Impact of NSAIDs on 8-year cumulative incidence of major cardiovascular events in patients with ankylosing spondylitis: a nationwide study

Abstract

Objectives

The objectives of this study were to describe the incidence of major adverse cardiovascular events (MACEs) in French patients newly benefiting from the French Long-term Illness scheme (LTI) for AS and to evaluate the effect of various treatments on the risk of MACE occurrence.

Methods

This national cohort study was based on the French national medico-administrative database SNDS containing data on hospitalization, the LTI, and outpatient care consumption. All French patients newly receiving LTI benefits for AS from 2010 to 2013 were included. The final follow-up date was 31 December 2018. The occurrences of MACEs [stroke and myocardial infarction (MI)] and comorbidities were identified from algorithms previously described in the literature. Competitive risk analysis using propensity score and inverse weighting was performed to calculate cumulative incidence functions and to determine subhazard ratios (SHRs) for the various treatments of interest.

Results

Between 2010 and 2013, 22 929 patients were included [mean age 43.0 (s.d. 13.9) years, 44.9% mal]. The 8-year cumulative incidences of MACE, stroke, and MI were 1.81% (1.61–2.05), 0.97% (0.83–1.14), and 0.85% (0.71–1.04), respectively. NSAIDs [SHR: 0.39 (0.32–0.50), P < 0.001] and anti-TNF [SHR 0.61 (0.46–0.80), P < 0.001], but not anti-IL17 [2.10 (0.79–5.57)] were associated with a lower risk of MACE occurrence.

Conclusion

MACE incidence rates at 8 years are low in patients newly benefiting from LTI for AS. Our results support the hypothesis of a protective role of NSAIDs and anti-TNF in cardiovascular risk in these patients.

Introduction

Spondyloarthritis (SpA) is one of the leading chronic inflammatory rheumatic diseases, with a prevalence of 0.3% in mainland France (and a prevalence of 0.15% for AS) [1]. Although the disease has a worse functional than vital prognosis, its extra-articular manifestations and the systemic inflammation it generates increase the occurrence of comorbidities, some of which may lead to death [2].

In particular, patients suffering from this disease seem to have increased cardiovascular morbidity and mortality, although the results of the various studies are heterogeneous [3]. A meta-analysis published in 2020 that included 16 studies published between 2011 and 2019 showed a relative risk compared with the general population of 1.52 (95% CI: 1.29, 1.80) for myocardial infarction (MI) and 1.21 (95% CI: 1.0, 1.47) for stroke [4]. However, the same study did not show an excess risk of all-cause mortality compared with the general population.

Thus, SpA, like RA, another chronic inflammatory rheumatism, constitutes an independent cardiovascular risk factor [5]. EULAR has issued recommendations on cardiovascular risk management for patients with inflammatory joint disorders [6].

The mechanisms involved are still poorly understood. Endothelial dysfunction exists similarly to diabetes or hypertension [7]. Indeed, several studies have shown alterations of the endothelial function of large vessels as well as the microcirculation, and an increased arterial stiffness in SpA [8]. These are the first steps of atherosclerotic plaque formation, which can induce ischaemic events such as MI or stroke.

SpA also leads to several specific cardiac complications, such as aortic insufficiency or conduction disorders, which may lead to heart failure or atrial fibrillation, the latter being a risk factor for stroke.

It should also be noted that the classical epidemiological risk factors (male sex, smoking, etc.) are frequent in patients with SpA, as is an association with diabetes and hypertension [9], which is likely to further increase cardiovascular risk in these patients.

In addition, drug treatments for the condition, which are necessary to achieve remission of the disease, may play a role. While studies for RA and psoriasis are numerous [10–13], there are few data on the potential cardiovascular protective effect of the main treatments used in SpA, whether symptomatic or disease-modifying. The main controversy is about the impact of NSAIDs, and this is still debated today. While this therapeutic class is known to increase the risk of MI in the general population [14], two studies have shown a reduction in this risk in SpA [15, 16], possibly due to the improvement of patient mobility and the reduction of systemic inflammation [17].

This study aimed to assess the incidence of major cardiovascular events (MACEs) in patients with SpA newly receiving Long-term Illness (LTI) benefits, and to compare the risk of MACEs in patients with different treatments after controlling for confounding factors.

Methods

Data source and study design

This cohort study is based on data from the National Healthcare Data System (SNDS), a French medico-administrative database containing administrative data, data on long-term illnesses, outpatient care and medication consumption, and hospitalizations of all beneficiaries of the French health insurance system. The French National Health Insurance provided us with access to an extraction of this database, in accordance with an expression of needs corresponding to the inclusion and non-inclusion criteria of the study. The data analysed covered the period from 1 January 2009 (i.e. at least 1 year before the inclusion date) to 31 December 2018 (i.e. at least 5 years after the inclusion date).

Data cleaning was performed prior to the statistical analysis by applying filters on the inclusion and non-inclusion criteria and deleting duplicates. The writing of this article is consistent with the REporting of studies Conducted using Observational Routinely-collected Data (RECORD) statement for observational studies using routinely collected health data [18].

Ethics statement

This study was approved by the Ethics and Scientific Committee for Health Research, Studies, and Evaluations (CESREES) and the French National Commission for Information Technology and Civil Liberties (CNIL) (authorization number: DR-2021–231). Written consent was not obtained from patients, in accordance with French law, because the study was based solely on routinely collected, anonymized medico-administrative data. Patients were informed of the study via the French Health Data Hub website and the Besançon University Hospital website.

Study population

Patients included were those newly enrolled in the LTI scheme with a severe SpA diagnosis between 1 January 2010 and 31 December 2013, of age 18 years or older, with an International Classification of Diseases (ICD)-10 code related to the enrolment of M45 (AS). The use of LTI allowed us to build a cohort of patients with a supposedly active disease; the SNDS, being a medico-administrative database, has no data on the activity of the disease. In France, because they are costly, almost all patients using biologic DMARDs (bDMARDs) and targeted synthetic DMARDs (tsDMARDs) are enrolled in the LTI scheme, which covers all the health-care costs related to the illness. Thus, all patients with active SpA despite NSAID treatment, and therefore requiring DMARDs, are theoretically newly enrolled in the LTI scheme.

Only patients covered by the general scheme of the French health insurance, i.e. most employees, students, social benefit recipients and ordinary residents (>85% of the insured), were included to limit missing data, since LTI data was not available for the other insurance schemes for the study period. Patients for whom stroke or MI was identified before the date of inclusion in the cohort were not included. We chose 2010 as the inclusion start date because LTI data in the SNDS are not reliable and/or comprehensive before that date.

Outcomes

The primary end point was the incidence rate of MACE, defined as the first occurrence of stroke or MI. These events were identified using previously described algorithms based on ICD-10 diagnoses of hospital discharge abstracts (Supplementary Table S1, available at Rheumatology online). The incidences of stroke, MI, and death were also assessed separately.

Covariates

Two socio-economic indicators were collected as proxies for social deprivation. At the individual level, indicators of assistance in accessing complementary health insurance [free complementary health coverage (CMU-C) or aid for the payment of complementary health insurance (ACS)]. The state medical aid (AME) provides emergency coverage for undocumented aliens. At the ecological level, the French social deprivation index (FDep) [19] at the scale of the city of residence was collected. Social deprivation is considered as an accumulation of material and social deprivations on a geographical scale. The score is derived from a principal component analysis of four parameters, then standardized. Its mean value is therefore 0, and the higher the value, the greater the social deprivation.

Regarding comorbidities, the existence of diabetes, hypertension, dyslipidaemia, smoking, obesity, chronic kidney disease (CKD), and depression were identified from algorithms previously described [20–25]. These were based on LTI benefits, reimbursement of drug treatments, and diagnoses or technical procedures during a hospital stay (Supplementary Table S2, available at Rheumatology online).

Regarding the treatments of interest, we collected outpatient reimbursements for NSAIDs (differentiating between COX2-selective and non-selective inhibitors) and three csDMARDs (MTX, LEF, SSZ). Outpatient reimbursements and inpatients prescriptions were recorded for bDMARDs: five anti-TNF drugs (adalimumab, certolizumab, etanercept, infliximab, golimumab) and two anti-IL17 drugs (secukinumab and ixekizumab). The treatment period was defined as the period from the date of the first reimbursement to the last available date of reimbursement.

Statistical analysis

A descriptive analysis was performed by calculating means and S.D.s for quantitative variables and frequencies and percentages for qualitative variables. Incidence rates for MACEs and their two components (stroke and MI) were calculated by computing cumulative incidence functions, considering the competing risk of death. A bivariate analysis was performed using Gray’s test for comorbidities and the Fine-Gray subdistribution hazard model with time-dependent variables considering the competing risk of death for the various treatments. To account for confounding by indication, we constructed a propensity score using a logistic regression with NSAIDs as the outcome and including in the model all covariates statistically associated with MACEs in univariate analysis, and covariates known to be associated with MACEs, even if not statistically significant in univariate analysis. Inverse probability treatment weighting was then performed using stabilized weights in a stratified analysis by NSAIDs treatment.

All tests were bilateral, with a threshold for statistical significance set at P < 0.05. All analyses were performed using SAS version 9.4 (SAS Institute Inc., Cary, North Carolina, USA).

Patient and public involvement

Patients were not involved in this study as it was based solely on medico-administrative data.

Results

A total of 117 892 observations of patients with LTI for SpA were present in the extraction provided by French Health Insurance. After removal of patients with an LTI-related ICD-10 code other than M45, of duplicates (n = 89 503), of patients enrolled in the LTI scheme before 2010 or after 2013 (n = 1304), of patients not covered by the general scheme of French Health Insurance (n = 3523), of patients aged <18 years at baseline (n = 499), and of patients for whom a history of MACEs was identified before the inclusion date (n = 134), a total of 22 929 individual patients were included. Their characteristics at baseline are described in Table 1. Mean (s.d.) follow-up time was 6.39 (1.09) years (min 0.01—max 9 years).

During follow-up, 344 MACEs (193 stroke, 151 MI) and 415 deaths occurred. The crude incidence rates were 2.35 (2.11–2.61) per 1000 person-years (PYs) for MACE, 1.34 (1.16–1.54) per 1000 PYs for stroke and 2.05 (1.83–2.30) PYs for MI. The 8-year cumulative incidence of MACEs, stroke, and MI were 1.81% (1.61–2.05), 0.97% (0.83–1.14), and 0.85% (0.71–1.04), respectively (Fig. 1). The cumulative incidence of death was 2.46% (2.20–2.74).

The univariate analysis of the levels of association between covariates and MACEs are presented in Table 2. There were not enough patients benefiting from state medical aid to estimate its impact on the occurrence of MACEs.

No socio-economic covariate was statistically associated with the occurrence of MACEs. Among comorbidities, diabetes, hypertension, dyslipidaemia, and atherosclerosis of the extremity arteries were statistically associated with the event of MACEs. Regarding treatments, NSAIDs and anti-TNF drugs were associated with a lower risk of MACEs. None of the csDMARDs were statistically associated with a risk of MACEs, and neither was anti-IL17.

After applying the stabilized propensity weighting, we obtained a pseudo-cohort in which the distribution of variables was similar, with a standardized difference of <0.1 between patients treated with NSAIDs and those not treated with NSAIDs (Supplementary Table S3, available at Rheumatology online). Analysis with inverse propensity score weighting showed a subhazard ratio of 0.39 (0.32–0.50) (P < 0.001) for NSAIDs. The stratified analysis for anti-TNF and anti-IL17 is shown in Table 3. Treatment with anti-TNF was associated with a lower subhazard ratio of MACEs for patients treated with NSAIDs compared with those not treated with NSAIDs. Anti-IL17 therapy was not related to the risk of MACE occurrence, regardless of whether or not patients were treated with NSAIDs.

Discussion

In this study, based on the medico-administrative data for 22 929 patients newly enrolled in the SpA LTI scheme, the risk of MACEs was found to be relatively low. Although direct comparison is difficult, the incidence rates appear to be similar to those for the general population [26, 27].

The incidence rate of MACEs was found to be lower in patients treated with NSAIDs, after controlling for potential confounders (using inverse stabilized weighting by propensity score). For NSAIDs, the effect was found in both COX2-selective and non-selective inhibitors. We did not find a statistically significant association between the risk of MACEs and anti-IL17 therapies. Our results support the hypothesis of a cardioprotective effect of NSAIDs, which is consistent with recent evidence [28].

We also found a lower rate of MACEs in patients treated with anti-TNF, as did an Australian study [29]. This could be explained by anti-TNF playing a therapeutic role in reducing subclinical inflammation and endothelial dysfunction.

In contrast, anti-IL17 therapies did not statistically alter the risk of MACEs. These data are consistent with a recent study based on the same database focusing on patients with PsA initiating biologic therapy or apremilast [30]. The pathophysiological explanation for this lack of association is not fully elucidated, but Th17 cells have an essential cardiovascular role, being both pro- and anti-atherogenic. However, the specific cardiovascular role of anti-IL17 therapies on the cardiovascular level is not well documented.

Among the baseline characteristics, we note there are a majority of women, whereas AS is classically a disease predominating in men. This female predominance, already known in the context of LTI in France, could be explained by the fact that disease activity of SpA is, on average, higher in women [31] and that patients with high activity will have a higher probability of benefiting from LTI.

In our study cohort, we also found relatively high prevalences of cardiovascular risk factors, including diabetes, dyslipidaemia, and hypertension, compared with prevalences in the general population. The values found in our study are somewhat lower than those reported in the literature in patients with a similar age [32], but this may be explained by a lack of sensitivity of the identification algorithms. On the other hand, there were very low rates of smoking and obesity, which can also be explained by a lack of sensitivity of the algorithms. In the first case, the algorithms only detect patients hospitalized or receiving treatment for smoking cessation. In the second case, they only detect patients hospitalized for morbid obesity or receiving surgical treatment for obesity. We also found a high prevalence of depression, consistent with what has been previously reported [33].

Regarding socio-economic environment, we found a rate of close to 60% of the patients in the study cohort were benefiting from free complementary health coverage, whereas this rate was 6.8% in the general population in 2014 [34]. We did not find similar studies reporting such a high rate of this indicator related to social benefits. This could be explained by the fact that patients with a low socio-economic environment might have a higher probability of being registered for and benefiting from LTI, since the latter offers 100% coverage of care related to the condition.

In our study, we chose to include patients based on their enrolment in LTI for SpA, assuming that this would allow us to build a cohort of subjects with active or progressive disease at inclusion, including subgroups being treated with a bDMARD or a tsDMARD. Our study found that 86% of the patients were receiving NSAIDs at inclusion, consistent with French recommendations, which state that the first-line treatment in SpA is NSAIDs, and that other treatments can only be considered if NSAIDs are ineffective [35]. About 6% of the patients had been exposed to anti-TNF drugs at inclusion. In almost half of the cases, this exposure began in the few months preceding the date of inclusion in LTI. Nevertheless, for some patients, the initial prescription could date several years before the date of admission to LTI. This could be explained by the fact that a patient can benefit from LTI for several pathologies (e.g. diabetes and SpA) and that there is a potential bias of underreporting a second LTI in the case of pre-existing benefit from a first LTI. In other studies, such as that of Pina-Vegas et al. [36], the subject identification algorithm is also based on hospital diagnoses, which do not allow assumptions to be made about disease severity.

It should be noted that there are recommendations from the Network for Better Use of National Health Data System (REDSIAM), based on a review of the literature, advising using M45 codes for AS [37]. However, these algorithms have not been evaluated, and the use of the LTI scheme for identifying these patients has not been studied. It can be assumed that there is little misdiagnosis related to registration in the LTI scheme, thanks to the use of ICD-10 codes, which offer good specificity. However, there is probably a selection phenomenon, since in some cases patients do not benefit from the LTI [e.g. when the disease is not very severe, i.e. is controlled with NSAIDs, or (in contrast) when it is very advanced], or patients refuse it (e.g. if they are covered by a mutual or a private insurance company).

Our study has several strengths: it was based on a large cohort, using a national and exhaustive database in a real-life setting. The restriction to people insured by the general health insurance scheme limited missing data. The identification of comorbidities was based on algorithms validated in the literature. To our knowledge, this is the first population-based study to investigate the effect of NSAIDs on the occurrence of MACEs. Using propensity score weighting limited confounding by indication, and our results are consistent with those of previous studies.

However, our study has several limitations: the source used did not provide data on the subjects’ clinical status, SpA activity, or biological inflammation level. The study of comorbidities was limited, and some algorithms used had low sensitivity, such as those for smoking and obesity. Finally, patients receiving any dose of NSAID were selected, whether it was being used continuously or on demand. Over-the-counter drug usage could not be taken into account because these purchases were not reimbursed and therefore not recorded in the database.

In conclusion, we found few MACEs in patients newly receiving LTI for SpA. NSAIDs and anti-TNFs may have a cardiovascular protective effect, unlike anti-IL17. These data are reassuring about the use of long-term NSAIDs in SpA.

Supplementary material

Supplementary material is available at Rheumatology online.

Data availability

All relevant data are reported in the article. Additional details can be provided by the corresponding author upon reasonable request.

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Disclosure statement: The authors have declared no conflicts of interest.

Acknowledgements

We wish to thank Mrs Robert for her help with the necessary administrative procedures to access the data.

References