Prescribed opioid use is associated with adverse cardiovascular outcomes in community‐dwelling older persons

Abstract

Aims

Prescribed opioids are commonly used in the older community‐dwelling population for the treatment of chronic pain. Although the harmful effects of opioid abuse and overdose are well understood, little is known about the long‐term cardiovascular (CV) effects of prescribed opioids. The aim of this study was to investigate the CV effects associated with prescribed opioid use.

Methods and results

A post hoc analysis of participants in the Aspirin in Reducing Events in the Elderly (ASPREE) trial was conducted. Participants in the ASPREE trial included community‐dwelling older adults without a prior history of CV disease (CVD). Prescribed opioid use was defined as opioid use at baseline and/or at the first annual visit (AV1). Cox proportional hazards regression was used to calculate hazard ratios and 95% confidence intervals (95% CI) for associations between opioid use and CVD events following AV1. Of the 17 701 participants included (mean age 75.2 years, 58.2% female), 813 took opioids either at baseline or at AV1. Over a median follow‐up period of 3.58 years (IQR 2.50–4.62), CVD events, most notably heart failure hospitalization, occurred in 7% (n = 57) amongst opioid users and 4% (n = 680) amongst non‐opioid users. After adjustment for multiple covariates, opiate use was associated with a 1.67‐fold (CI 1.26–2.23, P < 0.001) increase in the hazard ratio for CVD events.

Conclusions

These findings identify opioid use as a non‐traditional risk factor for CVD events in community‐dwelling older adults.

Introduction

Cardiovascular disease (CVD), including coronary heart disease, cerebrovascular disease, and heart failure, is a leading cause of death and disability globally. ¹ Although there are several traditional risk factors for CVD, including hypertension, smoking, dyslipidaemia, obesity, and diabetes, recent research is exploring the potential impact of non‐traditional, potentially modifiable factors. ²

The respiratory and psychosocial impacts of opioid abuse and overdose are well known; however, limited studies have investigated the potential cardiovascular (CV) effects of prescribed opioid use. ³ , ⁴ , ⁵ , ⁶ , ⁷ Findings from these limited studies remain conflicting, with some suggesting opioids increase the risk of CVD events, whereas others propose a potential cardioprotective effect. ⁸ , ⁹ , ¹⁰ Furthermore, opioids are frequently used as analgesia for the acute management of CV events, such as acute myocardial infarction (MI) to relieve chest pain and discomfort. Their analgesic effects are theoretically through the reduction in blood pressure and heart rate, thereby decreasing myocardial oxygen demand. Although their impact on CV outcomes is limited, increasing evidence suggests that their use in patients presenting with acute coronary syndromes is associated with higher mortality even after risk adjustment and matching. ¹¹ , ¹² , ¹³

Despite the current uncertainty around the CV effects of chronic opioid use, the use of opioids is increasing in the geriatric community‐dwelling population worldwide. ¹⁴ In 2017, opioid prescription was highest (26.8%) amongst people aged at least 65 years in the USA. ¹⁵ This high prevalence can largely be attributed to the frequent occurrence of chronic non‐cancer pain, which ranges between 25 and 50% in this population. ¹⁶ The most common causes of chronic non‐cancer pain in this population include osteoarthritis, musculoskeletal and neuropathic pain. ¹⁷ Prescribed opioids have a high potential to lead to psychological and physical dependence, with one in four adults who take opioid prescriptions for chronic non‐cancer pain struggling with lifelong addiction. ¹⁸ , ¹⁹ In 2018, at least 1 million elderly adults were estimated as having a substance use disorder with a concurrent rise in admission rates for substance use disorders in this population. ²⁰ , ²¹

This study sought to describe the prevalence of chronic opioid use amongst community‐dwelling people aged at least 65 years and the risk of CV outcomes associated with their use.

Methods

Data source and study population

The trial design, methods, and main results of the ASPREE trial have been published previously. ²² , ²³ , ²⁴ , ²⁵ The Aspirin in Reducing Events in the Elderly (ASPREE) trial was an international, multicentre, double‐blind, randomized controlled clinical trial that compared daily 100 mg aspirin with placebo. Participants were community‐dwelling adults living in Australia and the USA who were ≥70 years of age (≥65 years amongst US participants) and had no prior CVD events, or dementia, and major physical disability or any other chronic illness expected to limit survival to less than 5 years. The ASPREE trial concluded in 2017, after a median follow‐up of 4.7 years, and the study included annual clinic visits. ²² , ²³ , ²⁴ , ²⁵

Ascertainment of regular opioid exposure

All participants enrolled into the ASPREE trial completed two baseline visits to finalize eligibility and, after randomization, were seen annually for trial assessments conducted by trained staff. Opioid use was defined as use either at baseline or at the AV1. Participants who started opioids before AV1 and had a new diagnosis of cancer were adjusted for in the analysis. Information on opioid use (including types), along with the use of other concomitant prescription medications, were recorded at ASPREE trial entry and updated at annual intervals during follow‐up. Use of non‐steroidal anti‐inflammatory drugs (NSAIDs) was defined as use of salicylic acid and derivatives or non‐steroidal anti‐inflammatories (WHO ATC codes N02BA and MO1A). ²⁶ The information was collected by asking participants to bring all current or recently used medications to study visits.

Outcome measures

The primary outcome of the ASPREE trial was disability‐free survival, and results have been previously reported. ²⁴ Incident CVD was a prespecified, composite secondary endpoint of ASPREE, consisting of fatal coronary heart disease (death from MI, sudden cardiac death, cardiac failure death, or any other death in which the underlying cause was considered to be coronary heart disease), non‐fatal MI, fatal or non‐fatal stroke (haemorrhagic or ischaemic), non‐coronary cardiac or vascular death, or hospitalization for heart failure. ²² All CVD events were adjudicated by a panel of experts blinded to treatment group assignment.

In this post hoc analysis, we primarily used the prespecified composite CVD endpoint to explore the association between opioid use and incident CVD events. ²³ This outcome was measured after AV1 and any CVD events that occurred before AV1 were censored (Figure 1 ). We also explored the association of baseline opioid use with individual endpoints—ischaemic stroke, heart failure hospitalization, MI (non‐fatal/fatal), and all‐cause mortality.

Figure 1.

Analysis sample. Of the 19 114 individuals enrolled into ASPREE, 18 851 attended AV1; of these individuals, 18 019 had medication information available at baseline and AV1. There were 318 individuals who were censored at or before AV1 for CVD, resulting in a final sample of 17 701.

Statistical analyses

Continuous variables are presented as means ± standard deviations. Variables that were non‐normally distributed are recorded as median [interquartile range (IQR)]. Categorical data are presented as counts and percentages. For comparisons between opioid users and non‐users, Pearson's chi square (x ² ) test or Fisher's exact test was used as appropriate for categorical variables. T‐tests (for symmetrically distributed data) or non‐parametric tests (e.g. Mann–Whitney tests for non‐symmetrically distributed data) were used for continuous variables.

Cox proportional hazards regression was used to calculate hazard ratios (HR) and 95% confidence intervals (95% CI) for time to the first CVD event occurring during the follow‐up period after AV1. The unadjusted Cox proportional hazards model was followed with adjustment for accumulating sets of covariates, which included age and sex (Model 1); BMI, diabetes, hypertension, eGFR, smoking (current), country of residence, ethnicity, and aspirin use (Model 2); depression, frailty, dyslipidaemia, and other NSAID use (Model 3); and cancer (Model 4). In addition, cumulative incidence of the CVD endpoint as well as the other individual endpoints was computed according to the complement of the Kaplan–Meier estimates of event‐free survival. Cumulative incidence functions were computed for individual endpoints to account for competing risks. Analyses were repeated and stratified by sex.

Further, propensity‐matched samples based on propensity scores was used as a sensitivity analysis to minimize the effect of bias and confounding as well as the difference in the number of participants with or without opioid use at baseline. We used a 1:1 nearest neighbour approach to select similar numbers of samples for both groups. Briefly, the propensity score was generated for each participant based on key baseline characteristics: age, sex, hypertension, obesity, diabetes, dyslipidaemia, depression, and frailty status (Model 1) and further adjusted for smoking and other NSAID use (Model 2).

Results

Of the original 19 114 individuals enrolled into ASPREE, 18 851 attended AV1; of these individuals, 18 019 had medication information available at baseline and AV1. There were 318 individuals who were censored at, or before AV1 for CVD, resulting in a final sample of 17 701 (Figure 1 ). In the sub‐analysis of participants who had started opioid use at AV1 (n = 342), 300 had been diagnosed with cancer before AV1.

The mean (SD) age of included participants was 75.1 (4.5) years and 58.2% females. Prescribed opioids either at baseline or at AV1 were present in 813 participants. Table 1 shows the baseline characteristics clinical characteristics of the participants, stratified by opioid users and non‐users. Compared with participants who did not take opioids, those who used opioids were more likely to be female (70.7% vs. 56.6%, P < 0.001); have diabetes (13.5% vs. 10.9%, P = 0.018), hypertension (81.1% vs. 75.3%, P < 0.001), obesity (45.8% vs. 29.3%, P < 0.001), and smoke (6.9% vs. 3.4%, P < 0.001); and demonstrated increased frailty (62% vs. 40.3%, P < 0.001).

Table 1.

Patient baseline characteristics based on baseline opioid use

	All (n = 17 701)	Opioid use
		Yes (n = 813)	No (n = 16 888)	P‐value
Age (mean ± SD), years	75.1 ± 4.5	75.80 ± 4.76	75.12 ± 4.50	<0.001
Male (%)	42.8	29.3	43.4	<0.001
Weight (mean ± SD), (kg)	77.0 ± 15	80.0 ± 17.1	76.9 ± 14.9	<0.001
Height (m)	1.65 ± 0.1	1.63 ± 0.1	1.65 ± 0.1	<0.001
BMI (kg/m2)	28.1 ± 4.7	30.2 ± 5.9	28.1 ± 4.6	<0.001
BSA(m2)	3.6 ± 0.8	3.6 ± 0.9	3.6 ± 0.8	0.093
Diabetes (%)	11.0	13.5	10.9	0.001
Hypertension (%)	75.6%	81.1%	75.3%	<0.001
Systolic blood pressure (mmHg)	139.3 ± 16.5	138.4 ± 17.8	139.3 ± 16.4	0.417
Diastolic blood pressure (mmHg)	77.3 ± 10	76.7 ± 11	77.3 ± 10	0.411
Baseline heart rate	70.8 ± 10.7	72.2 ± 11.2	70.7 ± 10.7	0.051
Baseline eGFR	72.7 ± 13.9	72.0 ± 15.0	73.0 ± 13.8	0.149
Smoking status (%)	3.6%	6.9%	3.4%	<0.001
Familial history of coronary artery disease (%)	61.3%	62.5%	61.3%	0.49
Dyslipidaemia (%)	66.2%	66.7%	66.1%	0.757
Total cholesterol (mmol/L)	5.23 ± 0.98	5.19 ± 1.05	5.23 ± 0.98	0.006
HDL (mmol/L)	1.58 ± 0.46	1.60 ± 0.50	1.58 ± 0.46	0.500
LDL (mmol/L)	3.03 ± 0.88	2.90 ± 0.92	3.04 ± 0.87	<0.001
Triglycerides (mmol/L)	1.33 ± 0.66	1.51 ± 0.82 s	1.32 ± 0.65	<0.001
Frailty				<0.001
Not frail	58.8%	38.0%	59.8%
Pre‐frail	39.1%	55.1%	38.3%
Frail	2.2%	6.9%	2.0%

BMI, body mass index; BSA, body surface area; eGFR, estimated glomerular filtration rate; HDL, high‐density lipoproteins; LDL, low‐density lipoproteins.

Hypertension: Based on patient's Participant Medical History‐Baseline form. Diabetes: Based on patient's Participant Medical History‐Baseline form. Frailty: Baseline frailty according to modified Fried criteria.

Opioid users also used more concomitant medications (Tables 2 and S1), including beta‐blockers (10.9% vs. 8.3%, P = 0.007), angiotensin‐converting enzyme inhibitors (ACEI)/angiotensin receptor blockers (ARB) (52.5% vs. 43.2%, P < 0.001), lipid‐lowering drugs (40.6% vs. 35.2%, P = 0.002), psychotropic medications (5.2% vs. 2.4%, P < 0.001), and chemotherapy agents (3.9% vs. 1.8%, P < 0.001). This group were more likely to use other analgesics such as NSAIDs (43.1% vs. 20.5%, P < 0.001) and pregabalin (14.9 vs. 2.2%, P < 0.001). In addition, opioid users also had increased usage of anxiolytics including benzodiazepines (13.9% vs. 5.2%, P < 0.001).

Table 2.

Concomitant medication exposure

	All (n = 17 701)	Opioid use		P‐value
		Yes (n = 813)	No (n = 16 888)
Aspirin (%)	11	12.7	10.9	0.120
Beta‐blockers (%)	8.4	10.9	8.3	0.007
ACE inhibitors/ARB (%)	43.6	52.5	43.2	<0.001
Anti‐hyperlipidaemics (%)	35.5	40.6	35.2	0.002
NSAID use (%)	21.5	43.1	20.5	<0.001
Benzodiazepine (%)	5.6	13.9	5.2	<0.001
Psychotropic medications (%)	2.5	5.2	2.4	<0.001
Pregabalin (%)	2.8	14.9	2.2	<0.001
Chemotherapy (%)	1.9	3.9	1.8	<0.001
Alcohol use				<0.001
Current (%)	76.8	70.8	77.1
Former (%)	5.8	7.8	5.7
Never (%)	17.4	21.4	17.2

NSAID use: Salicylic acid and derivatives or non‐steroidal anti‐inflammatories (WHO ATC codes N02BA and MO1A); benzodiazepine (WHO ATC codes NO5CD); psychotropic medications (WHO ATC codes NO5A); pregabalin (WHO ATC codes NO3AX); chemotherapy (WHO ATC codes LO1). ²¹

Following AV1, over a median follow‐up period of 3.58 years (IQR 2.50–4.62), CVD events occurred in more opioid users compared to non‐ users: 21.2 per 1000‐person‐years vs. 11.5 per 1000‐person‐years, P < 0.001. In all, 57 (7%) opioid users compared with 680 (4%) of non‐users experienced one of the components of the composite CVD endpoint (Figure 2 and Table 3 ). In the univariate analysis (Model 1), a significantly increased risk for CVD events was observed, which persisted when analyses were adjusted for body mass index (BMI), diabetes, hypertension, estimated glomerular filtration rate (eGFR), smoking (current), country of residence, ethnicity, and aspirin use (Model 2) and further adjusted for depression, frailty, dyslipidaemia, and other NSAID use (Model 3) and cancer (Model 4). The increased risk of CVD events continued even after adjustment for other concomitant substances such as alcohol and benzodiazepines and the use of other medications such as psychotropic medications, pregabalin, and chemotherapy agents (Model 5) (Table 3 ). Propensity‐matched findings (Table 4 ) were consistent, with an increased incidence of CVD events associated with opioid use (HR = 1.74, CI 1.13–2.67, P = 0.011) compared with non‐users (Model 1). The propensity‐matched sample was further adjusted for smoking and other NSAID use with an HR 1.88 (IQR 1.21–2.91, P = 0.005) (Model 2).

Figure 2.

Cumulative incidence of first CVD event occurring after AV1. Following AV1, CVD events occurred in more opioid users compared with non‐users: 21.2 per 1000‐person‐years vs. 11.5 per 1000‐person‐years, P < 0.001. In all, 57 (7%) opioid users compared with 680 (4%) of non‐users experienced one of the components of the composite CVD endpoint. AV1, Annual Visit 1; CVD, cardiovascular.

Table 3.

Association between opioid use and CVD events occurring after AV1

	All			Male			Female
	Opioid use			Opioid use			Opioid use
CVD, N	Yes (n = 813)	No (n = 16 888)	P‐value	Yes (n = 238)	No (n = 7330)	P‐value	Yes (n = 575)	No (n = 9558)	P‐value
Observed CVD events, N (%)	57 (7.0%)	680 (4.0%)		18 (7.6%)	386 (5.3%)		39 (6.8%)	294 (3.1%)
Event rate per 1000‐person‐years	21.2	11.5	P < 0.001	24.3	15.2	P = 0.067	20.1	8.7	P < 0.001
Hazard ratio (95% CI)
Model 1	1.94 (1.48–2.54)		P < 0.001	1.56 (0.97–2.50)		P = 0.067	2.19(1.57–3.06)		P < 0.001
Model 2	1.82 (1.37–2.41)		P < 0.001	1.37 (0.83–2.27)		P = 0.222	2.12 (1.50–2.99)		P < 0.001
Model 3	1.69(1.26–2.24)		P < 0.001	1.26 (0.75–2.10)		P = 0.381	1.95(1.37–2.77)		P < 0.001
Model 4	1.67 (1.26–2.23)		P < 0.001	1.26 (0.75–2.10)		P = 0.38	1.93 (1.36–2.74)		P < 0.001
Model 5	1.67(1.25–2.23)		P < 0.001	1.30 (0.77–2.17)		0.325	1.90(1.33–2.72)		P < 0.001

CVD, cardiovascular disease.

Model 1: Adjusted for age and gender. Model 2: Model 1 plus adjustment for BMI, diabetes, hypertension, eGFR, smoking (current), country of residence, ethnicity, and aspirin use. Model 3: Model 2 plus adjustment for depression, frailty, dyslipidaemia, and other NSAID use. Model 4: Model 3 plus adjustment for cancer. Model 5: Plus adjustment for alcohol use, concomitant use of benzodiazepine, psychotropic medications, pregabalin, and chemotherapy.

Table 4.

CVD amongst propensity‐matched sample

	Coefficient	95%CI
CVD events, median (IQR)
Model 1 a	1.74	1.13, 2.67 (P = 0.011)
Model 2 b	1.88	1.21, 2.91 (P = 0.005)

^a Unadjusted.

^b Adjusted for smoking and other NSAIDs.

In the analysis of other individual end points following AV1, there were higher heart failure hospitalizations [18 (2.2%) vs. 123 (0.72%), P < 0.001], ischaemic strokes [17 (2.1%) vs. 224 (1.3%), P < 0.001], and deaths [67 (8.2%) vs. 762 (4.5%), P < 0.001] in opioid users compared with the non‐users (Table 5 ). The incidence of heart failure hospitalizations and deaths remained significantly higher in opioid users despite full covariate adjustment. Myocardial infarction (non‐fatal/fatal) was only significantly higher amongst opioid users in Model 1 HR = 1.77(1.11–2.83) P < 0.001 and Model 2 HR:1.68(1.03–2.73) P < 0.001, however not after further covariate analyses in Model 3 and 4. Figure 3 shows the cumulative incidence of ischaemic stroke, heart failure hospitalization, MI (non‐fatal/fatal), and all‐cause mortality.

Table 5.

Association between opioid use and individual components of CVD events occurring after AV1

	All			Male			Female
	Opioid use			Opioid use			Opioid use
Heart failure, N	Yes (n = 813)	No (n = 16 888)	P‐value	Yes (n = 238)	No (n = 7330)	P‐value	Yes (n = 575)	No (n = 9558)	P‐value
Observed CVD events, N (%)	18(2.2%)	123 (0.73%)		6 (2.5%)	68 (0.93%)		12(2.1%)	55 (0.58%)
Event rate per 1000‐person‐years	6.6	2.0	P < 0.001	7.9	2.6	P = 0.067	6.1	1.6	P < 0.001
Hazard ratio (95% CI)
Model 1	3.27 (1.99–5.39)		P < 0.001	2.80 (1.21–6.47)		P = 0.016	3.58(1.92–6.68)		P < 0.001
Model 2	2.72 (1.61–4.58)		P < 0.001	2.26 (0.089–5.69)		P = 0.085	2.94(1.54–5.60)		P = 0.001
Model 3	2.37(1.39–2.36)		P < 0.001	1.79 (0.69–4.67)		P = 0.231	2.62(1.35–5.09)		P = 0.005
Model 4	2.35 (1.38–4.02)		P = 0.002	1.76 (0.68–4.58)		P = 0.245	2.61 (1.35–5.08)		P = 0.005
Model 5	2.36 (1.36–4.07)		0.002	1.86 (0.71–4.87)		0.204	2.58(1.30–5.10)		P = 0.006

	All			Male			Female
	Opioid use			Opioid use			Opioid use
Stroke, N	Yes (n = 813)	No (n = 16 887)	P‐value	Yes (n = 238)	No (n = 7329)	P‐value	Yes (n = 575)	No (n = 9558)	P‐value
Observed CVD events, N (%)	17 (2.1%)	224 (1.3%)		3 (7.9%)	126 (17.2%)		14 (2.4%)	98 (1.0%)
Event rate per 1000‐person‐years	6.2	3.7	P = 0.056	4.0	4.9	P = 0.778	7.1	2.9	P = 0.004
Hazard ratio (95% CI)
Model 1	1.73 (1.05–2.83)		P < 0.001	0.80 (0.25–2.51)		P = 0.70	2.34 (1.34–4.10)		P = 0.003
Model 2	1.66(0.99–2.78)		P = 0.053	0.80 (0.25–2.53)		P = 0.70	2.20 (1.22–3.96)		P = 0.008
Model 3	1.61 (0.96–2.72)		P = 0.073	0.79 (0.25–2.52)		P = 0.69	2.15(1.18–3.92)		P = 0.012
Model 4	1.60 (0.95–2.70)		P = 0.078	0.79 (0.25–2.52)		P = 0.69	2.09 (1.15–3.81)		P = 0.016
Model 5	1.65 (0.97–2.79)		P = 0.063	0.82(0.26–2.64)		P = 0.742	2.17(1.18–3.98)		P = 0.013

	All			Male			Female
	Opioid use			Opioid use			Opioid use
Myocardial infarction (non‐fatal/fatal), N	Yes (n = 813)	No (n = 16 887)	P‐value	Yes (n = 238)	No (n = 7329)	P‐value	Yes (n = 575)	No (n = 9558)	P‐value
Observed CVD events, N (%)	19(2.3%)	262 (1.6%)		9(3.8%)	173(2.4%)		10 (1.7%)	89 (0.93%)
Event rate per 1000‐person‐years	6.9	4.4	P = 0.067	11.9	6.8	P = 0.117	5.0	2.6	P = 0.066
Hazard ratio (95% CI)
Model 1	1.77 (1.11–2.83)		P < 0.001	1.73(0.89–3.38)		P = 0.109	1.80(0.94–3.46)		P = 0.078
Model 2	1.68 (1.03–2.73)		P < 0.001	1.54(0.75–3.14)		P = 0.24	1.78 (0.91–3.46)		P = 0.091
Model 3	1.57(0.96–2.57)		P = 0.074	1.40 (0.67–2.90)		P = 0.368	1.70(0.86–3.36)		P = 0.127
Model 4	1.57 (0.96–2.58)		P = 0.073	1.40(0.68–2.90)		P = 0.37	1.69(0.86–3.34)		P = 0.13
Model 5	1.52 (0.92–2.51)		P = 0.102	1.38(0.66–2.89)		P = 0.39	1.59 (0.79–3.18)		P = 0.192

	All			Male			Female
	Opioid use			Opioid use			Opioid use
All‐cause mortality, N	Yes (n = 813)	No (n = 16 888)	P‐value	Yes (n = 238)	No (n = 7330)	P‐value	Yes (n = 575)	No (n = 9558)	P‐value
Observed CVD events, N (%)	67 (8.2%)	762 (4.5%)		27 (11.3%)	423 (5.8%)		40 (7.0%)	339 (3.5%)
Event rate per 1000‐person‐years	23.3	12.3	P < 0.001	34.2	4.9	P < 0.001	19.2	2.9	P < 0.001
Hazard ratio (95% CI)
Model 1	1.98 (1.54–2.55)		P < 0.001	2.13(1.45–3.15)		P < 0.001	1.90(1.37–2.63)		P < 0.001
Model 2	1.97 (1.52–2.56)		P < 0.001	1.97 (1.30–2.99)		P = 0.002	1.96 (1.40–2.73)		P < 0.001
Model 3	1.81(1.39–2.36)		P < 0.001	1.90 (1.24–2.91)		P = 0.003	1.73(1.23–2.43)		P = 0.002
Model 4	1.68 (1.29–2.18)		P < 0.001	1.84 (1.20–2.82)		P = 0.005	1.55(1.10–2.17)		P = 0.012
Model 5	1.63 (1.25–2.14)		P < 0.001	1.72(1.11–2.66)		P = 0.014	1.55(1.10–2.19)		0.012

Figure 3.

Cumulative incidence of individual components of CVD events occurring after AV1. In the analysis of individual end points following AV1, there were higher heart failure hospitalizations [18 (2.2%) vs. 123 (0.72%), P < 0.001], ischaemic strokes [17 (2.1%) vs. 224 (1.3%), P < 0.001], and deaths [67 (8.2%) vs. 762 (4.5%), P < 0.001] in opioid users compared with the non‐users.

In the sub‐analysis by sex, the rates of CVD and their individual components including heart failure, stroke, and all‐cause mortality were significantly elevated in female opioid users compared with non‐users (Tables 3 and 5 ). The rate of CVD events in females was 20.1 per 1000‐person‐years vs. 8.7 per 1000‐person‐years (P < 0.001) in opioid users compared with non‐users. These findings were consistent across all models: Model 1 HR = 2.19 (IQR 1.57–3.06), P < 0.001; Model 2 HR = 2.12 (1.50–2.99), P < 0.001; Model 3 HR = 1.95 (IQR 1.37–2.77), P < 0.001; Model 4 HR = 1.93 (IQR 1.36–2.74), P < 0.001; and Model 5 HR = 1.90 (IQR 1.33–2.72), P < 0.001. The HR associated with heart failure, stroke, and all‐cause mortality also remained significantly elevated after adjusting for all covariates. This contrasted to the male sub‐analysis, whereby the rate of CVD events in male opioid users compared with non‐users was not significantly different (24.3 per 1000‐person‐years vs. 15.2 per 1000‐person‐years, P = 0.067). Furthermore, within the individual components constituting CVD events, only all‐cause mortality remained significantly elevated after adjusting for all covariates, HR = 1.84 (IQR 1.20–2.82), P = 0.005.

Discussion

In this adjusted post hoc analysis of the ASPREE trial, participants who took opioids were at increased risk of the composite CVD endpoint, which included fatal coronary heart disease, nonfatal MI, stroke, hospitalization for heart failure, or all‐cause mortality. The risk of each of the individual endpoints including heart failure, stroke, and all‐cause mortality was also significantly elevated in opioid users compared with non‐users. We also observed important sex differences in the association of prescribed opioids and CVD endpoints. The impact of opioids on CVD is currently limited to observational data due to the unfeasibility of conducting randomized trials examining this relationship. As such, this study is integral to the growing body of literature highlighting the potential health risk in the use of therapeutic opioids, thereby steering clinical practice away from the prescription of opioids for chronic non‐cancer pain, especially for females.

Recent reports have suggested a relationship between opioid use and increased risk of coronary artery disease. ³ , ⁴ , ⁵ Li et al. reported a 1.28‐fold risk of MI in opioid users compared with non‐ users. ⁷ Carman et al. estimated that the incidence rate ratios for MI and MI/coronary revascularization in a cohort of chronic opioid users versus a matched cohort from the general population was 2.66 (95% CI, 2.3–3.08) and 2.39 (95% CI, 2.15–2.63), respectively. ²⁷ Finally, a recent study by Khodneva et al. found that prescribed opioid use was associated with significantly increased risk (1.24‐fold) of CV death. ²⁸

Although the present study is associative in nature, findings are biologically plausible. A prior histopathologic study of myocardial samples from subjects dying from illicit drug‐related deaths demonstrated a strong relationship between opioid use and myocardial fibrosis. ²⁹ This observation would be consistent with the present study's finding of a significant increase in heart failure events in opioid users. Opioid receptors in human myocardium may also play a role in neural transmission and regulation of myocardial function. ³⁰ Chronic and higher doses of opioid use have been shown to increase myocardial ischaemia and oxidative stress via opioid receptor‐dependent mechanisms, which interfere with cellular function. ²⁹ In rat myocardium, a dose‐dependent increased susceptibility to reperfusion injury after administration of remifentanil has been identified. ³¹ Reese et al. noted increased inflammatory markers such as C‐reactive protein and accelerated atherosclerosis in chronic opioid users. ³² Other potential pathophysiologic mechanisms include increased platelet aggregation through stimulation of the platelet/endothelial cell adhesion molecule 1 and glycoprotein IIb expression, thereby decreasing the protective effects of aspirin and perhaps providing an explanation for the observed increase in stroke rate. ³³

The relationship between chronic opioid use and the development of heart failure could potentially be attributable to the development of sleep‐disordered breathing including sleep apnoea with long‐term opioid use. ³⁴ In a recent study, Solhjoo demonstrated that nocturnal oxygen saturation was significantly reduced in a dose dependent manner in patients on chronic methadone therapy. ³⁵ Central sleep apnoea is present in 40% of patients with heart failure and is a strong independent marker of mortality. ³⁶ , ³⁷ The relationship between obstructive sleep apnoea and heart failure is complex but is likely attributable to increased sympathetic and neurohumoral activation secondary to hypoxia, coupled with increased oxidative stress, resulting in significant CV morbidity. ³⁸ , ³⁹ , ⁴⁰ , ⁴¹

Amongst participants using opioids, the use of concomitant NSAIDs was significantly higher (43.1% vs. 20.5%, P < 0.001). The use of NSAIDs is independently associated with an increase in the absolute risk of adverse CV events, ⁴² , ⁴³ , ⁴⁴ with higher doses and frequencies increasing the risk of MI. ⁴⁵ , ⁴⁶ , ⁴⁷ Furthermore, the combination of prescribed opioid use with acetaminophen has been more strongly associated with CV endpoints compared with non‐users of both. ²⁷

Chronic opioid use is often associated with polysubstance abuse, which could be a potential contributor to the development of heart failure and poorer outcomes in opioid users. ⁴⁸ , ⁴⁹ , ⁵⁰ Long‐term excess alcohol consumption is a known risk factor for the development of dilated cardiomyopathy. ⁵¹ Specka et al. reported that 90% of opioid users consumed at least one other psychoactive substance at admission into their study. Amongst those, alcohol and benzodiazepines were found to be the most common. ⁵² The use of concomitant substances with opioids could be attributed to the attenuation in processing of painful stimuli through sympathetic stimulation, hypothalamic–pituitary–adrenal axis dysregulation, and pro‐inflammatory immune‐system activation, resulting in an increased sensitivity to pain or decreased pain tolerance. ⁵³ , ⁵⁴ The relationship between chronic opioid use and polysubstance abuse in our cohort was variable. Participants using opioids were less likely to be current alcohol drinkers (70.8% vs. 77.1%, P < 0.001), and a greater percentage had never used alcohol (21.4% vs. 17.2%, P < 0.001). However, the concomitant use of benzodiazepines was significantly higher in opioid users compared with non‐users (13.9% vs. 5.2%, P < 0.001). The relationship between anxiolytics such as benzodiazepine use and heart failure has limited evidence. ⁵⁵ However, Zwas et al. noted that the treatment of heart failure patients with anxiolytics portended a worse prognosis irrespective of concomitant depression, and it remained a predictor of mortality. ⁵⁶ The possible confounding effect of both alcohol and benzodiazepine use was accounted for in the multi‐variate analysis in Model 5. Despite this, chronic opioid use was still associated with increased CVD events, HR = 1.67(IQR 1.25–2.23), P < 0.001.

Mood, sleep, and personality disorders can also aggravate pain symptoms and are frequently comorbid in patients with chronic pain. ⁵⁷ , ⁵⁸ , ⁵⁹ , ⁶⁰ , ⁶¹ This was reflected in our study whereby participants using opioids had an increased use of psychotropic medications (5.2% vs. 2.4%, P < 0.001). Although mental health disorders have been independently associated with an increased risk of CV mortality, the use of psychotropic drugs themselves are also associated with an increased risk of sudden cardiac death. ⁶² , ⁶³ Most notably, clozapine treatment is consistently linked with the development myocarditis and cardiomyopathy, with the risk of developing cardiomyopathy five times greater in patients treated with clozapine than the general population. ⁶⁴ , ⁶⁵ Depression, which is prevalent in 5–10% of community‐dwelling adults aged 65 years and older, has been accepted as a psychosocial risk factor for coronary heart disease with the 2016 European Guidelines on CVD prevention in clinical practice recommending screening and treatment of depression. ⁶⁶ , ⁶⁷ Depression and the use of psychotropic medications were accounted for in Model 3 and Model 5 of the multi‐variate analysis (Table 3 ); however, the risk of CVD events associated with opioid use was still elevated.

Sex differences in the association of CVD and opioid use, more pronounced in females, have been previously reported. ²⁸ Compared with males, females are also more likely to be prescribed opioids or prescribed higher doses and to use opioids for longer periods of time. ⁵⁷ One explanation for the higher CVD risks in female opioid users may be related to the finding that chronic opioid use decreases hypothalamic–pituitary–ovarian axis activity and may decrease oestrogen levels, potentially increasing CVD risk in females. ⁶⁸ In the REGARDS study, female opioid users were also more likely to use combined opioid–acetaminophen preparations than males prescribed opioids. The present study findings concurred with this sex difference, with the rates of CVD events in female opioid users being significantly higher compared with non‐users in all models of the multi‐variable analyses. ²⁸

A major study strength is the large sample of community‐dwelling participants, who, after robust screening at baseline, had no prior CVD events. The data were also obtained from a contemporary randomized trial in a large older population with age and sex distributions comparable with population norms. ⁶⁹ The ASPREE trial was subject to rigorous quality control, and all endpoints were formally adjudicated by experts blinded to treatment allocation. ²² , ²³ , ²⁴ , ²⁵ , ⁷⁰

Study limitations

Due to the observational design of the present study, causal relations between opioid use and the incidence of CV events cannot be established. Opioid use was defined as documented use either at baseline and/or AV1; however, the specific opioids were not included as overall duration and cumulative dose of opioid use cannot be ascertained. Ideally, a dose–response analysis would have been included to help strengthen the association for causality. Prescribed opioid use at subsequent annual visits was not included, and the measurement of incident CVD events was measured after AV1 to account for immortal‐time bias. Furthermore, some participants may have started or discontinued opioids after AV1, which may impact results. In the sub‐analysis of patients who started opioid use at AV1 (n = 342), 300 had been diagnosed with cancer prior to AV1. However, the diagnosis of cancer prior to AV1 has been adjusted for in Model 4. Despite thorough attempts to control for extensive confounders in the multi‐variate analysis, there are still unmeasured confounders. For example, we did not assess the effect of chronic pain per se on CV outcomes. As such, residual confounding bias cannot be ruled out.

Conclusions

Amongst community‐dwelling adults ≥65 years of age, opioid use, especially in women, may be associated with a higher incidence of CV events, stroke, heart failure hospitalization, and all‐cause mortality. The study's observational design limits causal inferences but strikes a cautionary note for all medical practitioners to limit the prescription of opioids in the setting of non‐cancer pain. Further research needs to be undertaken to fill the knowledge gaps surrounding the use of opioids in older individuals.

Conflict of interest

The remaining authors have nothing to disclose.

Funding

The work was supported by the National Institute on Aging and the National Cancer Institute at the National Institutes of Health (Grant Number U01AG029824), the National Health and Medical Research Council of Australia (Grant Numbers 334047 and 1127060); Monash University (Australia) and the Victorian Cancer Agency (Australia). J.N. is recipient of a research fellowship by the Deutsche Forschungsgemeinschaft (NE 2165/1‐1). C.M.R. is supported through an NHMRC Principal Research Fellowship (APP 1136372). J.M.C.N. is supported by an NHMRC Leadership Fellowship (IG 1173690). D.K. is supported by an NHMRC Senior Principal Research Fellowship (APP1136080).

Supporting information

Table S1. Concomitant Medication exposure.

Table S2. CVD among propensity matched samples.

Table S2. Association between opioid use and individual components of CVD events occurring after AV1.

Liew, S. M. , Chowdhury, E. K. , Ernst, M. E. , Gilmartin‐Thomas, J. , Reid, C. M. , Tonkin, A. , Neumann, J. , McNeil, J. J. , and Kaye, D. M. (2022) Prescribed opioid use is associated with adverse cardiovascular outcomes in community‐dwelling older persons. ESC Heart Failure, 9: 3973–3984. 10.1002/ehf2.14101.