Musculoskeletal corticosteroid injection and risk of acute coronary syndrome: a case control study

Abstract

Background

Musculoskeletal corticosteroid injection (CSI) is a frequently used treatment, considered safe with a low incidence of minor side effects.

Objective

To investigate whether the incidence of acute coronary syndrome (ACS) is increased following corticosteroid injection for musculoskeletal conditions.

Methods

Data were reviewed from 41,276 patients aged over 40 years and hospitalised with ACS between January 2015 and December 2019. Each ACS case was allocated up to 10 control patients from their primary care clinic, matched for age and sex. The cases and controls were reviewed for orthopaedic or rheumatological consultation including a CSI procedure and occurring prior to the hospital admission date. The incidence of CSI was compared between the case and control groups.

Results

Data from a total of 413,063 patients were reviewed, 41,276 ACS cases and 371,787 controls. The mean age was 68.1, standard deviation (SD) = 13.1, 69.4% male. In the week prior to their hospital admission, 118 ACS patients were treated with CSI compared with 495 patients in the control group; odds ratio (OR) = 1.95 (1.56–2.43). In total, 98% of CSI procedures were carried out by orthopaedic specialists. An association between ACS and prior CSI was strongest in the days immediately prior to hospitalisation: OR = 3.11 (2.10–4.61) for patients who were injected 1 day before ACS. The association between ACS and CSI declined with increasing time between injection and hospital admission: at 90 days OR = 1.08 (0.98–1.18). The association remained robust when cardiovascular risk factors, history of rheumatological disease, and other co-morbidity were taken into consideration.

Conclusions

Musculoskeletal corticosteroid injection appears to substantially increase the risk of acute coronary syndrome.

Introduction

Corticosteroid injections (CSIs) continue to be a recommended treatment for osteoarthritis and other musculoskeletal conditions.^1–4 The injections are considered safe and effective, particularly in providing short-term symptom relief.^5–8 Complications such as local trauma and infection are known, but their incidence is low and there have been no reports of significant systemic side effects.^9,10

Endocrinological side effects of corticosteroids were noted shortly after they were introduced as an anti-inflammatory agent 70 years ago,¹¹ and in the following decades the possibility was raised that cardiovascular side effects could be caused by corticosteroids.¹² While cardiovascular complications have not been noted following musculoskeletal corticosteroid injections, there have been reports of an increased incidence of ischemic heart disease during and following oral corticosteroid treatment^13,14 and also when treatment by exogenous corticosteroids by any route is reviewed.¹⁵

We have previously reported a substantial increase in the incidence of acute coronary syndrome (ACS) in the week following CSI provided in a primary care orthopaedic clinic; reviewing 60,856 consultations (OR = 7.3). However, that initial pilot analysis was a limited sample including few outcome events.¹⁶ Our current study again investigates a possible connection between ACS and corticosteroid injections for musculoskeletal conditions, here we have reviewed a large group of patients hospitalised with ACS and studied whether they received injections prior to their ischemic events.

Methods

Setting and population

Israel has a national health insurance scheme that guarantees medical care to all citizens through four healthcare providers, of which Clalit Health Services (CHS) is the largest with 4,746,408 patients. CHS provides all tiers of medical care to its members, who make up over 50% of Israel’s population.¹⁷ CHS members have comparable health characteristics with those of the other three Israeli healthcare providers,¹⁸ and the CHS population has been used as a proxy for the national population in previous reports.^19–21

To evaluate the incidence of ACS following corticosteroid injection we used the centralised CHS database. A digitalised personal medical file links each patient’s individual national identity number with data from multiple sources, including the medical records of primary care clinics, community-based specialist consultations, hospital-discharge letters, laboratory results, and pharmacy medication claims.

The study population included all CHS members aged 40 years or older between January 2015 and December 2019. Members hospitalised with an ACS during the 5-year study period were counted as cases. An ACS was considered to have occurred when an appropriate diagnosis was reported in non-elective hospital admission (ICD9 codes: 410.00; 410.01; 410.10; 410.11; 410.20; 410.21; 410.30; 410.31; 410.40; 410.41; 410.50; 410.51; 410.60; 410.61; 410.80; 410.81; 410.70; 410.71; 410.90; 410.91). We randomly allocated up to 10 controls to each ACS case, matched for age (within 5-year bands), sex, and primary care clinic.

In the Clalit system, orthopaedic or rheumatology specialists can administer CSIs during community-based consultations and are reimbursed if a relevant procedure code is documented in the medical file (‘fee-for-service’). Orthopaedic consultations are a primary care service in Israel (i.e. patients may self-refer), and are easily accessible (1.34 million visits took place in 2019 with an average waiting time of 23 days). Rheumatology consultations are also available in primary care, however, require a referral from a family physician and do have longer waiting lists. Using the time of the ACS hospitalisation as the index date for each case and their controls, medical files were reviewed for orthopaedic or rheumatological consultations which included an injection procedure code and occurred prior to the index date. Patients who had claimed a prescription for Arthrease (sodium hyaluronate, hyaluronic acid, Fidia Pharma), the only non-corticosteroid product available for musculoskeletal injections in Israel during the study period, were not regarded as corticosteroid-injected patients.¹

We extracted information on sociodemographic characteristics and medication claims (i.e. purchases of prescription medications) from the CHS centralised data warehouse. Background medical conditions and cardiovascular risk factors were ascertained through information from the Clalit chronic disease register.²²

Statistical analysis

The study participants’ baseline sociodemographic and clinical characteristics were determined by using proportions for categorical variables and means and standard deviations for normally distributed continuous variables. We fitted conditional logistic regression models to assess the association between receiving a corticosteroid injection and a subsequent ACS event, within 7 days. We calculated the crude associations (implicitly adjusted to the matching variables through the use of conditional regression); then adjusted for age (included as a continuous variable), and ethnicity; and produced a final fully adjusted model which additionally accounts for background morbidity (i.e. smoking, hypertension, diabetes, hyperlipidaemia, obesity, ischemic heart disease, peripheral vascular disease, chronic renal failure, arthropathies, gout, rheumatoid arthritis, and dementia). We calculated odds for all study participants, as well as stratified on sex and age group. To explore the temporal nature of the association, we then conducted a secondary analysis, using varying periods of time between the injection and the index date (1–15, 30, 90, 180, 270, 365, and 1,826 days).

Finally, a number of sensitivity analyses were conducted to assess for potential bias: including only the first event for those with more than one ACS event during the study period; restricting the analysis to those with an ST-elevation myocardial infarction; including only those who died within 30 days of an ACS; including only those admitted for 3 days or longer after an ACS (to assure the exclusion of elective procedures); including only those who had a previous orthopaedic or rheumatologic visit during the study period, or before the index date.

All data were anonymised. Stata (version 15.1 IC) was used to perform all statistical analyses. The study was approved by the institutional ethics committee of Clalit Healthcare Services (approval 0207-17-COM2).

Results

Between 1st January 2015 and 31st December 2019, 42,397 CHS patients over 40 years of age were hospitalised with ACS. After excluding those with incomplete data or without available matches, the final analysis included 41,276 ACS case patients (97.4%) and their 371,787 matched controls. The mean age at index date was 68.1 ± 13.1 years (± standard deviation [SD]); 69.4% of patients were male. Sociodemographic and clinical characteristics of study participants are shown in Table 1.

After accounting for potential confounders in the fully adjusted model, people hospitalised for an ACS were almost twice as likely to have received a CSI within the week before their admission, compared to their respective controls (odds ratio [OR] = 1.95; 95% confidence intervals [CI] 1.56–2.43; P ≤ 0.001). In total, 98% of consultations including a CSI procedure were carried out by orthopaedic specialists, and 2% by rheumatologists. The association between CSI and ACS appears to be stronger among males (OR = 2.44; 95%CI 1.86–3.19; P ≤ 0.001) compared to females (OR = 1.28; 95%CI 0.86–1.92; P = 0.235); and was apparent across all age groups, with the exception of those over the age of 80 (OR = 1.28; 95%CI 0.86–1.92; P = 0.246) (Table 2). The mean time between receiving a corticosteroid injection and admission with acute coronary syndrome is shown in Table 3.

The association between ACS and prior CSI was strongest 1 day after the injection (OR = 3.11; 95% CI 2.10–4.61; P ≤ 0.001), and gradually decreased as the time between the injection and hospitalisation increased reaching at 90 days a non-significant odds ratio (OR = 1.08; 95%CI 0.98–1.18; P = 0.109) (Fig. 1, Supplementary Table 1).

The results remained robust in a series of sensitivity analyses, but the association between ACS and prior CSI was not apparent when the analysis was limited only to sets where the case patient died within 30 days of the ACS (n = 46) (Supplementary Table 2).

Discussion

Principal findings

In a case control analysis of 413,063 patient files, patients who had received a musculoskeletal corticosteroid injection were twice as likely to be hospitalised with an acute coronary syndrome within a week, compared to those who had not received an injection. The association was strongest 1 day after CSI and diminished gradually for 90 days. The findings continued to be consistent after accounting for various potential confounders including cardiovascular risk factors, rheumatological conditions and other co-morbidities.

Potential mechanisms and comparison to other studies

An association between corticosteroid treatment and increased incidence of cardiovascular disease has been previously reported in large observational studies examining oral^13,14 and all exogenous corticosteroid use.¹⁵ Our previous study suggested a significant increase in the incidence of ACS following corticosteroid injection for musculoskeletal conditions.¹⁶ Corticosteroids are known to affect the cardiovascular system by potentially exacerbating hypertension, hyperlipidaemia, hyperglycaemia, and coagulopathy.^23–25 Additionally, the presence of glucocorticoid receptors in cardiovascular tissues raises the possibility of a localised effect on atherosclerosis.²⁶

The physiological changes associated with systemic corticosteroid treatment have also been demonstrated after CSI, with increases in blood glucose and blood pressure in the hours and days following injections.²⁷ These changes, which typically occur soon after injection, could explain why we found that the largest increase in ACS incidence is in the days immediately following CSI.

An increased incidence of cardiovascular disease has been reported among patients with inflammatory arthropathies,^28,29 possibly associated with oral corticosteroid use.³⁰ However, a particular association between cardiovascular disease and CSI has not been reported in these population groups. Rheumatological conditions were accounted for in the statistical analysis suggesting that the increased incidence of ACS demonstrated in our study was not a result of inflammatory arthropathies.

Previous studies examining the potential side effects of corticosteroid injections for musculoskeletal conditions have generally been small and of low quality with systemic reviews including relatively few total numbers of patients.^5–7,9,10 It is therefore possible that the association between CSI and ACS demonstrated here was a result of the far larger patient population studied.

In our study, there is variation between the patient and control groups with respect to background morbidity and cardiovascular risk factors with the ACS patients having higher incidences of both. Other studies considering corticosteroid use and cardiovascular outcomes have also found differences in morbidity and risk factors between case and control groups.^13,14 While we cannot be sure of what shaped clinicians’ decisions, it may be that patients with greater morbidity were more likely to be selected for injection in our study, while healthier patients were treated with other medications or referred for surgical interventions. However, our fully adjusted statistical analysis does take into account cardiovascular risk factors and background morbidity with the association between CSI and increased ACS remaining robust.

Our study shows a relatively low overall incidence of cardiovascular risk factors and disease which reflects the known rates in the Israeli population as a whole.^31,32 Arab Israelis were considered in our statistical analysis as a separate ethnic group, as previous studies have indicated increased cardiovascular morbidity and mortality in the Arab population. This increased cardiovascular morbidity appears to be due to a higher incidence of cardiovascular risk factors likely combined with other social determinants of health.³³ However, our study did not find differences between the proportion of patients of Arab and non-Arab ethnic backgrounds in the control and ACS patient groups nor in the occurrence of ACS following CSI.

Strengths and limitations of this study

This is a large study that reviewed over 40,000 hospitalisations with ACS using a centralised database which enabled comprehensive assessment of patients’ files. The validity of the findings was increased by the universal healthcare coverage and accessible primary and secondary care in Israel which, reduced the risk of selection bias.

The limitations of this study include that we collected data from community-based orthopaedic and rheumatological consultations but did not include injections carried out by family physicians or in hospital-based clinics. However, it is worth noting that community-based specialist consultations comprise the vast majority of specialist consultations in CHS while corticosteroid injection is rarely carried out by family physicians in Israel. Any additional injections which were not captured by us were likely to have been distributed similarly between cases and controls (non-differential misclassification), potentially making our results an underestimation of the true association between CSI and ACS.

Another limitation in our data collection was the lack of uniformity of diagnoses in the orthopaedic and rheumatological consultations with 25% of consultations lacking a site-specific diagnosis. Although we have presumed that all the injections were for ‘musculoskeletal conditions’, we cannot confirm their site nor whether the injection was intra-articular or placed in soft tissue. However, we believe the magnitude of such misclassification in the context of the Israeli health system is small, and that since systemic injections were equally likely to be recorded as ICS among both cases and controls, this would lead to non-differential misclassification, making our results an underestimate of the true association.

It is possible that cardiac ischemia was misdiagnosed as a musculoskeletal condition in some of the patients in our study. However, it is unlikely that misdiagnosis accounts for all the cases reviewed here, particularly those patients with lower limb complaints. Furthermore, any such misclassification would have occurred independently of the subsequent allocation into the case or control groups, making any remaining bias non-differential.

We have presumed that injections were corticosteroids by excluding patients who purchased hyaluronic acid prior to their specialist consultation. However, information on particular corticosteroid products, their dose and whether they were combined with a non-steroid product was not included in our data set.

As this was an observational study using routinely collected data with some arbitrary definitions of variables, we assessed the sensitivity of the main results to various potential sources of bias in our supplementary analysis notably, amongst the patients who died within 30 days of admission with ACS the odds ratio of having received ICS was 0.63 (95% CI = 0.18–2.15). However, this finding represented only three deaths among the cases, whilst the total number of ACS cases in our study is over 40,000 unfortunately, we are not able to draw conclusions about this particular small group.

Implications for research and practice

This study suggests an association between receiving a musculoskeletal corticosteroid injection and suffering from an ischaemic cardiovascular event in the days and weeks after the procedure. Further research is needed to confirm the association between CSI and ACS and to identify the patients who are at particular risk. In addition, research should investigate whether all sites of injection are equally likely to trigger ischaemia and whether other injectable products pose similar dangers. In the meantime, it is perhaps necessary to reconsider corticosteroids as a ‘safe’ option, particularly for patients at higher risk for cardiovascular events.

Conclusion

Corticosteroid injections for musculoskeletal conditions appear to substantially increase the risk of acute coronary syndrome in the days and weeks following the injection. The absolute risk of ACS after receiving these injections is small. However, the risk is important given that CSI is a frequently used treatment and that ACS is a life-threatening side effect.

Author contributions

Katharine Thomas is the principal author of this study. Yochai Schonmann contributed significantly to the design, statistical analysis, and review of the study and has approved the final version.

Funding

No funding was provided for this study.

Ethical approval

The study was approved by the institutional ethics committee of Clalit Healthcare Services (approval 0207-17-COM2).

Conflict of interest

There are no known conflicts of interest in the production of this study.

Data availability

The data underlying this article are available in the article and in its online supplementary material. Further details can be shared upon request to the corresponding author.

References

Footnotes