the GP practice, (6) no new prescription for an antipsychotic medication recorded in the data once the previous prescription has run out, and (7) achieved the maximum follow-up period (5 years since cohort entry).

Patients who had been prescribed an antipsychotic drug more than 6 months prior to cohort entry within the up-to-standard period were excluded from the study to ensure that all prescriptions were new prescriptions related to the schizophrenia diagnosis. We excluded patients aged under 18 at cohort entry and those with history of stroke or myocardial infarction before cohort entry. We also limited our cohort to patients eligible for linkage to either hospital episode statistics or MHSDS.

We formed 3 cohorts from the study population, one for each study outcome: Patients who did not have an ICD-10 or READ code for or had not received a previous prescription for a medication for diabetes (cohort 1), hypertension (cohort 2), or hyperlipidemia (cohort 3) more than 6 months prior to inclusion into the study.Figure 1 presents the cohort flow diagram. The 3 cohorts were analyzed separately because each outcome has well-established treatment and prevention strategies, which generally target each individually, increasing the clinical utility of this study. All READ/ICD-10/BNF codes used to define study cohorts can be found on GitHub (https://github.com/eyles-ec/polypharmacy-codes).¹

Fig. 1.

Clinical Practice Research Datalink cohort flow diagram.

Exposure

The exposure of interest was time in days exposed to antipsychotic polypharmacy vs. antipsychotic monotherapy. We also examined the impact of the generations of drug combinations when exposed to APP which were either (1) combinations of first-generation antipsychotics only, (2) combinations of first- and second-generation antipsychotic drugs, or (3) combinations of second-generation drugs only.

For each patient, we used the CPRD database to identify all prescriptions for antipsychotic drugs that were filled between the date of cohort entry and the end of follow‐up. The continuous periods of exposure to each antipsychotic drug substance were derived from the date and the duration of each prescription. The duration of each prescription was used as a proxy for duration of medication use and was determined from the quantity prescribed in milligrams divided by daily dose in milligrams. We added a 90-day washout period after the end of each prescription duration to allow for the time taken for the drug to clear from the patients’ system. Patients were considered to be exposed to the impact of a prescribed drug throughout the washout period. Overlapping days between prescriptions of the same drug were ignored unless prescribed on the same day, in which case the overlapping days were added. The dosage for all prescriptions in each continuous period of exposure was summed, then averaged out over the days in that continuous period of time. This includes all prescription types, including depot medications given every 2–4 weeks. We identified continuous periods of exposure to each antipsychotic compound during the follow‐up and classified overlapping periods of antipsychotics as polypharmacy.

Outcomes

Our outcomes were time to incidence of diabetes, hypertension, or hyperlipidemia defined as having a prescription of an antihypertensive, lipid-lowering or anti-diabetic drug and/or ICD-10 or READ code for diabetes, hypertension, or hyperlipidemia.¹

Baseline Characteristics

Several covariates were controlled for in the analysis, and included: (1) gender, (2) age, (3) IMD score, (4) ethnicity (patient-reported), (5) smoking status (never/nonsmoker, current or ex-smoker) recorded in the 5 years prior to cohort entry, (6) alcohol drinking status (“problem user” or not) recorded in the 12 months prior to cohort entry, (7) illicit drug use (“user” or not) recorded in the 12 months prior to cohort entry, (8) body mass index (BMI) recorded in the 5 years prior to cohort entry, (9) GP practice region, (10) year of schizophrenia diagnosis, to account for changes in prescribing patterns over time, (11) number of comorbidities including diagnoses of anorexia/bulimia or depression/anxiety recorded in the 12 months prior to cohort entry or diagnoses of dementia, Parkinson’s disease, chronic liver disease, learning difficulties, diabetes, hypertension, or hyperlipidemia recorded at any time prior to cohort entry, (12) number of consultations, hospitalizations, and mental health service encounters (proxy variable for likelihood of engaging with primary and secondary care) recorded in the 12 months prior to cohort entry, and (13) prescribed daily dose (PDD) of antipsychotics/ defined daily dose (DDD) for each drug taken, ie, PDD/DDD ratio. The DDD of antipsychotic medication (inversely related to drug potency)²⁸ is the average maintenance dose per day for a drug used for its main indication in adults. The actual dose taken by the patient was used as the PDD. If a patient took multiple regimens within the same drug, the PDD was averaged over the days it was taken for that specific continuous exposure period, including depot medications. Time-varying PDD/DDD ratio was calculated for each drug and summed for all drugs taken within an antipsychotic medication class to measure the drug load. Supplementary material 1 presents the specific antipsychotic drugs prescribed, as well as additional psychotropic prescriptions, per cohort.

Statistical Analysis

We used an event-time stratified multivariable Cox proportional hazards model with time-varying exposures on daily data to determine the effect of antipsychotic polypharmacy and prescribed drug generation on the incidence of diabetes, hypertension, and hyperlipidemia, following Xu et al.²⁹ We calculated stabilizing weights as described in Xu et al²⁷ for every unique event time of follow-up period for each individual, including the time-varying exposures and covariates. We then fitted the stratified Cox model with stabilizing weights as the adjustment to obtain hazard ratios of risk comparing those on monotherapy vs those on polypharmacy. This type of model accounts for changes in risk due to the start or end of a particular treatment, in this case, antipsychotic mono- or polypharmacy. Data were stratified by unique event times, ie, days of polypharmacy exposure.

Missing data on baseline covariates of BMI, IMD score, and smoking status were imputed using multiple chained equations under the missing at-random assumption. Imputation models were derived for each missing variable (BMI, IMD score, and smoking status) and included as covariates: the exposures of interest, outcome (hypertension, hyperlipidemia, or diabetes), and all other variables with or without missing data. A total of 20 complete datasets were constructed to reduce sampling variability from the imputation process³⁰ and the results were combined using Rubin’s Rules.^31,32

We conducted multiple sensitivity analyses to investigate the impact of alternative definitions of exposures: Not including dose in the model; not using the stabilizing weights approach; and the impact of potential interaction between exposure categories and actual duration of polypharmacy over the follow-up using Poisson regression models.

The Cox model makes several assumptions (proportional hazards; linear relationship between hazard function and covariates) that were assessed prior to analysis and were found to be plausible in this setting.

Stata versions 16 and 17³³ were used for data management and analyses.

Ethics

The study protocol was reviewed and approved by the Independent Scientific Advisory Committee for Medicines and Healthcare Products Regulatory Agency (MHRA) database research (protocol number 17_263). The committee waived the need for patient informed consent because data used in this study were de-identified.

Results

The study cohorts comprised of 1663, 1268, and 1668 patients with schizophrenia to be followed for diabetes, hypertension, and hyperlipidemia outcomes, respectively. During follow-up, 4.4% (n = 73) of the first cohort developed diabetes, 6.6% (n = 84) of the second cohort developed hypertension and 3.8% (n = 63) of the third cohort developed hyperlipidemia. The follow-up time in this study was up to a maximum of 5 years, with a median time of approximately 14 months or 418 days (interquartile range [IQR], 152–1153 days) in the first cohort, 421 days (IQR, 151–1129 days) in the second cohort and 420 days (IQR, 150–1153 days) in the third cohort. For the primary analysis, the total analytic duration of the first and second cohort were 958 days, and for the third was 630 days, due to data requirements in calculating the stabilizing weights.²⁹ Participants who were diagnosed with the metabolic outcomes were markedly older at entry in all 3 cohorts and had more medical comorbidities. For both hypertension and hyperlipidemia, participants who were diagnosed with these outcomes were exposed to greater cumulative doses of antipsychotic drugs. Table 1 reports baseline characteristics of each cohort prior to imputation, and table 2 reports baseline characteristics post-imputation. Supplementary table 2 reports the complete case vs. imputed data: It is important to note here that patients who previously were not included due to incomplete case data, because of missing data for BMI, IMD score, and/or smoking status, were included in the imputed dataset. Only BMI, IMD score, and smoking status were imputed.

Table 1.

Characteristics of the Study Cohorts by Outcome at Baseline, Pre-imputation

	Cohort 1 (n = 1663)2			Cohort 2 (n = 1268)2			Cohort 3 (n = 1668)2
Characteristics at Cohort Entry 1	No Diabetes	Diabetes	P	No Hyper- Tension	Hyper- Tension	P	No Hyper- Lipidaemia	Hyper- Lipidaemia	P
Total n	1590 (95.6%)	73 (4.4%)		1184 (93.4%)	84 (6.6%)		1605 (96.2%)	63 (3.8%)
Follow-up duration, days	681.8 ± 631.7	576.1 ± 525.2	.159	686.9 ± 620.4	509.0 ± 496.2	.010	672.3 ± 625.4	692.0 ± 498.1	.805
Antipsychotics prescription pattern, days per patient3			.895			.070			.142
Monopharmacy days	557.9 ± 540.0	497.9 ± 467.2	.354	558.9 ± 528.7	437.8 ± 446.9	.048	548.2 ± 535.4	561.3 ± 436.1	.852
Polypharmacy	175.9 ± 239.1	99.8 ± 79.3	.191	179.2 ± 245.33	170.5 ± 199.9	.870	173.3 ± 229.9	195.4 ± 215.9	.667
Antipsychotics generation prescriptions, for polypharmacy days only, days per patient3
1st generation only days	233.5 ± 407.4	75.0 ± 97.6	.592	145.6 ± 243.6	57.0 ± 48.1	.535	162.1 ± 281.3	276.0 ± 313.4	.510
1st and 2nd generation days	168.8 ± 214.0	103.1 ± 80.0	.237	188.7 ± 246.7	200 ± 218.3	.844	172.3 ± 221.8	182.0 ± 204.8	.856
2nd generation only days	423.3 ± 540.9	—	—	61.0 ± 19.9	82.5 ± 45.9	.505	423.3 ± 540.9	—	—
Age, years	46.8 ± 21.9	56.7 ± 21.1	<.001	39.7 ± 17.6	47.9 ± 20.3	<.001	44.9 ± 21.4	53.6 ± 14.7	.001
Defined daily dose (DDD)	121.50 ± 198.1	89.85 ± 156.7	.346	118.3 ± 195.9	82.1 ± 157.7	.000	120.8 ± 198.0	115.1 ± 177.0	.000
Prescribed daily dose (PDD)	60.51 ± 137.6	64.98 ± 174.9	.013	68.9 ± 153.6	40.8 ± 113.9	.000	62.2 ± 140.9	90.7 ± 187.4	.000
Daily drug load (PDD/DDD)	0.62 ± 0.53	0.67 ± 0.54	.113	0.66 ± 0.50	0.66 ± 0.70	.856	0.61 ± 0.53	1.00 ± 0.78	.000
Body mass index (BMI)			.120			.008			<.001
Underweight	61 (3.8%)	4 (5.5%)		41 (3.5%)	4 (4.8%)		66 (4.1%)	1 (1.6%)
Healthy	434 (27.3%)	14 (19.2%)		332 (28.0%)	13 (15.5%)		441 (27.5%)	9 (14.3%)
Overweight	229 (14.4%)	43 (58.9%)		141 (11.9%)	16 (19.0%)		208 (13.0%)	20 (31.7%)
Obese	133 (8.4%)	11 (15.1%)		90 (7.6%)	12 (14.3%)		139 (8.7%)	5 (7.9%)
(Missing)	733 (46.1%)	31 (42.5%)		580 (49.0%)	39 (46.4%)		751 (46.8%)	28 (44.4%)
Ethnicity			.094			.882			.783
White	1306 (86.8%)	57 (79.1%)		924 (83.5%)	65 (82.3%)		1294 (85.1%)	51 (83.6%)
Asian	60 (4.0%)	5 (6.9%)		60 (5.4%)	4 (5.1%)		71 (4.7%)	2 (3.3%)
Black	60 (4.0%)	6 (8.3%)		55 (5.0%)	5 (6.3%)		69 (4.5%)	5 (8.2%)
Mixed	15 (1.0%)	2 (2.8%)		13 (1.2%)	2 (2.5%)		15 (1.0%)	1 (1.6%)
Other	25 (1.7%)	2 (2.8%)		29 (2.6%)	2 (2.5%)		31 (2.0%)	1 (1.6%)
Unknown	39 (2.6%)	0 (0.0%)		26 (2.4%)	1 (1.3%)		40 (2.6%)	1 (1.6%)
GP practice region			.458			.449			.771
North East	40 (2.5%)	0 (0.0%)		25 (2.1%)	3 (3.6%)		36 (2.2%)	2 (3.2%)
North West	294 (18.5%)	13 (17.8%)		215 (18.2%)	21 (25.0%)		293 (18.3%)	14 (22.2%)
Yorkshire and the humber	57 (3.6%)	2 (2.7%)		38 (3.2%)	3 (3.6%)		53 (3.3%)	3 (4.8%)
East Midlands	36 (2.3%)	2 (2.7%)		31 (2.6%)	0 (0.0%)		39 (2.4%)	1 (1.6%)
West Midlands	160 (10.1%)	13 (17.8%)		116 (9.8%)	12 (14.3%)		169 (10.5%)	6 (9.5%)
East of England	160 (10.1%)	7 (9.6%)		123 (10.4%)	7 (8.3%)		160 (10.0%)	4 (6.3%)
South West	211 (13.3%)	11 (15.1%)		155 (13.1%)	10 (11.9%)		220 (13.7%)	5 (7.9%)
South Central	219 (13.8%)	6 (8.2%)		158 (13.3%)	7 (8.3%)		217 (13.5%)	7 (11.1%)
London	237 (14.9%)	13 (17.8%)		196 (16.6%)	14 (16.7%)		249 (15.5%)	14 (22.2%)
South East Coast	176 (11.1%)	6 (8.2%)		127 (10.7%)	7 (8.3%)		169 (10.5%)	7 (11.1%)
Index of Multiple Deprivation (IMD) 2015			.667			.236			.001
1 (Least deprived)	249 (15.7%)	13 (17.8%)		181 (15.3%)	14 (16.7%)		255 (15.9%)	4 (6.3%)
2	286 (18.0%)	12 (16.4%)		195 (16.5%)	14 (16.7%)		281 (17.5%)	7 (11.1%)
3	285 (17.9%)	17 (23.3%)		216 (18.2%)	12 (14.3%)		288 (17.9%)	13 (20.6%)
4	374 (23.5%)	13 (17.8%)		285 (24.1%)	14 (16.7%)		382 (23.8%)	10 (15.9%)
5 (Most deprived)	393 (24.7%)	18 (24.7%)		305 (25.8%)	30 (35.7%)		396 (24.7%)	29 (46.0%)
(Missing)	3 (0.2%)	0 (0.0%)		2 (0.2%)	0 (0.0%)		3 (0.2%)	0 (0.0%)
Sex			.154			.070			.505
Female	693 (43.6%)	38 (52.1%)		446 (37.7%)	40 (47.6%)		705 (43.9%)	25 (39.7%)
Male	897 (56.4%)	35 (47.9%)		738 (62.3%)	44 (52.4%)		900 (56.1%)	38 (60.3%)
Alcohol problem user			.075			.512			.331
No	1392 (87.5)	69 (94.5%)		1043 (88.1%)	76 (90.5%)		1413 (88.0%)	58 (92.1%)
Yes	198 (12.5%)	4 (5.5%)		141 (11.9%)	8 (9.5%)		192 (12.0%)	5 (7.9%)
Illicit drug user			.415			.572			.438
No	1542 (97.0%)	72 (98.6%)		1142 (96.5%)	82 (97.6%)		1551 (96.6%)	62 (98.4%)
Yes	48 (3.0%)	1 (1.4%)		42 (3.5%)	2 (2.4%)		54 (3.4%)	1 (1.6%)
Smoking status			.765			.499			.662
Never/non smoker	520 (32.7%)	28 (38.4%)		366 (30.9%)	27 (32.1%)		519 (32.3%)	20 (31.7%)
Ex-smoker	184 (11.6%)	10 (13.7%)		118 (10.0%)	5 (6.0%)		182 (11.3%)	10 (15.9%)
Current smoker	498 (31.3%)	22 (30.1%)		409 (34.5%)	25 (29.8%)		504 (31.4%)	21 (33.3%)
(Missing)	388 (24.4%)	13 (17.8%)		291 (24.6%)	27 (32.1%)		400 (24.9%)	12 (19.0%)
Number of comorbidities 4	0.8 ± 0.9	1.2 ± 0.8	<.001	0.6 ± 0.8	0.9 ± 0.9	<.001	0.9 ± 1.0	1.2 ± 1.1	.011
Healthcare use encounters 5	27.2 ± 20.4	39.2 ± 30.0	<.001	23.1 ± 17.3	23.5 ± 18.6	.985	26.8 ± 20.8	24.4 ± 17.5	.440

¹Characteristics are presented as mean ± SD or n (%).

² P-values are generated by chi-squared test for categorical, t-test for continuous and Mann–Whitney U test for count variables: The P-values represent a comparison between the outcome groups.

³Mean days of monotherapy and polypharmacy were calculated as means of the total count of day type per patient, excluding any days of nonadherence.

⁴Comorbidities include anorexia/bulimia, depression/anxiety, dementia, Parkinson’s disease, chronic liver disease, learning difficulties, diabetes, hypertension, and hyperlipidemia.

⁵Healthcare use encounters includes count of GP consultations, hospitalizations, and mental health service encounters within the 12 months prior to cohort entry.

Table 2.

Characteristics of the Study Cohorts by Outcome at Baseline, Imputed Data^*

	Cohort 1 (n = 1663)2			Cohort 2 (n = 1268)2			Cohort 3 (n = 1668)2
Characteristics at Cohort Entry 1	No Diabetes	Diabetes	P	No Hyper-Tension	Hyper-Tension	P	No Hyper- Lipidaemia	Hyper- Lipidaemia	P
Total n	1590 (95.6%)	73 (4.4%)		1184 (93.4%)	84 (6.6%)		1605 (96.2%)	63 (3.8%)
Follow-up duration, days	681.8 ± 631.7	576.1 ± 525.2	.159	686.9 ± 620.4	509.0 ± 496.2	.010	672.3 ± 625.4	692.0 ± 498.1	.805
Antipsychotics prescription pattern, days per patient3			.895			.070			.142
Monopharmacy days	557.9 ± 540.0	497.9 ± 467.2	.354	558.9 ± 528.7	437.8 ± 446.9	.048	548.2 ± 535.4	561.3 ± 436.1	.852
Polypharmacy	175.9 ± 239.1	99.8 ± 79.3	.191	179.2 ± 245.33	170.5 ± 199.9	.870	173.3 ± 229.9	195.4 ± 215.9	.667
Antipsychotics generation prescriptions, for polypharmacy days only, days per patient3
First-generation only days	233.5 ± 407.4	75.0 ± 97.6	.592	145.6 ± 243.6	57.0 ± 48.1	.535	162.1 ± 281.3	276.0 ± 313.4	.510
First and second-generation days	168.8 ± 214.0	103.1 ± 80.0	.237	188.7 ± 246.7	200 ± 218.3	.844	172.3 ± 221.8	182.0 ± 204.8	.856
Second-generation only days	423.3 ± 540.9	—	—	61.0 ± 19.9	82.5 ± 45.9	.505	423.3 ± 540.9	—	—
Age, years	45.7 ± 23.1	60.4 ± 23.6	.002	39.7 ± 17.6	47.7 ± 20.8	<.001	44.9 ± 21.4	53.6 ± 14.7	.001
Defined daily dose (DDD)	119.7 ± 189.0	121.7 ± 186.3	.927	118.3 ± 195.9	82.1 ± 157.7	.000	120.8 ± 198.0	115.1 ± 177.0	.000
Prescribed daily dose (PDD)	56.2 ± 265.9	131.2 ± 489.4	.025	68.9 ± 153.6	40.8 ± 113.9	.000	62.2 ± 140.9	90.7 ± 187.4	.000
Daily drug load (PDD/DDD)	0.59 ± 1.34	0.84 ± 1.26	.115	0.66 ± 0.50	0.66 ± 0.70	.856	0.61 ± 0.53	1.00 ± 0.78	.000
Body mass index (BMI)			.809			.393			.005
Underweight	185 (11.6%)	9 (12.3%)		124 (10.5%)	12 (14.3%)		226 (14.1%)	3 (4.8%)
Healthy	691 (43.5%)	28 (38.4%)		527 (44.5%)	30 (35.7%)		717 (44.7%)	21 (33.3%)
Overweight	443 (27.9%)	21 (28.8%)		331 (28.0%)	25 (29.8%)		416 (25.9%)	27 (42.9%)
Obese	271 (17.0%)	15 (20.6%)		202 (17.1%)	17 (20.2%)		246 (15.3%)	12 (19.1%)
Ethnicity4			.004			.774			.577
White	1306 (82.1%)	57 (78.1%)		924 (78.0%)	65 (77.4%)		1294 (80.6%)	51 (81.0%)
Black, Asian, mixed, other	160 (10.1%)	15 (20.6%)		157 (13.3%)	13 (15.5%)		186 (11.6%)	9 (14.3%)
Unknown	124 (7.8%)	1 (1.4%)		103 (8.7%)	6 (7.1%)		125 (7.8%)	3 (4.8%)
GP practice region			.458			.449			.771
North East	40 (2.5%)	0 (0.0%)		25 (2.1%)	3 (3.6%)		36 (2.2%)	2 (3.2%)
North West	294 (18.5%)	13 (17.8%)		215 (18.2%)	21 (25.0%)		293 (18.3%)	14 (22.2%)
Yorkshire and the humber	57 (3.6%)	2 (2.7%)		38 (3.2%)	3 (3.6%)		53 (3.3%)	3 (4.8%)
East Midlands	36 (2.3%)	2 (2.7%)		31 (2.6%)	0 (0.0%)		39 (2.4%)	1 (1.6%)
West Midlands	160 (10.1%)	13 (17.8%)		116 (9.8%)	12 (14.3%)		169 (10.5%)	6 (9.5%)
East of England	160 (10.1%)	7 (9.6%)		123 (10.4%)	7 (8.3%)		160 (10.0%)	4 (6.3%)
South West	211 (13.3%)	11 (15.1%)		155 (13.1%)	10 (11.9%)		220 (13.7%)	5 (7.9%)
South Central	219 (13.8%)	6 (8.2%)		158 (13.3%)	7 (8.3%)		217 (13.5%)	7 (11.1%)
London	237 (14.9%)	13 (17.8%)		196 (16.6%)	14 (16.7%)		249 (15.5%)	14 (22.2%)
South East Coast	176 (11.1%)	6 (8.2%)		127 (10.7%)	7 (8.3%)		169 (10.5%)	7 (11.1%)
Index of Multiple Deprivation (IMD) 2015			.671			.237			.002
1 (Least deprived)	249 (15.7%)	13 (17.8%)		182 (15.4%)	14 (16.7%)		255 (15.9%)	4 (6.3%)
2	286 (18.0%)	12 (16.4%)		196 (16.5%)	14 (16.7%)		281 (17.5%)	7 (11.1%)
3	286 (18.0%)	17 (23.3%)		216 (18.2%)	12 (14.3%)		289 (18.0%)	13 (20.6%)
4	374 (23.5%)	13 (17.8%)		285 (24.1%)	14 (16.7%)		382 (23.8%)	10 (15.9%)
5 (Most deprived)	395 (24.8%)	18 (24.7%)		305 (25.8%)	30 (35.7%)		389 (24.8%)	29 (46.0%)
Sex			.154			.070			.505
Female	693 (43.6%)	38 (52.1%)		446 (37.7%)	40 (47.6%)		705 (43.9%)	25 (39.7%)
Male	897 (56.4%)	35 (47.9%)		738 (62.3%)	44 (52.4%)		900 (56.1%)	38 (60.3%)
Alcohol problem user			.075			.512			.331
No	1392 (87.5)	69 (94.5%)		1043 (88.1%)	76 (90.5%)		1413 (88.0%)	58 (92.1%)
Yes	198 (12.5%)	4 (5.5%)		141 (11.9%)	8 (9.5%)		192 (12.0%)	5 (7.9%)
Illicit drug user			.415			.572			.438
No	1542 (97.0%)	72 (98.6%)		1142 (96.5%)	82 (97.6%)		1551 (96.6%)	62 (98.4%)
Yes	48 (3.0%)	1 (1.4%)		42 (3.5%)	2 (2.4%)		54 (3.4%)	1 (1.6%)
Smoking status			.680			.111			.720
Never/non smoker	683 (43.0%)	35 (48.0%)		482 (40.7%)	44 (52.4%)		710 (44.2%)	25 (39.7%)
Ex-smoker	662 (41.6%)	27 (37.0%)		542 (45.8%)	31 (36.9%)		660 (41.1%)	27 (42.9%)
Current smoker	245 (15.4%)	11 (15.1%)		160 (13.5%)	9 (10.7%)		235 (14.6%)	11 (17.5%)
Number of comorbidities 5	0.8 ± 0.9	1.0 ± 0.8	.071	0.5 ± 0.8	0.7 ± 0.8	.039	0.8 ± 1.0	0.9 ± 1.1	.533
Healthcare use encounters 6	27.2 ± 20.4	39.2 ± 30.0	<.001	23.1 ± 17.3	23.5 ± 18.6	.845	26.8 ± 20.8	24.4 ± 17.5	.369

^*The variables that were imputed were BMI, Smoking, and index of multiple deprivation.

¹Characteristics are presented as mean ± SD or n (%).

² P-values are generated by chi-squared test for categorical, t-test for continuous and Mann–Whitney U test for count variables: The P-values represent a comparison between the outcome groups.

³Mean days of monotherapy and polypharmacy were calculated as means of the total count of day type per patient, excluding any days of nonadherence.

⁴Ethnicity was collapsed into three categories at imputation, due to lack of data in smaller categories.

⁵Comorbidities include anorexia/bulimia, depression/anxiety, dementia, Parkinson’s disease, chronic liver disease, learning difficulties, diabetes, hypertension, and hyperlipidemia.

⁶ Healthcare use encounters includes count of GP consultations, hospitalizations, and mental health service encounters within the 12 months prior to cohort entry.

Table 3 shows the unadjusted and adjusted hazard ratios from the Cox models, as well as the baseline cumulative hazard. Relative to exposure to antipsychotic monotherapy, there was no evidence in the adjusted models that exposure to antipsychotic polypharmacy was associated with increased risk of developing diabetes or hyperlipidemia. There was evidence of an association between antipsychotic polypharmacy and increased risk of developing hypertension (unadjusted HR 2.35, 95% CI 1.19 to 4.64, P = .014; adjusted HR 3.16, 95% CI 1.19 to 8.38 P = .021).

Table 3.

Unadjusted and Adjusted Hazard Ratios for Developing an Outcome Among Schizophrenia Patients by Antipsychotic Prescription Pattern

	Cohort 1 (n = 1663)				Cohort 2 (n = 1268)				Cohort 3 (n = 1668)
	No Diabetes vs. Diabetes				No Hypertension vs. Hypertension				No Hyperlipidemia vs. Hyper-lipidaemia
	Baseline Cumulative Hazard2	HR	[95%CI]	P-value	Baseline Cumulative Hazard2	HR	[95%CI]	P-value	Baseline Cumulative Hazard2	HR	[95%CI]	P-value
Antipsychotics prescription patterns	0.029				0.036				.016
Unadjusted estimates
Monotherapy		1.00				1.00				1.00
Polypharmacy		1.38	[0.55,3.49]	.491		2.35	[1.19, 4.64]	0.014		0.40	[0.05, 2.91]	.362
Adjusted estimates1
Monotherapy		1.00				1.00				1.00
Polypharmacy		1.32	[0.87, 2.01]	.190		3.16	[1.19, 8.38]	0.021		0.46	[0.04, 6.11]	.560
Antipsychotics generation prescriptions only for polypharmacy days	0.029				0.052				.015
Unadjusted estimates
1st generation only (reference)		1.00				1.00				1.00
1st and 2nd generation		0.38	[0.17, 0.86]	.020		0.44	[0.22, 0.90]	0.024		0.36	[0.13, 1.01]	.052
2nd generation only		1.60	[0.90, 2.83]	.110		1.18	[0.68, 2.06]	0.559		1.37	[0.64, 2.92]	.413
Adjusted estimates1
1st generation only (reference)		1.00				1.00				1.00
1st and 2nd generation		0.36	[0.12, 1.15]	.085		0.29	[0.09, 0.94]	0.039		0.69	[0.22, 2.12]	.514
2nd generation only		1.41	[0.65, 3.03]	.392		1.15	[0.51, 2.58]	0.737		1.16	[0.40, 3.35]	.590

¹Adjusted to baseline covariates of age, sex, alcohol problem use, illicit drug use, smoking status, ethnicity, GP practice region, index of multiple deprivation score, BMI, year of schizophrenia diagnosis, no of comorbidities, no of health service encounters, defined daily dose.

²Calculated as the mean cumulative baseline hazard over all imputations.

Secondary analyses comparing the effect of combining the different generations of antipsychotic medications involved in polypharmacy showed inconsistent results between the different metabolic outcomes. For hypertension, the only outcome associated with polypharmacy in the primary analyses, there was evidence for a reduced risk associated with combined first- and second-generation polypharmacy compared to exclusive first-generation polypharmacy which remained after adjustment (unadjusted HR 0.44, 95% CI 0.22 to 0.90, P = .024, adjusted HR 0.29, 95% CI 0.09 to 0.94, P = .039). Regarding hyperlipidemia, there was a trend towards a reduced risk of hyperlipidemia in patients with combined first- and second-generation polypharmacy compared to patients with exclusive first-generation polypharmacy in the unadjusted analysis (HR 0.36, 95% CI 0.13 to 1.01, P = .052) which was no longer evident after adjustment (HR 0.69, 95% CI 0.22 to 2.12, P = .514). For diabetes, unadjusted models were suggestive of a reduced risk of diabetes associated with combined first- and second-generation polypharmacy compared to exclusive first-generation polypharmacy, but this disappeared after adjustment (unadjusted HR 0.38, 95% CI 0.17 to 0.86, P = .020, adjusted HR 0.36, 95% CI 0.12 to 1.15, P = .085). There was no evidence that exclusive second-generation antipsychotic polypharmacy altered the risk of diabetes, hypertension, or hyperlipidemia, relative to treatment exclusively with first-generation antipsychotics, but exposure to exclusively second-generation drug combinations was limited in our sample.

Discussion

We examined the prospective association between APP and development of diabetes, hypertension, and hyperlipidemia for up to 5 years, in patients with schizophrenia managed in UK primary care.

Patients with a treatment history of APP had a higher risk of developing hypertension, compared to patients exclusively exposed to monotherapy. This suggests that the risks of developing hypertension and its cardiovascular consequences over time need to be considered when making decisions on whether APP is justified. Where APP is judged to be necessary, regular blood pressure monitoring, and where necessary, instigation of antihypertensive treatment, is important to reduce cardiovascular risk. Mitigation of other factors associated with hypertension should be considered.

No evidence of an association between exposure to APP and either diabetes or hyperlipidemia was found, either before or after adjustment. While clinicians should remain vigilant for these outcomes when co-prescribing two or more antipsychotics, unlike with hypertension, exposure to APP does not appear to elevate risk relative to monotherapy.

All previous studies investigating APP and metabolic disorders were conducted in a secondary care setting; a small number of studies suggested a protective effect. Correll et al²⁵ described a univariate association between polypharmacy and elevated rates of metabolic syndrome; however, in multivariable analyses, found that this association was unlikely to be independent of risk factors for metabolic syndrome. Our longitudinal study has found an effect on hypertension even when risk factors are included. This is also in contrast to the findings of Galling et al,³⁴ who, in a systematic review, found the effect of APP on hypertension was the same as monotherapy.

Plausible biological mechanisms for deleterious effects of APP irrespective of total dose might involve receptor and neurotransmitter systems beyond the dopamine system (which is related to the dose variable for which we adjusted) that some or all antipsychotic drugs are known to modulate. These might include interactions involving affinities for serotonergic, alpha-adrenergic, histaminergic, or cholinergic and other receptor systems.^10,35 For example, many antipsychotic drugs are antagonists of one or more serotonin receptors, while serotonergic neurons in key brainstem areas are known to be implicated in blood pressure control.³⁶ The finding in our secondary analysis that the risk of developing hypertension was markedly greater with polypharmacy based exclusively on first-generation antipsychotics than with combinations of first- and second-generation drugs, which held both before and after adjustment is of particular interest. While second-generation antipsychotics are known to increase the risk of various adverse metabolic outcomes,^7,37a direct comparison between first and second-generation antipsychotics suggested that the risk of hypertension is higher in the former.³⁸

Strengths and Limitations

The key strength of our study is the use of a prospectively collected population-based database of primary and secondary care data deemed to be representative of the age, sex, and geographic regions of the United Kingdom.³⁹ This reduced the risk of selection bias. CPRD includes detailed medical information, symptoms, and signs in a well-defined, representative, and stable population, and the completeness and accuracy of the clinical records have been validated externally.³⁹ We have also employed time-varying exposures in the Cox models, meaning that treatment with antipsychotics is allowed to vary through time, in our case exposure days, which allows for less biased estimates of treatment effects.

We have attempted to include all known confounding variables; however, some residual confounding may still influence our results. Confounding by indication, the possibility that people who experience APP have a more severe or difficult-to-treat form of schizophrenia and thereby are more unwell, may contribute to the results. It is feasible that treatment-resistant schizophrenia is associated both with a higher likelihood of polypharmacy, and to increased incidence of a physical/cardiovascular co-morbidity such as hypertension. However, this issue is mitigated to an extent by the large number of covariates we were able to include in our adjusted models, such as cardiovascular risk factors and health service utilization. The ability to adjust for total antipsychotic drug dose ascertained from reliable prescription data registers is a particular strength in eliminating the effects of total drug dose exposure, which is inevitably higher across individuals prescribed two drugs concurrently rather than a single antipsychotic agent.

Another important limitation is that patients with unstable schizophrenia are managed in secondary/tertiary care, and prescription data on these patients was not available in this study. It is possible that the associations between drug treatment and metabolic illness in these patients could be different from the associations observed in primary-care-managed patients. Furthermore, because this is an observational study, we are unable to establish a causal link between APP and metabolic disorders.

There might be some misclassification of the exposure through prescriptions not being captured, or patients filling prescriptions that they do not use, which would result in under- or overestimation of actual duration of use. However, if repeated prescriptions are present, it is reasonable to assume that total prescription duration is similar to the total duration of use.

Furthermore, while antipsychotic prescribing trends have changed over time, and advances in the risk of, prevention of and management of hypertension, diabetes, and hyperlipidemia may have impacted on the results. To mitigate this, we have included the year of schizophrenia diagnosis in the analysis. Furthermore, the 2 most important innovations in pharmacotherapy (the increasing use of second-generation antipsychotics relative to first-generation, and the introduction of statins) both occurred towards the beginning of the time span of the study which also mitigates the impact of the time period of entry.

Median follow-up time in this study was around 14 months. Follow-up was limited for various reasons, including patients reaching the metabolic outcome in the cohort, ceasing to be prescribed antipsychotic medications (or having their prescriptions provided elsewhere such as in secondary or tertiary care), transferring out of their GP surgery, losing contact with the GP practice or by death. Although no associations were detected for polypharmacy with hyperlipidemia and diabetes, it is possible that these might have emerged over a longer period. It is also plausible that hypertension yielded a positive result because it is typically detected and treated more rapidly than the other 2 outcomes, blood pressure checks being easier and cheaper than ascertaining lipid and diabetes parameters. If this were the case, blood pressure monitoring might be seen as an early indicator of future cardio-metabolic issues including diabetes and hyperlipidemia.

There are small numbers of exposure to solely second-generation antipsychotic combinations in our sample, which limits our ability to draw conclusions about their impact on metabolic disorders.

A further limitation is the potential for misclassification of the hypertension outcome through identification using drugs, which have utility both in mental health disorders related to schizophrenia, and as antihypertensives. The most important example is the use of specific beta-blockers which, while bonafide hypertension treatments, may also be prescribed for anxiety disorders (eg, propranolol) or depression (eg, pindolol). Some participants may be deemed to have reached the hypertension endpoint due to these drug prescriptions when in fact they were prescribed for psychiatric symptoms co-morbid with schizophrenia.

Conclusions

We found that, relative to exposure to monotherapy, polypharmacy was associated with an increased risk of developing hypertension, but there is no evidence of an increased risk of diabetes or hyperlipidemia. Compared to polypharmacy exclusively with first-generation antipsychotics, we found evidence of a lower risk for hypertension for those taking first- and second-generation antipsychotics in combination, compared to those treated with solely first-generation drugs, but no evidence of a change in risk for hyperlipidemia or diabetes.

Future research employing larger sample sizes, longer follow-up, and methodologies capable of overcoming limitations such as confounding by indication would be valuable to confirm the associations found in this study. Furthermore, the finding that treatment exclusively with first-generation antipsychotic combinations was associated with an excess risk of hypertension is worth further investigation.

The clinical implication of our study is that there may be an excess risk of hypertension when employing antipsychotic polypharmacy relative to monotherapy which is not explained fully by polypharmacy being associated with higher cumulative antipsychotic dose. Prescribers should therefore be especially vigilant for changes in blood pressure and proactive in managing cardiovascular risk when antipsychotic polypharmacy is the chosen treatment for schizophrenia.

Funding

This research was funded by the National Institute for Health Research Applied Research Collaboration West (NIHR ARC West; NIHR200181). The views expressed in this article are those of the author(s) and not necessarily those of the NIHR or the Department of Health and Social Care. The funder had no influence on the design and conduct of the study; collection, management, analysis, and interpretation of the data; and preparation, review, or approval of the manuscript.

Author Contributors

SS, SD, BB, MTR, DK, and PM were responsible for the study concept and design and developing study protocol. EE, RM, and HE carried out statistical analyses and drafted the manuscript supervised by MTR. SD, BB, DK, and PM. MTR. SD, BB, DK, and PM advised on clinical aspects of the study and interpretation of the results. All the authors were responsible for the critical revision of the manuscript for important intellectual content. The corresponding author attests that all listed authors meet authorship criteria and that no others meeting the criteria have been omitted. All authors have approved the submitted version. MTR and SS are guarantors.

Competing Interests

All authors declare no support from any organization for the submitted work, no financial relationships with any organizations that might have an interest in the submitted work in the previous 3 years, no other relationships or activities that could appear to have influenced the submitted work.

Ethical Approval

This study protocol was reviewed and approved for CPRD by Independent Scientific Advisory Committee for Medicines and Healthcare Products Regulatory Agency (MHRA) database research (protocol number 17_263). This was secondary analysis of data submitted to the CPRD, no patient consent forms were required to access this dataset.

Data Availability

The data controller of the data analyzed are the CPRD. Patient-level data are available subject to their information governance requirements.

Transparency

The lead authors EE, RM, and HE affirm that the manuscript is an honest, accurate, and transparent account of the study being reported; that no important aspects of the study have been omitted; and that any discrepancies from the study as planned have been explained.