SEARCH

SEARCH BY CITATION

Keywords:

  • GPRD ;
  • myocardial infarction;
  • nested case–control;
  • noncancer pain;
  • Opioids

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conflict of interest statement
  8. Acknowledgements
  9. References
  10. Appendix

Backgrounds

With increasing use of opioids for chronic noncancer pain comes concern about safety of this class of drugs. Opioid-induced hypogonadism, which could increase the risk for myocardial infarction (MI), has recently come to the attention of clinicians. To evaluate this concern we examined the association between opioid use for noncancer pain and risk of MI amongst adults.

Methods

We conducted a nested case-control study using the UK General Practice Research Database. Amongst 1.7 million opioid users during 1990–2008, we identified 11 693 incident MI cases aged 18–80 years, and randomly selected up to four controls matched by age, gender, index date (date of onset symptoms or diagnosis of first-ever MI) and general practice via risk-set sampling. Cases and controls were required to have no cancer and no major risk factors for MI before the index date. Adjusted odds ratios (ORs) and 95% confidence intervals (CIs) were estimated from conditional logistic regression.

Results

Compared with nonuse, current use of opioids was associated with a 1.28-fold (95% CI 1.19–1.37) risk of MI. Cumulative use of opioids with 11–50 (OR = 1.38, 95% CI: 1.28–1.49) or > 50 (OR = 1.25, 95% CI: 1.11–1.40) prescriptions, was also marginally associated with increased risk of MI. The risk was particularly increased in users of morphine (OR = 1.71, 95% CI: 1.09–2.68), meperidine (OR = 2.15, 95% CI: 1.24–3.74) and polytherapy (OR = 1.46, 95% CI: 1.22–1.76).

Conclusions

Current use of any opioids and cumulative use of 11 or more prescriptions are associated with a small increased risk for MI compared to nonuse and the risk was greater in morphine, meperidine and polytherapy users. Residual confounding, particularly confounding by indication, should be considered in interpreting our results.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conflict of interest statement
  8. Acknowledgements
  9. References
  10. Appendix

Chronic noncancer pain (CNCP) is a common condition which can undermine overall physical, psychological, and social well-being [1, 2]. With increased use of opioid analgesics in patients with CNCP over the last decade, comes concern about the safety of this class of drugs [1, 3-6]. That opioids decrease plasma testosterone levels has long been documented in narcotic addicts [7-10]. Now similar hormonal changes not only limited to testosterone, but also oestrogen, luteinizing hormone, gonadotrophin releasing hormone and dehydroepiandrosterone, are also observed in those with administration of opioids for CNCP [11-18]. The primary mechanism for opioid-induced hypogonadism is suppression of the hypothalamic-pituitary-gonadal axis [19-21].

It is well known that the age-adjusted morbidity and mortality rates from coronary artery disease are 2.5- to 4.5-fold higher in men than in women and that the gender gap narrows after menopause [22], suggesting that either gender or sex hormones may exert a significant influence on the cardiovascular system [23]. Testosterone was reported to suppress activation of pro-inflammatory cytokines which are believed to play a central pathogenic role in the initiation and progression of coronary atheroma and its clinical consequences [24]. Men with high serum testosterone levels were observed to be independently associated with a reduced risk for aortic atherosclerosis or lower arterial stiffness compared to men with low testosterone levels [25, 26], and low testosterone levels were reported to predict 1.38-fold risk of cardiovascular mortality in men [27]. Sex hormones are also associated with other cardiovascular risk factors such as adverse lipid profiles and insulin resistance [28, 29].

Although opioid-induced hypogonadism has recently come to the attention of clinicians [30-35], there is still limited clinical awareness of the endocrine effects of opioids, together with the lack of information on their long-term effects [6, 19, 34]. Thus, we hypothesized that opioid use for noncancer pain was associated with an increased risk of myocardial infarction (MI) amongst adults. To test this hypothesis we conducted a nested case–control study using the UK-based General Practice Research Database (GPRD).

Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conflict of interest statement
  8. Acknowledgements
  9. References
  10. Appendix

Data source and study sample

Data were derived from the GPRD that is housed and organized for research purposes at the Boston Collaborative Drug Surveillance Program. The GPRD is a large primary care database that has been previously described in detail [36-38]. Briefly, the GPRD is an ongoing longitudinal database that has collected data from over 500 general practices in the UK since 1987. It contains information on more than 5 million patients, of which over 3.5 million are currently active (approximately 5% of the UK population), and has a representative age- and gender- distribution of the entire UK population. The information recorded includes patient demographics and characteristics (e.g. height, weight and smoking status), clinical diagnoses, drug prescriptions, consultant referrals and hospitalizations. Validation studies have shown the high quality of recorded drug exposure and diagnoses [36-40], including diagnosis of MI [41-43].

The study sample comprised 1.7 million noncancer patients who had a record for at least one opioid prescription (Table A1) between 1 January 1990 and 31 December 2008 in the GPRD. Around 70% were new opioid users who had at least 1 year of recorded medical history before the first documented opioid prescription. This study was reviewed and approved by the Independent Scientific Advisory Committee for UK Medicines and Healthcare products Regulatory Agency database research.

Case selection

Amongst the study sample we identified all incident MI cases, and defined the index date as the date of the first-ever recorded MI during 1990–2008, or the date of onset of prodromal symptoms of MI (e.g. chest pain, chest pressure, radiating pain, etc.) recorded within 28 days before the first MI diagnosis, where available. We required each case to be aged 18–80 years on the index date and to have at least 2 years of medical history recorded before the index date. We excluded any patients with a history of conditions such as cancer (except nonmelanoma skin cancer), ischaemic heart disease, heart failure, stroke, congenital heart disease, heart transplantation, cardiac arrhythmias, treated hypertension, diabetes, alcohol/drug abuse, or serious liver or renal disease any time before the index date. We also excluded any patients with a cardiac surgery recorded within 90 days before the index date. This was done to exclude secondary MIs caused by recent cardiovascular events or procedures.

Control selection

For each MI case we randomly selected up to four controls matched on age, gender, index date and general practice (GP) from the study sample via risk-set sampling [44]. The same inclusion and exclusion criteria applied to cases were also applied to controls.

Opioid exposure

We identified all opioid prescriptions recorded any time before the index date for cases and controls from the computerized records without knowledge of disease status. We classified someone as a ‘nonuser’ if there was no recorded opioid prescription within 2 years before the index date (reference), including subjects with no opioid exposure any time before the index date and those whose opioid therapy ended > 2 years before the index date. For those with opioid prescriptions recorded within 2 years before the index date, we evaluated exposure in terms of timing of use, cumulative use, and individual opioid type. On the basis of the end date of the most recent opioid prescription's supply prior to the index date we categorized timing of use as current use (0–30 days), recent use (31–365 days), or past use (366–730 days). We defined cumulative use as the number of opioid prescriptions recorded any time before the index date and categorized it as 1–2, 3–10, 11–50, >50 prescriptions. Note that we did not restrict cumulative use to prescriptions recorded within 2 years before the index date because we intended to estimate the potential maximum effects of extensive opioid use. We also categorized exposure by individual drug type based on the last opioid prescribed before the index date. If someone had different types of opioids prescribed simultaneously for the last prescription, or if the supply of a second opioid overlapped his/her last opioid prescription, he/she was classified as an opioid polytherapy user.

We further grouped current opioid users by cumulative use, i.e. the number of opioid prescriptions received any time before the index date (1–2, 3–10, 11–50, >50 prescriptions), and by the last opioid type prescribed before the index date.

Covariates

We collected many covariates including lifestyle risk factors, co-morbidities, concomitant medications, pain symptoms and trauma before the index date. Lifestyle risk factors included smoking status (never, current, former, unknown) and body mass index (BMI, <18.5, 18.5–24.9, 25.0–29.9, ≥30.0 kg m−2, missing) recorded closest to and before the index date; co-morbidities included a history of hyperlipidaemia, untreated hypertension, peripheral vascular disease, asthma, chronic obstructive pulmonary disease, osteoarthritis, rheumatoid arthritis, systemic lupus erythematosus, multiple sclerosis, fibromyalgia, Parkinson disease, depression and psoriasis any time before the index date; concomitant medications included low-dose aspirin (defined as ≤ 325 mg per day), lipid modifying therapy (i.e. statins, fibrates, etc.), steroids, antidepressants, antipsychotics, antiepileptics, nonsteroidal anti-inflammatory drugs (NSAIDs), other analgesics (i.e. acetaminophen, and other acetylsalicylic acid), oral contraceptives, hormone replacement therapy and testosterone therapy recorded within 90 days before the index date; and pain symptoms and trauma included neuralgia, headache, abdominal and pelvic pain, musculoskeletal pain, other pain, noncardiac surgery and injury (mainly fracture) recorded within 90 days before the index date. All co-morbidities, concomitant medications except low-dose aspirin, and pain symptoms and trauma were coded as dichotomous variables. For those with concomitant low-dose aspirin use within 90 days before the index date, based on the number of prescriptions they received within 2 years before the index date we classified them into 1–9 and ≥10 prescriptions groups. We also collected the number of GP visits within 2 years before the index date (<10, 10–19, 20–39, ≥40) and years of medical history recorded before the index date (<5, 5–10, >10 years). These are listed in Table 1.

Table 1. Characteristics of cases and controls
 CasesControls
Characteristicsn = 11,693 (%)n = 44,897 (%)
  1. SD, standard deviation; GP, general practice.

  2. a

    Associated with the risk of myocardial infarction in a conditional logistic regression including all variables listed in this table (P < 0.05).

  3. b

    All associated with either opioid exposure timing of use or cumulative use, or both amongst controls except untreated hypertension, systemic lupus erythematosus, psoriasis and testosterone therapy (< 0.05).

  4. c

    Assessed the number of prescriptions recorded within 2 years before the index date.

Age, mean (SD), year61.8 (11.2)61.6 (11.2)
Male sex8052 (68.9)30,833 (68.7)
Calendar period of index date
1990–19942170 (18.6)8357 (18.6)
1995–19993775 (32.3)14,541 (32.4)
2000–20043651 (31.2)14,038 (31.3)
2005–20082097 (17.9)7961 (17.7)
Smokinga,b
Never3354 (28.7)19,489 (43.4)
Current4515 (38.6)10,460 (23.3)
Former1794 (15.3)6823 (15.2)
Unknown2030 (17.4)8125 (18.1)
Body mass indexa,b (kg m−2)
Low weight (<18.5)137 (1.2)524 (1.2)
Normal (18.5–24.9)3021 (25.8)12,965 (28.9)
Overweight (25.0–29.9)3706 (31.7)13,573 (30.2)
Obese (≥ 30.0)1614 (13.8)5057 (11.3)
Missing3215 (27.5)12,778 (28.5)
History of co-morbidities recorded any time before index dateb
Hyperlipidaemiaa980 (8.4)2384 (5.3)
Untreated hypertensiona681 (5.8)489 (1.1)
Peripheral vascular diseasea381 (3.3)662 (1.5)
Asthma1410 (12.1)5168 (11.5)
Chronic obstructive pulmonary disease595 (5.1)1626 (3.6)
Osteoarthritis2929 (25.0)10,885 (24.2)
Rheumatoid arthritisa379 (3.2)795 (1.8)
Systemic lupus erythematosusa37 (0.3)60 (0.1)
Multiple sclerosis51 (0.4)150 (0.3)
Fibromyalgia127 (1.1)501 (1.1)
Parkinson diseasea39 (0.3)244 (0.5)
Depression2122 (18.1)7208 (16.1)
Psoriasis487 (4.2)1642 (3.7)
Concomitant medications recorded within 90 days before index dateb
Low-dose aspirina,c  
1–9 prescriptions321 (2.8)401 (0.9)
≥ 10 prescriptions214 (1.8)555 (1.2)
Lipid modifying therapya433 (3.7)927 (2.1)
Steroidsa1133 (9.7)3259 (7.3)
Antidepressants964 (8.2)2930 (6.5)
Antipsychotics326 (2.8)992 (2.2)
Antiepilepticsa201 (1.7)805 (1.8)
Nonsteroidal anti-inflammatory drugsa1863 (15.9)5768 (12.8)
Other analgesics460 (3.9)1499 (3.3)
Oral contraceptives50 (0.4)188 (0.4)
Hormone replacement therapy242 (2.1)895 (2.0)
Testosterone therapy2 (<0.1)22 (<0.1)
Pain symptoms and trauma recorded within 90 days before index dateb
Neuralgia29 (0.2)61 (0.1)
Headache111 (0.9)339 (0.8)
Abdominal and pelvic paina342 (2.9)793 (1.8)
Musculoskeletal pain917 (7.8)3057 (6.8)
Other paina243 (2.1)573 (1.3)
Noncardiac surgery704 (6.0)2337 (5.2)
Injury184 (1.6)711 (1.6)
Number of GP visits recorded within 2 years before index datea,b
Mean (SD)21.0 (19.9)18.4 (17.3)
< 104030 (34.5)17,379 (38.7)
10–192704 (23.1)10,899 (24.3)
20–393304 (28.3)11,845 (26.4)
≥401655 (14.2)4774 (10.6)
Years of medical history recorded before index datea,b
Mean (SD)8.1 (4.3)8.1 (4.4)
<53055 (26.1)11,617 (25.9)
5–105271 (45.1)20,039 (44.6)
>103367 (28.8)13,241 (29.5)
Statistical analysis

We conducted conditional logistic regression analyses to estimate the multivariate adjusted odds ratios (ORs) of MI and 95% confidence intervals (CIs) amongst those with opioid exposure compared to nonusers. We evaluated each covariate for potential confounding by separately evaluating the associations with opioid exposure and MI, and then examining the strength of potential confounding by assessing the effect of each confounder on the crude association between opioid exposure and MI, one confounder at a time. None of the covariates affected the crude association by 10% or more. Despite this, we retained all confounders that were associated with both opioid exposure and MI, because each of them made a small but nonzero contribution to the association of opioid exposure with MI. We also controlled for new versus prevalent opioid use in the analyses to minimize the impact of including prevalent users. Amongst current users we conducted stratified analyses to examine whether the associations between each category of opioid exposure and MI were homogeneous across gender. Finally, we examined whether the association between opioid timing of use and risk of MI was modified by age (18–50, 51–60, 61–70, and 71–80 years), calendar year of the index date (1990–1994, 1995–1999, 2000–2004, and 2005–2008) and other common risk factors for MI amongst men and women, separately. To evaluate whether the stratified factors were effect modifiers we constructed multivariate models that included interaction terms for each separately with the opioid exposure variable.

We performed four sensitivity analyses to examine whether the primary findings were robust. First, we repeated the analyses restricted to the sets where both cases and matched controls were new opioid users. Second, because the effects of cumulative opioid use on risk of MI may be diluted towards the null because of extended gaps (>90 days) within prescriptions, we redefined current cumulative use based on continuous use with no gaps of >90 days between any consecutive opioid prescriptions and re-examined the effects. Third, we repeated analyses restricted to the sets where both cases and matched controls had the same length of medical history to estimate the potential impact of differential availability of confounding information due to varying lengths of medical history recorded before the index date amongst cases and controls. Finally, we conducted stratified analyses to examine whether the associations between each category of opioid exposure and risk of MI were homogeneous across different lengths of medical history recorded before the index date (<5, 5–10, and >10 years).

All analyses were conducted using SAS version 9.2 (SAS Institute, Cary, NC).

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conflict of interest statement
  8. Acknowledgements
  9. References
  10. Appendix

We identified 11 693 incident MI cases to which we matched 44 897 controls. Cases had a mean age of 62 years and around 70% were men. Cases and controls had a similar distribution of length of medical history recorded before the index date with an average of 8 years for each. Cases were more likely than controls to be current smokers, to have a BMI ≥25 kg m−2, to have more co-morbidities and concomitant medications and to have more pain symptoms. Correspondingly, cases had more GP visits than controls within 2 years before the index date. See Table 1.

After controlling for potential confounders as well as new versus prevalent opioid use the OR for current opioid use was 1.28 (95% CI: 1.19–1.37) compared with nonuse, which decreased with increasing time from last opioid prescription to the index date and reduced to the null for past use. Compared with nonuse, any cumulative use of 11–50 prescriptions was associated with a 1.38-fold (95% CI: 1.28–1.49) risk of MI, and extensive use of >50 prescriptions yielded a similar effect (adjusted OR = 1.25, 95% CI: 1.11–1.40). Use of meperidine and morphine was strongly associated with increased risks for MI compared to nonuse: adjusted ORs were 2.15 (95% CI: 1.24–3.74) and 1.71 (95% CI: 1.09–2.68) respectively. Polytherapy use was also associated with a small increased risk (adjusted OR = 1.46, 95% CI: 1.22–1.76). See Table 2.

Table 2. Risk of myocardial infarction associated with opioid timing of use, cumulative use and individual opioid type
 CasesControls  
 = 11,693 (%)= 44,897 (%)OR (95% CI)Adjusted ORa (95% CI)
  1. OR, odds ratio; CI, confidence interval.

  2. a

    Adjusted for smoking, body mass index, number of general practice visits, years of medical history, opioid new versus prevalent use, co-morbidities (hyperlipidaemia, peripheral vascular disease, rheumatoid arthritis, Parkinson disease), concomitant medications (aspirin, lipid modifying therapy, steroids, antiepileptics, nonsteroidal anti-inflammatory drugs), abdominal and pelvic pain and other pain in addition to matching variables

Nonuse6678 (57.1)28,598 (63.7)ReferenceReference
Timing of use
Current (≤30 days)1788 (15.3)4922 (11.0)1.59 (1.49–1.69)1.28 (1.19–1.37)
Recent (31–365 days)2135 (18.3)7139 (15.9)1.30 (1.23–1.37)1.17 (1.10–1.24)
Past (366–730 days)1092 (9.3)4238 (9.4)1.11 (1.04–1.20)1.06 (0.98–1.14)
Cumulative use, No. of prescriptions
1–21462 (12.5)5542 (12.3)1.14 (1.07–1.21)1.10 (1.03–1.18)
3–101452 (12.4)5148 (11.5)1.23 (1.15–1.31)1.09 (1.02–1.17)
11–501525 (13.0)4039 (9.0)1.66 (1.55–1.77)1.38 (1.28–1.49)
>50576 (4.9)1570 (3.5)1.63 (1.48–1.81)1.25 (1.11–1.40)
Opioid type
Buprenorphine13 (0.1)40 (<0.1)1.41 (0.75–2.63)1.18 (0.62–2.24)
Morphine33 (0.3)59 (0.1)2.39 (1.56–3.67)1.71 (1.09–2.68)
Meperidine21 (0.2)38 (<0.1)2.43 (1.43–4.15)2.15 (1.24–3.74)
Tramadol195 (1.7)593 (1.3)1.42 (1.20–1.68)1.19 (1.00–1.42)
Codeine1911 (16.3)6675 (14.9)1.23 (1.16–1.31)1.10 (1.03–1.17)
Dihydrocodeine1124 (9.6)3617 (8.1)1.34 (1.25–1.44)1.17 (1.08–1.26)
Propoxyphene1509 (12.9)4792 (10.7)1.37 (1.28–1.46)1.20 (1.12–1.29)
Meptazinol10 (<0.1)29 (<0.1)1.53 (0.74–3.13)1.47 (0.70–3.09)
Others16 (0.1)31 (<0.1)2.24 (1.22–4.10)1.54 (0.82–2.90)
Polytherapy183 (1.6)425 (0.9)1.87 (1.57–2.23)1.46 (1.22–1.76)

Although the magnitude of the association for current cumulative use of more than 10 prescriptions was similar to the magnitude for any cumulative use of more than 10 prescriptions, current use of 1–2 prescriptions was associated with a 1.37-fold risk of MI (95% CI: 1.14–1.64), in contrast to the null effect for use at any time (adjusted OR = 1.10, 95% CI: 1.03–1.18). Current users of meperidine and morphine as well as polytherapy had consistently increased risks for MI. In addition, current users of propoxyphene had a small increased risk of MI (adjusted OR = 1.37, 95% CI: 1.22–1.52). See Table 3.

Table 3. Risk of myocardial infarction associated with current opioid exposure combined with cumulative use and individual opioid type
 CasesControls  
 = 11,693(%)n = 44,897(%)OR (95% CI)Adjusted ORa (95% CI)
  1. OR, odds ratio; CI, confidence interval.

  2. a

    Adjusted for smoking, body mass index, number of general practice visits, years of medical history, opioid new versus prevalent use, co-morbidities (hyperlipidaemia, peripheral vascular disease, rheumatoid arthritis, Parkinson disease), concomitant medications (aspirin, lipid modifying therapy, steroids, antiepileptics, nonsteroidal anti-inflammatory drugs), abdominal and pelvic pain and other pain and recent/past opioid use in addition to matching variables.

Nonuse6678 (57.1)28,598 (63.7)ReferenceReference
Current cumulative use, No. of prescriptions
1–2173 (1.5)490 (1.1)1.54 (1.29–1.84)1.37 (1.14–1.64)
3–10305 (2.6)947 (2.1)1.40 (1.23–1.60)1.20 (1.05–1.38)
11–50834 (7.1)2169 (4.8)1.68 (1.54–1.83)1.34 (1.22–1.48)
>50476 (4.1)1316 (2.9)1.60 (1.43–1.79)1.19 (1.05–1.34)
Current use, opioid type
Buprenorphine3 (<0.1)14 (<0.1)0.94 (0.27–3.28)0.68 (0.19–2.41)
Morphine23 (0.2)35 (<0.1)2.84 (1.68–4.82)1.89 (1.09–3.30)
Meperidine6 (<0.1)4 (<0.1)7.11 (1.99–25.3)4.92 (1.31–18.5)
Tramadol95 (0.8)247 (0.6)1.67 (1.31–2.12)1.28 (0.99–1.65)
Codeine615 (5.3)1824 (4.1)1.47 (1.34–1.62)1.21 (1.09–1.35)
Dihydrocodeine344 (2.9)1020 (2.3)1.47 (1.30–1.67)1.18 (1.03–1.35)
Propoxyphene582 (5.0)1529 (3.4)1.67 (1.51–1.84)1.37 (1.22–1.52)
Meptazinol5 (<0.1)6 (<0.1)3.96 (1.20–13.1)3.16 (0.92–10.8)
Others9 (<0.1)15 (<0.1)2.58 (1.13–5.90)1.73 (0.74–4.05)
Polytherapy106 (0.9)228 (0.5)2.05 (1.62–2.59)1.48 (1.16–1.90)

There was a significant interaction between gender and current cumulative opioid use (P = 0.049). Compared to nonuse, current use of 1–2 opioid prescriptions was associated with a 1.53-fold risk of MI (95% CI: 1.23–1.91) amongst men, whilst there was no association amongst women (adjusted OR = 1.07, 95% CI: 0.75–1.51). Conversely, current cumulative opioid use of 11–50 prescription was associated with a slightly higher OR amongst women than men (1.52 vs. 1.23), and extensive use of >50 prescriptions was marginally associated with an increased risk of MI in women (adjusted OR = 1.30, 95% CI: 1.07–1.57), but not in men (adjusted OR = 1.12, 95% CI: 0.95–1.32). There was no significant interaction between gender and opioid type amongst current users (P = 0.21). Despite this, there was a strong association between current polytherapy use and risk of MI in women (adjusted OR = 2.15, 95% CI: 1.50–3.09), but not in men (adjusted OR = 1.05, 95% CI: 0.73–1.49). See Table 4.

Table 4. Risk of myocardial infarction associated with current opioid exposure combined with cumulative use and opioid type stratified on gender
 MaleFemale
CasesControlsAdjusted ORCasesControlsAdjusted OR
= 8052 (%)= 30,833 (%)(95% CI)a= 3641 (%)= 14,064 (%)(95% CI)a
  1. OR, odds ratio; CI, confidence interval.

  2. a

    Adjusted for smoking, body mass index, number of general practice visits, years of medical history, opioid new versus prevalent use, co-morbidities (hyperlipidaemia, peripheral vascular disease, rheumatoid arthritis, Parkinson disease), concomitant medications (aspirin, lipid modifying therapy, steroids, antiepileptics, nonsteroidal anti-inflammatory drugs), abdominal and pelvic pain and other pain and recent/past opioid use in addition to matching variables.

Nonuse4962 (61.6)20,613 (66.9)Reference1716 (47.1)7985 (56.8)Reference
Current cumulative use, No. of prescriptions
1–2128 (1.6)316 (1.0)1.53 (1.23–1.91)45 (1.2)174 (1.2)1.07 (0.75–1.51)
3–10185 (2.3)578 (1.9)1.18 (0.99–1.41)120 (3.3)369 (2.6)1.24 (0.99–1.56)
11–50445 (5.5)1253 (4.1)1.23 (1.08–1.39)389 (10.7)916 (6.5)1.52 (1.31–1.77)
>50244 (3.0)709 (2.3)1.12 (0.95–1.32)232 (6.4)607 (4.3)1.30 (1.07–1.57)
P-value for interaction0.049
Current use, opioid type
Buprenorphine2 (<0.1)12 (<0.1)0.59 (0.13–2.70)1 (<0.1)2 (<0.1)0.90 (0.08–10.2)
Morphine13 (0.2)23 (<0.1)1.86 (0.91–3.80)10 (0.3)12 (<0.1)1.91 (0.78–4.68)
Meperidine4 (<0.1)4 (<0.1)3.86 (0.92–16.2)2 (<0.1)0NA
Tramadol63 (0.8)161 (0.5)1.35 (0.99–1.84)32 (0.9)86 (0.6)1.20 (0.77–1.86)
Codeine350 (4.3)1075 (3.5)1.14 (0.99–1.31)265 (7.3)749 (5.3)1.33 (1.13–1.57)
Dihydrocodeine202 (2.5)591 (1.9)1.19 (1.00–1.42)142 (3.9)429 (3.1)1.16 (0.93–1.44)
Propoxyphene314 (3.9)835 (2.7)1.35 (1.17–1.56)268 (7.4)694 (4.9)1.40 (1.18–1.66)
Meptazinol4 (<0.1)5 (<0.1)3.05 (0.78–11.9)1 (<0.1)1 (<0.1)4.11 (0.25–66.8)
Others6 (<0.1)12 (<0.1)1.39 (0.51–3.80)3 (<0.1)3 (<0.1)3.93 (0.74–20.9)
Polytherapy44 (0.5)138 (0.4)1.05 (0.73–1.49)62 (1.7)90 (0.6)2.15 (1.50–3.09)
P-value for interaction0.21

Amongst men, there were no significant interactions with age or calendar year of index date and opioid exposure timing of use (P = 0.12 and 0.10, separately), although current use was significantly associated with a small increased risk of MI in the calendar period of 1990–1999, but not in 2000–2008. Similarly, there were no interactions with age or calendar year of index date and opioid exposure timing of use amongst women (P = 0.47 and 0.85, separately). See Fig. 1. The effects of opioid timing on risk of MI were not materially different when the data were stratified by smoking (current/former/never), BMI (<25 or ≥25 kg m−2) and history of hyperlipidaemia (yes/no) (data not shown).

image

Figure 1. Adjusted odds ratios (OR) [adjusted for smoking, body mass index, number of general practice visits, years of medical history, opioid new versus prevalent use, co-morbidities (hyperlipidaemia, peripheral vascular disease, rheumatoid arthritis, Parkinson disease), concomitant medications (aspirin, lipid modifying therapy, steroids, antiepileptics, nonsteroidal anti-inflammatory drugs), abdominal and pelvic pain and other pain and recent/past opioid use in addition to matching variables.] and 95% confidence intervals (CI) for current opioid exposure in relation to risk of myocardial infarction stratified on age and calendar year of index date amongst men and women separately

Download figure to PowerPoint

When we repeated the analyses restricted to new opioid users, we found results consistent with the main findings. See Table 2A–3A. The effects of current cumulative opioid use on the risk of MI did not materially change when we analysed the number of opioid prescriptions based on continuous use with no gaps more than 90 days (Table 4A). Nor did the results differ in the analyses restricted to cases and matched controls who had the same length of recorded medical history before the index date (Table 5A). When we conducted stratified analyses based on the length of recorded medical history before the index date, we found the risk of MI associated with current opioid use was lower amongst those with the longest history than that amongst those with the shortest history (Table 6A).

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conflict of interest statement
  8. Acknowledgements
  9. References
  10. Appendix

In this study, we found 1.3-fold adjusted ORs of MI for current opioid users overall and for those with more than 10 opioid prescriptions in comparison with nonusers amongst noncancer adults. We also found that use of morphine, meperidine, and polytherapy were each associated with an elevated risk of MI compared with no exposure.

The association between opioids and risk of MI has not been widely reported. Recently, Solomon et al. reported that compared with nonselective NSAIDs opioid use was associated with a 2.25-fold risk of MI (95% CI: 1.32–3.84) in the elderly [45]. More recently, Carman et al. reported that chronic opioid therapy (180-day cumulative use) was associated with an increased incidence rate ratio of 2.66 (95% CI: 2.30–3.08) for MI compared to the general population [46]. Compared to Solomon's and Carman's study populations, our study subjects had a higher proportion of men (16%, 40% and 69% respectively) and were relatively healthy as we restricted study subjects to those with no important risk factors for MI. There were also differences in exposure definition, comparison group selection, and covariate adjustment. Most importantly, Solomon's study addressed a different question (opioids vs. NSAIDs) to ours.

In our study, entry into the study cohort did not mark the patient's opioid initiation; rather, it was an indication that all subjects had pain symptoms at some time in their record, though it may have occurred after the index date for any individual. The rationale for this study design was to limit the effects of pain itself on the development of MI and also to optimize the exposure prevalence in the study sample (the base from which all cases and controls were selected), thus ensuring an efficient ratio of exposed to unexposed individuals. In addition, we did not restrict the study subjects to new opioid users. Although we captured more extensive opioid users who were more likely to be prevalent users due to this process, the result was that cumulative use could not be accurately determined for prevalent users. To minimize the potential impact of including prevalent use on the effects of each category of opioid exposure, we controlled opioid new and prevalent use in the analyses. More importantly, we repeated all analyses restricted to new opioid users who accounted for most of the study subjects and found results consistent with the primary findings.

The risk of MI levelled off after cumulative use of 11 or more opioid prescriptions in these data. In our study, the average length for an opioid prescription was 16 days, and use of 11 or more prescriptions was equivalent to around 6 or more month cumulative use. Although a dose-response relationship is usually considered important evidence to prove causation, we cannot rule out the possibility that long-term opioid use could significantly improve patient's function through alleviation of pain, which would in turn reduce the risk of MI. In this scenario, the covariates that reflect pain symptoms are on a causal pathway between opioid use and risk of MI. In this study when we controlled for pain symptoms, the observed ORs between opioid use and risk of MI were biased towards the null which supports the suggestion that decreased pain leads to a decreased risk of MI which in turn could underestimate the direct effect of opioids on the risk of MI [47]. It is also likely that some opioid prescriptions were separated by extended gaps that were not accounted for in our analyses, which may have biased the results towards the null if there was a continuous cumulative opioid effect. Re-examination of duration effects based on continuous use restricted to those with no gaps of >90 days, however, did not yield any material difference.

In clinical practice, morphine can be used for the treatment of MI. Although we defined the index date to eliminate this potential bias, we cannot completely rule out the possibility that morphine was used to treat prodromal pain of MI, because these early symptoms may not have been recorded in the electronic files. When we reviewed the records of all 23 current morphine users, we found seven cases who had been prescribed one single morphine prescription immediately before the index date that was likely to be treatment for prodromal pain of MI. Excluding these seven cases and their matched controls resulted in a decrease in the OR from 1.89 to 1.29, which suggests that the morphine association was at least in part, if not completely, explained by morphine use to treat prodromal pain of MI.

Meperidine, also known as pethidine, has a toxic metabolite, normeperidine, with a longer half-life than meperidine, consequently it is not recommended for use in chronic pain [48]. When we reviewed the records of all six current meperidine users, we found that four of them were acute users with only one prescription indicated for either renal colic or epigastric pain. As there were a very limited number of meperidine users in our study, this warrants further study to explore whether the strong association between use of meperidine and risk of MI is due to meperidine itself or other reasons. Although we found current propoxyphene use was associated with a small increased risk of MI, propoxyphene had been withdrawn from the UK market in 2005 [49]. More recently, it was also removed from the US market because there were significant changes to the electrical activity of the heart when propoxyphene was taken at therapeutic doses [50].

In this study, we found that the effects of current cumulative opioid use on risk of MI were significantly different between men and women as were the effects of current polytherapy use. Compared with women, men with acute coronary syndromes are significantly more likely to report chest pain or discomfort [51]. It is likely that the increased risk associated with current use of 1–2 prescriptions amongst men, but not amongst women, was due to opioid use for relief of prodromal pain from MI. Compared with men, women display greater sensitivity to pain stimuli [52]. Epidemiological studies have also confirmed that women are at increased risk for chronic pain [53-55]. So whether the differences in the effects of current cumulative use of 11 or more prescriptions as well as polytherapy use on the risk of MI between men and women are related to effect modification of gender or to other reasons needs further study.

Confounding bias is a major concern in interpreting the marginally increased risk of MI in our study. As we discuss above, the increased risk of MI associated with morphine and current use of 1–2 opioid prescriptions amongst men is likely to be explained by confounding by indication. That the effects of current opioid use varied across calendar period also suggests the possibility of confounding, partly because compared to those with index dates in the 1990s, those in the 2000s were more likely to have information on covariates because of their relatively long medical history recorded before the index date, and partly because opioids were probably prescribed for pain related to more severe conditions in the 1990s compared to the recent years due to improved awareness of under treatment of CNCP. Although the results did not materially change when we restricted the study subjects to the sets where cases and controls had the same length of medical history recorded before the index date, stratifications by length of medical history yielded a lower risk of MI amongst those with the longer history, lending some credence to this possibility.

Misclassification of exposure is another potential limitation. It is possible that we missed some use of over-the-counter opioids including codeine in combination with paracetamol or ibuprofen and dihydrocodeine in combination with paracetamol. However, this is unlikely to have had much impact on long-term users because people who receive more than minimal analgesia get their prescriptions from their GPs to avoid paying for them out of pocket. It is also possible that people who become prescription opioid abusers seek illegal opioids, which are not recorded in the GPRD. To minimize the impact of such misclassification, we excluded subjects with a personal history of alcohol and/or drug abuse which are strong predictors of opioid misuse by chronic pain patients [56]. In addition, if misclassification of exposure due to illegal usage existed in our study, it is more likely to occur in young adults than in elderly adults, because young age is another predictor of opioid misuse in chronic pain patients [56]. However, stratification by age did not yield material differences in MI risk associated with opioid use.

In summary, we observed marginally increased risks for MI in current opioid users or in those with cumulative use of more than 10 opioid prescriptions in a relatively healthy adult population. However, the results should be interpreted cautiously because these small, increased ORs could be partially or completely explained by the presence of confounding or other bias. Although the risk was greater in users of morphine and meperidine, and female polytherapy users, these findings warrant further study because of small numbers.

Conflict of interest statement

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conflict of interest statement
  8. Acknowledgements
  9. References
  10. Appendix

This study was not funded. All authors declare no competing interests.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conflict of interest statement
  8. Acknowledgements
  9. References
  10. Appendix

This work represents research conducted as part of L Li's doctoral dissertation in the Department of Epidemiology at the Boston University School of Public Health. We thank Drs. James Kaye and Tuhina Neogi for their critical review. We thank all the general practitioners who contribute information to the GPRD for their continuing effort and cooperation.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conflict of interest statement
  8. Acknowledgements
  9. References
  10. Appendix

Appendix

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conflict of interest statement
  8. Acknowledgements
  9. References
  10. Appendix
Table A1. List of study opioid analgesics
Class of OpioidsDrug Name
Pure AgonistsMorphine
Oxycodone
Dihydrocodeine
Diamorphine
Codeine
Hydromorphone
Meperidine
Fentanyl
Methadone
Propoxyphene
Tramadol
Dipipanone
Partial AgonistsBuprenorphine
Mixed Agonists- AntagonistsPentazocine
Nalbuphine
Meptazinol
OthersDextromoramide
Papaveretum
Table A2. Risk of myocardial infarction associated with opioid timing of use, cumulative use and individual opioid type, amongst new opioid users
 CasesControlsORAdjusted ORa
 n = 9173 (%)n = 29,019 (%)OR(95% CI)
  1. OR, odds ratio; CI, confidence interval.

  2. a

    Adjusted for smoking, body mass index, number of general practice visits, years of medical history, co-morbidities (hyperlipidaemia, peripheral vascular disease, rheumatoid arthritis, Parkinson disease), concomitant medications (aspirin, lipid modifying therapy, steroids, antiepileptics, nonsteroidal anti-inflammatory drugs), abdominal and pelvic pain and other pain in addition to matching variables

Nonuse5656 (61.7)19,584 (67.5)ReferenceReference
Timing of use
Current (≤30 days)1046 (11.4)2329 (8.0)1.55 (1.43–1.69)1.28 (1.17–1.40)
Recent (31–365 days)1598 (17.4)4360 (15.0)1.28 (1.19–1.36)1.16 (1.08–1.25)
Past (366–730 days)873 (9.5)2746 (9.5)1.10 (1.01–1.19)1.05 (0.96–1.14)
Cumulative use, No. of prescriptions
1–21349 (14.7)4085 (14.1)1.14 (1.06–1.22)1.10 (1.03–1.19)
3–101088 (11.9)3044 (10.5)1.25 (1.15–1.35)1.10 (1.01–1.19)
11–50861 (9.4)1799 (6.2)1.66 (1.52–1.82)1.39 (1.26–1.54)
>50219 (2.4)507 (1.7)1.55 (1.31–1.83)1.16 (0.97–1.38)
Opioid type
Buprenorphine8 (<0.1)24 (<0.1)1.12 (0.50–2.51)0.96 (0.42–2.19)
Morphine22 (0.2)31 (0.1)2.52 (1.45–4.38)2.04 (1.15–3.62)
Meperidine16 (0.2)21 (<0.1)2.65 (1.37–5.12)2.72 (1.36–5.44)
Tramadol144 (1.6)368 (1.3)1.37 (1.13–1.67)1.19 (0.97–1.46)
Codeine1424 (15.5)4003 (13.8)1.23 (1.15–1.32)1.11 (1.03–1.19)
Dihydrocodeine781 (8.5)2085 (7.2)1.29 (1.18–1.42)1.16 (1.05–1.28)
Propoxyphene992 (10.8)2667 (9.2)1.28 (1.18–1.39)1.15 (1.06–1.26)
Meptazinol6 (<0.1)17 (<0.1)1.25 (0.49–3.19)1.19 (0.45–3.15)
Others12 (0.1)20 (<0.1)2.03 (0.98–4.20)1.35 (0.62–2.95)
Polytherapy112 (1.2)199 (0.7)1.93 (1.52–2.44)1.61 (1.26–2.06)
Table A3. Risk of myocardial infarction associated with current opioid exposure combined with cumulative use and individual opioid type, amongst new opioid users
 CasesControlsORAdjusted ORa
 n (%)n (%)(95% CI)(95% CI)
  1. OR, odds ratio; CI, confidence interval.

  2. a

    Adjusted for smoking, body mass index, number of general practice visits, years of medical history, co-morbidities (hyperlipidaemia, peripheral vascular disease, rheumatoid arthritis, Parkinson disease), concomitant medications (aspirin, lipid modifying therapy, steroids, antiepileptics, nonsteroidal anti-inflammatory drugs), abdominal and pelvic pain and other pain and recent/past opioid use in addition to matching variables.

Nonuse5656 (61.7)19,584 (67.5)ReferenceReference
Current cumulative use, No. of prescriptions
1–2161 (1.8)377 (1.3)1.52 (1.25–1.83)1.36 (1.12–1.66)
3–10250 (2.7)587 (2.0)1.46 (1.25–1.70)1.25 (1.06–1.47)
11–50453 (4.9)941 (3.2)1.64 (1.46–1.85)1.34 (1.18–1.52)
>50182 (2.0)424 (1.5)1.52 (1.27–1.82)1.12 (0.92–1.35)
Current use, opioid type
Buprenorphine2 (<0.1)4 (<0.1)1.60 (0.29–8.86)1.12 (0.20–6.38)
Morphine14 (0.2)14 (<0.1)3.87 (1.84–8.14)2.73 (1.24–6.01)
Meperidine4 (<0.1)1 (0)17.5 (1.94–158)16.4 (1.71–157)
Tramadol64 (0.7)151 (0.5)1.48 (1.10–1.99)1.17 (0.85–1.60)
Codeine390 (4.3)912 (3.1)1.49 (1.31–1.68)1.24 (1.08–1.41)
Dihydrocodeine197 (2.2)471 (1.6)1.43 (1.21–1.71)1.19 (0.99–1.43)
Propoxyphene311 (3.4)680 (2.3)1.57 (1.36–1.80)1.30 (1.12–1.51)
Meptazinol3 (<0.1)5 (<0.1)2.59 (0.59–11.3)2.07 (0.46–9.30)
Others8 (<0.1)9 (<0.1)3.18 (1.20–8.41)2.15 (0.77–5.99)
Polytherapy53 (0.6)82 (0.3)2.36 (1.66–3.35)1.80 (1.25–2.59)
Table A4. Risk of myocardial infarction associated with current cumulative opioid use with no gaps of more than 90 days
 CasesControlsORAdjusted ORa
 n (%)n (%)OR(95% CI)
  1. OR, odds ratio; CI, confidence interval.

  2. a

    Adjusted for smoking, body mass index, number of general practice visits, years of medical history, opioid new versus prevalent use, co-morbidities (hyperlipidaemia, peripheral vascular disease, rheumatoid arthritis, Parkinson disease), concomitant medications (aspirin, lipid modifying therapy, steroids, antiepileptics, nonsteroidal anti-inflammatory drugs), abdominal and pelvic pain and other pain and recent/past opioid use in addition to matching variables.

Nonuse6678 (57.1)28,598 (63.7)ReferenceReference
Current cumulative use No. of prescriptions
1–2579 (5.0)1725 (3.8)1.46 (1.33–1.62)1.25 (1.13–1.39)
3–10395 (3.4)1023 (2.3)1.68 (1.49–1.90)1.36 (1.20–1.55)
11–50529 (4.5)1390 (3.1)1.67 (1.50–1.86)1.30 (1.15–1.46)
>50285 (2.4)784 (1.8)1.60 (1.39–1.84)1.20 (1.03–1.40)
Table 5A. Risk of myocardial infarction associated with opioid timing of use, cumulative use and individual opioid type, amongst cases and matched controls with the same lengths of medical history recorded before index date
 CasesControlsORAdjusted ORa
 n = 11,693 (%)n = 26,321 (%)(95% CI)(95% CI)
  1. OR, odds ratio; CI, confidence interval.

  2. a

    Adjusted for smoking, body mass index, number of general practice visits, opioid new versus prevalent use, co-morbidities (hyperlipidaemia, peripheral vascular disease, rheumatoid arthritis, Parkinson disease), concomitant medications (aspirin, lipid modifying therapy, steroids, antiepileptics, nonsteroidal anti-inflammatory drugs), abdominal and pelvic pain and other pain in addition to matching variables.

Nonuse6678 (57.1)16,552 (62.9)ReferenceReference
Timing of use
Current (≤30 days)1788 (15.3)3028 (11.5)1.59 (1.48–1.71)1.30 (1.19–1.41)
Recent (31–365 days)2135 (18.3)4300 (16.3)1.33 (1.24–1.42)1.20 (1.11–1.28)
Past (366–730 days)1092 (9.3)2441 (9.3)1.12 (1.03–1.22)1.07 (0.98–1.16)
Cumulative use, No. of prescriptions
1–21462 (12.5)3268 (12.4)1.16 (1.07–1.25)1.12 (1.04–1.21)
3–101452 (12.4)3118 (11.9)1.25 (1.16–1.35)1.12 (1.03–1.21)
11–501525 (13.0)2480 (9.4)1.65 (1.52–1.78)1.39 (1.27–1.52)
>50576 (4.9)903 (3.4)1.66 (1.47–1.87)1.27 (1.11–1.46)
Opioid type
Buprenorphine13 (0.1)33 (0.1)1.05 (0.52–2.10)0.87 (0.42–1.77)
Morphine33 (0.3)37 (0.1)2.31 (1.41–3.77)1.70 (1.02–2.85)
Meperidine21 (0.2)21 (<0.1)2.67 (1.39–5.13)2.58 (1.30–5.13)
Tramadol195 (1.7)258 (1.0)1.36 (1.09–1.68)1.13 (0.90–1.42)
Codeine1911 (16.3)3780 (14.4)1.26 (1.18–1.35)1.12 (1.04–1.21)
Dihydrocodeine1124 (9.6)2145 (8.2)1.36 (1.25–1.48)1.20 (1.09–1.32)
Propoxyphene1509 (12.9)3219 (12.2)1.37 (1.27–1.48)1.20 (1.11–1.30)
Meptazinol10 (<0.1)16 (<0.1)1.73 (0.75–3.99)1.58 (0.67–3.73)
Others16 (0.1)16 (<0.1)2.48 (1.19–5.19)1.74 (0.79–3.83)
Polytherapy183 (1.6)244 (0.9)1.94 (1.58–2.40)1.56 (1.25–1.95)
Table A6. Risk of myocardial infarction associated with opioid timing of use and stratified by length of medical history recorded before index date
 Length of medical history recorded before index date
Adjusted OR
(95% CI) a
<5 years5–10 years>10 years
  1. OR, odds ratio; CI, confidence interval.

  2. a

    Estimates were obtained from a logistic regression model including age, gender, region, calendar year, smoking, body mass index, number of general practice visits, opioid new versus prevalent use, co-morbidities (hyperlipidaemia, peripheral vascular disease, rheumatoid arthritis, Parkinson disease), concomitant medications (aspirin, lipid modifying therapy, steroids, antiepileptics, nonsteroidal anti-inflammatory drugs) and abdominal and pelvic pain and other pain.

NonuseReferenceReferenceReference
Timing of use
Current (≤30 days)1.35 (1.17–1.56)1.25 (1.13–1.39)1.14 (1.01–1.30)
Recent (31–365 days)1.22 (1.09–1.37)1.14 (1.04–1.24)1.12 (1.00–1.25)
Past (366–730 days)1.08 (0.93–1.24)1.09 (0.98–1.22)0.95 (0.82–1.09)