Article Text
Abstract
Background Dual lipid-lowering and anti-inflammatory properties of statins may lead to survival benefits in patients with rheumatoid arthritis (RA). However, data on this topic are limited, and the role of statins in RA remains unclear.
Objectives To examine the association of statin use with overall mortality among patients with RA in a general population context.
Methods We conducted an incident user cohort study with time-stratified propensity score matching using a UK general population database. The study population included individuals aged ≥20 years who had a diagnosis of RA and had used at least one disease-modifying antirheumatic drug (DMARD) between January 2000 and December 2012. To closely account for potential confounders, we compared propensity score matched cohorts of statin initiators and comparators (non-initiators) within 1-year cohort accrual blocks.
Results 432 deaths occurred during follow-up (mean 4.51 years) of the 2943 statin initiators for an incidence rate of 32.6/1000 person-years (PY), while the 513 deaths among 2943 matched comparators resulted in an incidence rate of 40.6/1000 PY. Baseline characteristics were well-balanced across the two groups. Statin initiation was associated with a 21% lower risk of all-cause mortality (HR=0.79, 95% CI 0.68 to 0.91). When we defined RA by its diagnosis code alone (not requiring DMARD use), the corresponding HR was 0.81 (95% CI 0.74 to 0.90).
Conclusions Statin initiation is associated with a lower risk of mortality among patients with RA. The magnitude of association is similar to that seen in previous randomised trials among the general population.
- Rheumatoid Arthritis
- Outcomes research
- Treatment
Statistics from Altmetric.com
Introduction
Rheumatoid arthritis (RA) is a chronic, inflammatory arthritis associated with systemic inflammation and characterised by substantial disability and premature mortality.1–4 This increased risk of mortality is associated with more severe disease activity and is probably due in part to higher levels of systemic inflammation in patients with RA.5–8
Hydroxymethylglutaryl-coenzyme A reductase inhibitors, or ‘statins’, reduce mortality by up to 30% in patients with coronary artery disease and congestive heart failure.9–13 Statins also confer a survival benefit in primary prevention in the general population, with two recent meta-analyses showing a 9% and 14% reduced risk of mortality associated with statin use.14 ,15 The JUPITER (Justification for the Use of Statins in Primary Prevention: an Intervention Trial Evaluating Rosuvastatin) trial demonstrated that rosuvastatin decreases the risk of cardiovascular (CV) events or CV mortality in otherwise healthy individuals who have low levels of low-density lipoprotein cholesterol but elevated high-sensitivity C reactive protein.16 Subjects treated with rosuvastatin also had a 20% reduction in all-cause mortality (p=0.02).
Statins have lipid-lowering effects, and also anti-inflammatory properties that involve regulating leucocyte-endothelial cell adhesion, reducing nitric oxide production and decreasing levels of inflammatory cytokines such as tumour necrosis factor α, interleukin 1 and interleukin 6.17 ,18 A randomised controlled trial of patients with RA found that the addition of statins to standard disease-modifying antirheumatic drug (DMARD) treatment improved clinical parameters such as Disease Activity Score and swollen joint count as well as inflammatory markers.19 The dual benefits of lipoprotein improvement and anti-inflammatory effects might be expected to confer a survival benefit on patients with RA, at least similar to, and potentially greater than, that seen among the general population without a history of cardiovascular disease (CVD) or inflammatory disease. A recent randomised trial was initiated to examine the impact of atorvastatin in preventing CV events in patients with RA (Trial of Atorvastatin for the Primary Prevention of Cardiovascular Events in Patients with Rheumatoid Arthritis (TRACE-RA)); however, the trial was terminated early owing to the low event rate.20 ,21 To help deal with this knowledge gap, we evaluated the association of statin initiation with the risk of death among individuals with RA in a general population context.
Methods
Data source
We used The Health Improvement Network (THIN), a computerised medical record database entered by general practitioners (GPs) in the UK. Data on approximately 10.2 million patients are systematically recorded and sent anonymously to THIN. About 97% of the UK population is registered with a GP,22 making THIN data representative of the general population of the UK.23 Information in THIN includes demographic information, details from GP visits, diagnoses from specialists’ visits and hospital admissions, laboratory tests, and additional information including height, weight, alcohol use and smoking status. The Read classification system is used to code specific diagnoses,24 and prescriptions written by the GP are coded in THIN using the Multilex classification system, a standard drug terminology library used in the UK that includes information on drug formulation and strength. During office visits, GPs record patient health information using a computerised system, and they receive training and feedback to ensure that data recording is accurate.25 Data from THIN therefore reflect routine medical practices in a population-based setting and have been shown to be valid for use in epidemiological and clinical studies.26
Study population
Our study population included individuals aged ≥20 years who had a recorded diagnosis of RA and use of at least one DMARD between January 2000 and December 2012 (see online supplementary file for RA diagnostic codes and DMARDs). This definition of RA has been shown to have a specificity of 96% in the UK General Practice Research Database27 and has been used to accurately identify RA cases in healthcare utilisation data.28 Study cohort members were required to have ≥1 years of enrolment with the general practice before entering the study cohort. Individuals were excluded if they were current or prior statin users or if they had incomplete records of covariates (cholesterol level, body mass index (BMI), drinking status and smoking status). A secondary definition of RA using only an RA diagnostic code without requiring the use of a DMARD was applied as a sensitivity analysis.
Propensity score matched cohorts stratified by time blocks
Confounding by indication, where baseline characteristics of compared groups differ in ways that alter outcome risk, can be a major concern in observational studies of medication effects; we therefore employed propensity score matching to deal with this threat to validity. Further, to address potential secular trends in statin use and mortality risk, cohorts were matched within the same calendar year (ie, 13 blocks from January 2000 to December 2012).
Within each annual cohort accrual block, statin initiators were defined as patients who started a statin during that year, while eligible patients who did not start a statin were their potential matches. Propensity scores (predicted probability of statin initiation) were estimated using logistic regression, separately for each cohort accrual block. For each statin initiator, we identified a propensity score matched subject who did not start a statin during the accrual block using a 5-to-1 digit ‘greedy matching’ algorithm.29 ,30 The initiation date was used as the index date for that patient, and a random date within the 1-year block was assigned as the index date for non-initiators.
The variables included in the logistic regression model comprised RA duration before the index date, sociodemographic factors, BMI, lifestyle factors, comorbidities, medication use and healthcare utilisation. Specifically, sociodemographic information included age at the index date, sex and the Townsend Deprivation Index score, an indicator of economic status.31 ,32 Lifestyle factors included smoking status and alcohol use. Comorbidities included myocardial infarction, atrial fibrillation, ischaemic heart disease, peripheral vascular disease, congestive heart failure, valvular heart disease, transient ischaemic attack, other circulatory disease, stroke, hypertension, chronic obstructive pulmonary disease, chronic kidney disease (stage ≥3), liver disease, diabetes, cancer, pneumonia or infection, dementia, depression, venous thromboembolism and varicose veins. Drugs included nitrates, angiotensin-converting enzyme inhibitors, angiotensin receptor blockers, β blockers, calcium channel blockers, aspirin, non-steroidal anti-inflammatory drugs (NSAIDs), loop diuretics, hydrochlorothiazide, thiazide-like diuretics, potassium-sparing diuretics, insulin, DMARDs, glucocorticoids, biological drugs and anticoagulants. Laboratory measurements included total cholesterol levels. Healthcare utilisation variables included the number of general practitioner visits, hospitalisations, referrals and number of drugs as assessed over the 2 years before the index date.
Statistical analysis
The outcome of interest was all-cause mortality as defined by the date of death recorded in THIN. Statin initiators and matched non-initiators began accruing risk time from the index date until death, disenrolment in THIN or the end of the study period, whichever came first. Statin initiators and non-initiators in each cohort retained their original exposure status throughout the follow-up. This approach maintains the comparability of these two exposure groups for baseline characteristics and provides conservative estimates, similar to an intention-to-treat approach used in clinical trials.
We used Cox proportional hazard models to estimate the association of statin initiation with the risk of all-cause mortality, stratified by 1-year cohort accrual blocks. Survival plots were generated as estimates of cumulative mortality to identify time trends in the occurrence of death. We repeated the same analyses limited to individuals with only a record of RA but not requiring use of a DMARD. We conducted subgroup analyses stratified by sex, age group (<60 years vs 60–69 years vs ≥70 years), the Townsend Deprivation Index score (≤2 vs >2), RA duration (≤5 years vs >5 years) and history of circulatory disease. As a sensitivity analysis, we used symmetrical trimming, with the cut-off point at the 5th and 95th centiles of the propensity score matching.33 Finally, we performed the same analysis excluding patients with a history of CVD (ie, myocardial infarction, ischaemic heart disease, transient ischaemic attack, stroke). To deal with the potential for statin initiators to discontinue the drug over time, we performed sensitivity analyses with follow-up time truncated at 1 year, 2 years and 3 years for all subjects. For all HRs, we calculated 95% CIs. All p values were two-sided. All statistical analyses were performed using SAS V.9.2 (SAS Institute Inc, Cary, North Carolina, USA).
Results
There were an average 264 statin initiators and 3495 statin non-initiators among the patients with RA who were eligible for propensity score matching and cohort inclusion in each 1-year accrual block. When statin initiators were compared with randomly chosen non-initiators using 1:1 matching, statin initiators had a higher prevalence of CVD, coronary heart disease risk factors, other comorbidities and use of CV-related and other drugs (n=3438 in each group; table 1 and see online supplementary table S1). During the follow-up of these unmatched cohorts, 532 died in the statin initiator group and 374 died in the non-initiator group (HR=1.46, 95% CI 1.28 to 1.67; see online supplementary figure S1).
In comparison, after propensity score matching, the baseline characteristics were well balanced between the two groups (n=2943 in each group; table 1 and see online supplementary table S1). The standardised differences of the baseline characteristics ranged from 0% to 3% (all p values >0.2), which were well within the often used guideline of 10%.34 Patients in the matched cohort were 34% male and had a mean age of 65 years. There were 432 deaths in the statin initiator group and 513 deaths in the non-initiator group during the follow-up period (mean 4.51 years and 4.29 years in the initiator and comparator groups, respectively; table 2). The corresponding incidence rates were 32.6 and 40.6 per 1000 person-years, respectively, and the HR for mortality associated with statin initiation was 0.79 (95% CI 0.68 to 0.91) (table 2, figure 1). The HRs in the analyses with follow-up truncated at 1 year, 2 years and 3 years were 0.65 (95% CI 0.48 to 0.89), 0.70 (95% CI 0.55 to 0.88) and 0.76 (95% CI 0.62 to 0.93), respectively (table 2).
Sensitivity analyses
When we repeated the analysis with a less restrictive definition of RA (diagnostic code only), more patients qualified (n=6401 in each group), but the association between statin initiation and mortality did not change materially (HR=0.81, 95% CI 0.74 to 0.90) (table 2). When we examined the association between statin initiation and mortality in a number of different subgroups, the inverse association with statin initiation persisted (table 3). We excluded patients with a prior history of CVD and found that statin use was still associated with a 25% reduction in all-cause mortality (HR=0.75, 95% CI 0.61 to 0.92). Finally, after we performed the 5th and 95th centile symmetry trimming of our propensity score matched cohorts, our results remained similar (HR=0.77, 95% CI 0.66 to 0.90).
Discussion
In this large cohort study of patients with RA derived from the general population, we found that statin initiation was associated with a 21% lower risk of all-cause mortality among patients with RA. This association was apparent from the first year through the subsequent years of follow-up. These associations were independent of age, sex, BMI, socioeconomic status, relevant comorbidities, CV medication use, total cholesterol levels and healthcare utilisation. Furthermore, subgroup analyses suggested no meaningful heterogeneity of the inverse association according to key covariates (age, sex, socioeconomic status, RA duration, baseline cholesterol levels and history of circulatory disease). Importantly, the magnitude of inverse association between statin use and mortality was similar to that found in many previous studies of non-RA patients.9 ,10 ,16 However, the magnitude was slightly greater than that seen in two recent meta-analyses of statin use for primary prevention in the general population.14 ,15 Although the differences were small, this finding may not be surprising as patients with RA are at a higher risk for CVD than the general population, and might benefit from the dual anti-inflammatory19 and lipid-lowering effects of statins in a way that the general population might not.35
The increased risk of CVD and premature mortality in RA, as well as the contribution of inflammation to this risk, has given rise to considerable interest in the cardioprotective role of statins in RA. To our knowledge, our study is the first to evaluate the association of statin initiation with all-cause mortality in patients with RA irrespective of whether they had a history of CVD. One previous study attempted to assess the potential survival effect of the prevalent use of statins in 430 patients with RA, and reported a 64% reduction in mortality.36 However, approaches investigating prevalent users (as opposed to incident users) are vulnerable to selecting patients who are already tolerating and benefiting from statins (ie, selection bias). Our study adopted an incident user design and an intention-to-treat analytical approach to help emulate the paradigm of randomised controlled trials. An additional study of statins in secondary prevention of CVD showed that patients with inflammatory joint disease treated with high-dose statins, including 199 patients with RA, had a similar reduction in lipid levels and CV events as those without RA.37 Given the relatively small number of patients with RA in that analysis, it remains unclear if there was sufficient power to detect a germane subgroup effect (by presence of RA), although it is interesting to note that the hazard ratio in that study was almost identical to the one in our study.
Our findings expand previous evidence for the beneficial effects of statins in RA, which have been indirectly drawn from (1) studies evaluating intermediate markers of CVD and premature mortality in RA;38–40 (2) antirheumatic and lipid findings from studies evaluating RA disease outcomes19 and (3) studies evaluating statin effects in other patient populations, such as the JUPITER trial.16 For example, statins have shown promising results including improvement of endothelial function,38 a reduction in arterial stiffness39 and regression of carotid artery plaque height40 in patients with RA. The TARA (Trial of Atorvastatin in Rheumatoid Arthritis) trial showed that RA disease activity improved significantly in the atorvastatin group compared with placebo, and total cholesterol, low-density lipoprotein cholesterol and triglyceride levels also declined substantially.19 Finally, the applicability of findings from the JUPITER trial to patients with RA has been argued on the basis that RA itself is a risk factor for CVD and early mortality.
The strengths of our study are worthy of discussion. This was a large-scale study conducted among the general population with a substantial number of statin initiators and a sufficient number of outcomes to produce meaningful estimates. The strength of a large-scale observational study such as this one is of particular importance in light of the recent TRACE-RA trial that was terminated prematurely owing to a low event rate.20 ,21 In pharmacoepidemiological studies such as ours, confounding by indication can pose a problem and bias results. Indeed, in our unmatched analysis, the baseline characteristics of patients with RA who were taking statins were quite different from those who were not taking statins; in particular, the initiators had more CVD and comorbidities and greater use of CV drugs. Thus, one would expect that statin use might actually be associated with a higher risk of mortality owing to this increased prevalence of comorbidities among statin initiators, and this was the case in the comparison of our unmatched cohorts (see online supplementary figure S1). However, after we employed propensity score matching to our cohorts (thereby balancing baseline covariate distributions), we found that statin initiation was associated with a lower risk of mortality. In addition, matching within 1-year blocks allows for changes in the relative importance of potential confounding variables to be accounted for at different times. Although matching may result in less generalisability, we were able to match nearly 90% of the statin initiators, making generalisability less of a concern.
Potential limitations of our study also deserve comment. Although our study found a lower risk of mortality associated with statin initiation among patients with RA, we were unable to examine cause-specific mortality, as these data are generally incomplete within THIN. Nevertheless, knowledge of the overall lower mortality risk associated with statin use among patients with RA is itself critically significant. We hypothesise that the lower mortality rate associated with statin use stems from the reduction of CV-specific mortality in patients with RA, and this speculation calls for future studies that examine cause-specific mortality outcomes. Because our study was a complete case analysis that excluded patients with missing values, it is possible that this might have affected the pool of statin non-users as they may be less likely to have cholesterol values recorded. However, exclusion of these patients would be unlikely to affect our results significantly and, if anything, would bias towards more conservative estimates of the role of statins. Although we do not have information about disease activity in the THIN database, it is unlikely that it would have significantly influenced a physician's decision to initiate a statin, making it improbable that disease activity would be acting as a confounder. Furthermore, our matched cohorts were well balanced for use of glucocorticoids, NSAIDs, biological agents and by definition, all patients were required to be using a DMARD, making it unlikely that there was meaningful imbalance in use of these drugs between the two groups. Finally, while our propensity score matching allowed us to control many variables, our study was observational, and thus we cannot exclude the possibility that there was residual or unknown confounding.
In conclusion, we found that statin initiation was associated with a 21% lower risk of mortality among patients with RA. The magnitude of the inverse association between statin use and mortality among patients with RA was similar to that seen in previous randomised trials among non-RA patients. While awaiting definitive randomised trial evidence on the topic (which is challenging to obtain, as shown by the recent experience of the TRACE-RA trial20 ,21), statin initiation appears to be associated with a lower risk of mortality in patients with RA.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
- Data supplement 1 - Online supplement
- Data supplement 2 - Online supplement
Footnotes
Handling editor Tore K Kvien
Contributors All authors participated in the conception, design and analyses of the study. SRS and HKC drafted the manuscript and are guarantors. All authors contributed to interpretation of the results.
Funding This work was supported in part by grants from the NIH (P60AR047785 and T32AR007258-36A1).
Competing interests None declared.
Ethics approval The Health Improvement Network (THIN) institutional review board.
Provenance and peer review Not commissioned; externally peer reviewed.