Effect of disease modifying anti-rheumatic drugs on major cardiovascular events: a meta-analysis of randomized controlled trials

Disease modifying anti-rheumatic drugs (DMARDs) were developed to treat joint inflammation. There is growing evidence that anti-inflammatory drugs prevent major cardiovascular events (MACE). The aim of this systematic review and meta-analysis was to examine whether DMARDs reduce the risk of MACE. A systematic literature search was performed to identify randomized controlled trials (RCTs) testing the effect of DMARDs on cardiovascular events. The primary outcome was MACE defined as the first occurrence of non-fatal myocardial infarction (MI), non-fatal stroke or cardiovascular death. Secondary outcomes were myocardial infarction or stroke alone and all-cause mortality. Safety was assessed by fatal or life threatening infection. Meta-analyses were performed using random effect models and reported as risk ratios (RR) and 95% confidence intervals (CI). Study quality and publication bias were assessed using the Cochrane Collaboration’s tool for assessing risk of bias and funnel plots. Twelve RCTs involving 18,056 participants testing three different DMARDs subclasses (Tumor Necrosis Factor inhibitors—4 trials; Janus Kinase inhibitors—5 trials; Interleukin inhibitors—3 trials) were included. Meta-analysis suggested that none of the DMARD subclasses had any effect on MACE, MI alone, stroke alone, risk of fatal or life threatening infection or death. Risk of bias was high, low and unclear in five, six and one studies respectively. Funnel plots suggested a low possibility of publication bias. This meta-analysis suggests that DMARDs do not affect the incidence of MACE. More trials are needed for firm conclusions.

and psoriasis [13][14][15][16][17] . Prior clinical trials report both positive 6 and negative findings 18,19 on the effect of DMARDs on MACE. Further, there is no consensus on whether all DMARDs are effective in reducing MACE, with prior systematic reviews reporting both positive 20,21 and negative 22,23 findings by comparing between biologic and conventional synthetic DMARDs and also within subclasses of biologic DMARDs.
Prior meta-analyses have been restricted to specific disease populations, such as people with psoriasis 24 or rheumatoid arthritis 25 , limiting the power to test the effect of DMARDs on MACE. Previous systematic reviewers have also incorporated observational data together with findings from randomized controlled trials limiting the ability to interpret findings due to lack of comparison to placebo controlled groups 22,[25][26][27][28][29][30][31] . Furthermore, randomized trials specifically examining the effect of DMARDs on MACE have been published since these previous meta-analyses were reported; meaning that more up to date data are available but have not been systematically appraised 6,14,32,33 . In view of this, and past lack of consensus and limitations of previous systematic reviews, there is a need for an up to date comprehensive meta-analyses. This systematic review and meta-analysis aimed to provide an up to date test of whether DMARDs reduce the risk of cardiovascular events, using data from all available randomized controlled trials.

Methods
Search strategy and eligibility criteria. The systematic review was performed according to the Preferred Reporting the Items for Systematic Review and Meta-Analysis (PRISMA) statement and registered in the PROS-PERO database (Registration Number: CRD42020166140). The literature search was conducted by one author (ST). The databases PubMed, Cochrane Central Register for Controlled Trials and Scopus were searched on 12th January 2021. The search strategy included the synonyms or similar terms of "Cardiovascular events" AND "DMARDs" AND "Randomized controlled trials" (See supplementary material for full search terms). No date and language restriction were applied. This review was restricted to biologic and targeted synthetic DMARDs that target TNF-α, IL-1β, IL-12/23 and JAK1/2. To be eligible for inclusion, studies had to adopt: A randomized controlled design; test drugs in the DMARD class that target TNF, IL or JAK; include a placebo control group; and either reported the number of CV events (at least one of MACE, MI alone or stroke) or all-cause mortality. Both full text and abstracts of studies meeting entry criteria were included as long as minimum data were available either within the report, within the clinical trial registration entry or by contacting the corresponding author. Minimum data included DMARD tested, dose used, mean age of patients in control and intervention groups, duration of treatment and reported number of MACE, MI, stroke or death in both intervention and control groups. Exclusion criteria included trials adopting an open label or non-randomized design and studies where minimum data were not available. Included articles were identified by one author (ST) and reviewed by two more authors (JP, SMK) to confirm that they met entry criteria. Any discrepancies were resolved by discussion with another author (JM).
Data extraction and outcomes. Data were extracted on a customized spreadsheet by four authors (ST, JP, SMK, JG). Any inconsistencies were resolved through discussion with another independent author (JM). The primary outcome of this study was MACE incidence, defined as the first occurrence of any non-fatal MI or non-fatal stroke or cardiovascular death. Death occurring due to MI, stroke or any other cardiovascular related conditions was defined as cardiovascular death. Secondary outcomes were the incidence of MI, stroke or cardiovascular death alone, or all-cause mortality. Safety was assessed through the incidence of fatal or life threatening infection or infections requiring hospital admission. Outcome data were reported relative to the number of patients in whom results were reported for. The following information was also extracted from the included trials: Primary diagnosis of the population, cardiovascular risk factors, sample size for both control and intervention groups, name, dose and type of DMARD drug tested, entry and exclusion criteria for the trial, duration of treatment period, age, sex, history of MI and stroke and use of DMARDs other than the drug being tested (including TNF-α or receptor inhibitors, IL-1β or IL-12/23 inhibitors and JAK-1/2 inhibitors) at baseline. Quality assessment. Three authors independently assessed the risk of bias of all included studies using the Cochrane Collaboration's tool which assessed key aspects of the reports including guideline definitions for included patient groups, random sequence generation, blinding of participants and assessors, statistical sample size estimate, reporting the number of patients who completed the study, percentage of patients who had incomplete data, a clear primary outcome, intent-to-treat analysis and other biases 34 . Any inconsistencies were resolved through discussion between the authors until a consensus was reached. Each aspect was rated as low, high or unclear risk of bias. Each trial was then given an overall rating of risk of bias. If any aspect of the trial was considered high (or unclear) risk of bias this overall rating was given. A low risk of bias rating required all aspects to be given this assessment (Supplementary Table 14). Data analysis. Meta-analyses were performed for any of the primary or secondary outcomes where data were available from at least three trials. The main analyses pooled data from individual DMARD classes, specifically, TNF-α inhibitors, JAK-1/2 inhibitors or IL-1β, IL-12/23 inhibitors, where data for one or more outcomes were available from a minimum of three trials. In trials that had no events in either one of the groups, 'zero' event was imputed by adding one event equally to all groups in the study. Due to the small number of eligible studies, meta-analyses were performed using the inverse variance method with random effects models and applied Sidik-Jonkman method with Hartung-Knapp modification to estimate the between-study variance (tau2) 35 . This estimator uses the Q-profile method to provide a conservative and broader CI to minimize the risk of false positive results. Random effects analyses were preselected with an anticipation of high heterogeneity between studies. Statistical heterogeneity between studies was assessed using the I 2 statistic and interpreted as low (0 to  36 . The contribution of each study to the pooled estimates of outcome measures were assessed by performing leave-one-out sensitivity analyses where individual studies were excluded to inspect if they influenced the overall estimated effects. Publication bias was assessed by funnel plots comparing the overall effect size of each study to the standard error of its log-transformed effect 37 . Meta-analysis was conducted using 'meta' package and sensitivity analysis was performed using 'dmetar' package of R software version 3.4.4. All statistical tests were two-sided and p value of ≤ 0.05 was considered significant.

Results
Included trials and participants. The literature search identified 4043 unique records of which 12 trials including a total of 18,056 participants ultimately met the inclusion criteria ( Fig. 1 15 and MI patients were included in trials testing Canakinumab 6 and Anakinra 44 . The medical management of the included participants, such as prescription of statins, anti-platelet and anti-hypertensive medications, was poorly and variably reported ( Table 1). The follow-up periods in the trials ranged between 12 and 54 weeks (Table 1). Eligibility criteria, follow-up and study centers are shown in Supplementary Table 2. The outcomes of included studies are shown in Table 2.

MI alone. Six included trials reported the incidence of MI alone. Three trials testing Certolizumab pegol,
Baricitinib and Anakinra reported no significant effect on the incidence of MI 40,42,44 . Two trials reported that briakinumab and baricitinib increased the incidence of MI 15,41 . One trial reported that canakinumab significantly reduced the incidence of MI 6 .
A meta-analysis of three trials testing the effect of JAK inhibitors in 1660 participants suggested no significant effect on the risk of MI (RR = 0.63; 95% CI 0.09, 4.53) 14,41,42 . A meta-analysis of three trials testing the effect of IL inhibitors in 11,625 participants suggested no significant effect on the risk of MI (RR = 0.88; 95% CI 0.58, 1.33) 6,15,44 (Fig. 3). Sensitivity analyses suggested that the results were robust (Supplementary Tables 5, 6). Funnel plots suggested no publication bias in trials testing JAK inhibitors, but was noted to be asymmetrical in the meta-analysis on IL inhibitors ( Supplementary Fig. 2). Insufficient trials testing TNF inhibitors reported MI alone to qualify for meta-analysis.
Stroke alone. Eight of the trials testing DMARDs and all reported no significant effect on the incidence of stroke alone 6,14,15,33,[40][41][42]44 . Meta-analysis of five trials testing JAK inhibitors in 2819 participants (RR = 0.58; 95% CI 0.22, 1.50) 14,33,41-43 and three trials testing IL inhibitors in 11,625 participants (RR = 0.93; 95% CI 0.78, 1.09) 6,15,44 suggested that DMARDs had no significant effect on the relative risk of stroke (Fig. 4). Sensitivity analyses and funnel plots showed that the findings were consistent and removal of individual trials did not change the overall results (Supplementary Fig. 3 and Supplementary Tables 7 and 8). Insufficient trials testing TNF inhibitors reported the outcome of stroke alone to qualify for a meta-analysis.

Cardiovascular death alone.
A total of three trials reported cardiovascular death alone 6,15,39 . One trial tested TNF inhibitors in 535 participants and reported 1 and 3 cardiovascular deaths in participants' allocated placebo and etanercept respectively 39 . Two trials tested IL inhibitors in 11,526 participants and reported 182, 319 and 1 cardiovascular deaths in participants' allocated placebo, canakinumab and briakinumab respectively 6,15 . Insufficient trials reported this outcome to qualify for meta-analyses.

Author, Year
Group/dose (no. of patients)

Previous history of CHF (%)
Use

Discussion.
The main finding of this meta-analysis was that the three DMARD drug classes tested had no significant effect on the incidence of MACE, MI alone, stroke alone and mortality and the findings were robust in sensitivity analyses. One of the included trials, testing the IL inhibitor Canakinumab, reported a significant reduction in MACE but this finding could not be replicated in the meta-analysis which included all IL inhibitor trials identified 6 . It should be acknowledged however that the analyses are limited by the number of available trials and heterogeneity of the included populations and further trials are needed for robust conclusions. It was not possible to perform meta-analyses for the outcome of cardiovascular related mortality alone as this was infrequently reported. The recent European League Against Rheumatism (EULAR) task force recommendations suggested that there was no increase in MACE or cancer but an increased risk of infection with the use of biological (b)DMARDs as compared with conventional synthetic (cs)DMARDs 45 . The meta-analysis suggested that IL inhibitors did not have any effect on the incidence of MACE or the risk of serious or severe infections similar to the current study. These findings suggest that DMARDs may be safe to use, although longer trials are needed. Many of the included trials were very short meaning conclusions on safety should be guarded.
Most of the included trials did not examine cost effectiveness. Data from the Finnish registry suggested that bDMARDs were not cost-effective due to a high incremental cost effectiveness ratio (ICER) 46 . It should be noted that the cost effectiveness were calculated for the treatment of rheumatoid arthritis patients. This is therefore not relevant to its effects in reducing MACE. A previous systematic review suggested that TNF inhibitors were cost effective when applying a threshold of 50,000 Euro per Quality adjusted life years (QALY) as a treatment of joint inflammation 47 . Further, cost-effectiveness of bDMARDs have been shown to be superior to csDMARDs when analysed using the number needed to treat (NNT) method 48 . With respect to the MACE, the CANTOS trial was the only large scale trial focused on secondary prevention of recurrent cardiovascular events and was estimated www.nature.com/scientificreports/ to increase QALY by 0.13 for a cost of $832,000, yielding an ICER of 6.4 million per QALY gained. Based on that trial Canakinumab was not considered cost-effective. Targeted use of DMARDs, particularly if they can be offered at a lower cost might lead to a cost-effective treatment but this remains to be shown 49 .
The findings of this meta-analysis should be interpreted after considering a number of limitations. The number of included trials was small limiting the statistical power. Five trials had a high risk of bias 18,33,39,40,43 due to failure to follow intent-to-treat principles and other methodological weaknesses. The participants included were noted to have heterogeneous diagnoses which complicates the interpretation of the findings. Only one risk of bias assessment was used based on the Cochrane collaborative tool. The Grading of Recommendations, Assessment, Development and Evaluation (GRADE) criteria were not reported. Of particular note the majority of the trials did not report the medical treatments, such as anti-platelet, lipid modifying, and anti-hypertensive medications, prescribed to participants and thus it was not possible to investigate if any variation in this may have impacted on the results.
In conclusion, this meta-analysis suggested that the three DMARD sub-classes tested did not significantly affect the risk of MACE. Further trials in people at high risk of cardiovascular events are needed before firm conclusions can be drawn.