Relation of NSAIDs, DMARDs, and TNF Inhibitors for Ankylosing Spondylitis and Psoriatic Arthritis to Risk of Total Hip and Knee Arthroplasty

METHODS

We conducted a nested case-control study using data from the OptumLabs Data Warehouse (OLDW), which includes deidentified medical and pharmacy claims, laboratory results, and enrollment records for commercial and Medicare Advantage enrollees. The database contains longitudinal health information on enrollees and patients, representing a diverse mixture of ages, ethnicities, and geographical regions across the US. Members in the database had full insurance coverage for physician, hospital, and prescription drug services.¹⁸ A nested case-control study design was selected to maximize power, given the relatively low number of outcomes (knee and hip replacements).

Study sample. Cases were defined as subjects with AS or PsA with THA or TKA, and controls were defined as subjects with AS or PsA without such procedures. International Classification of Diseases, 9th and 10th revisions (ICD9 and ICD10) codes were used to identify adults with AS and PsA aged 18 years or older from January 1, 1994, to December 31, 2018. These codes (Supplementary Table 1, available with the online version of this article) have been validated in AS and PsA in prior studies.^19,20 Subjects were required to have medical and pharmacy coverage at least 6 months prior to AS or PsA diagnosis, and were eligible for inclusion on the date of their second AS or PsA diagnosis claim. If a subject had both AS and PsA diagnoses, we categorized them by the first diagnosis listed. THA/TKAs were defined by ICD 9/10 procedure codes and current procedural terminology (CPT) codes, and the date of first THA/TKA during the study period was defined as the index date. We made the assumption that subjects were undergoing THA/TKA for end-stage joint damage due to their AS or PsA. We did not include arthroplasty revisions in this outcome. We did not exclude subjects who had had a single hip/knee replacement, to allow for THA/TKA on both left and right sides within each subject. Additional exclusion criteria were (1) history of 2 THAs or 2 TKAs prior to AS/PsA diagnosis, and (2) a diagnosis of inflammatory bowel disease any time prior to AS/PsA diagnosis. Subjects with inflammatory bowel disease were excluded because they may have different treatment patterns than AS and PsA subjects. We excluded hip and knee fractures 30 days prior to the outcome because this is the time frame that we would expect that any THA or TKA would be related to the fracture. As part of our internal validation study, we reviewed 100 random subject profiles, and all claims listed 6 months prior to study entry to 6 months after exit of the study. We obtained claim records with the codes of interest from the OLDW through SAS (SAS Institute); the data were then exported to Excel for review. We were able to confirm appropriate identification of diseases of interest (AS or PsA), exposures (NSAID, DMARD, and/or TNFi), and outcomes (THA or TKA).

Study design. The outcome of interest was THA or TKA, as defined by procedure codes. THA or TKA cases were each matched to up to 4 controls by sex, age (± 3 yrs), SpA subtype (AS/PsA), AS/PsA diagnosis year (± 1), insurance type (commercial or Medicare Advantage), obesity (defined as BMI ≥ 30 kg/m²), and any hip or knee replacement prior to study entry. Subjects were matched using risk-set sampling. Thus, subjects could be included as controls before they became a case. We assessed AS/PsA treatment in the 6-month period prior to THA or TKA. We assessed for use of DMARDs or TNFi stratified by NSAID use. This is because those taking NSAIDs in the 6 months prior to arthroplasty may have done so due to pain from end-stage joint disease rather than as a treatment for their AS or PsA (confounding by indication). Medical therapies were identified according to records of filled prescriptions. DMARDs included apremilast, auranofin, azathioprine, baricitinib, chloroquine, cyclophosphamide, cyclosporine, hydroxychloroquine (HCQ), leflunomide, methotrexate (MTX), mycophenolate mofetil, sulfasalazine (SSZ), thiomalate, and tofacitinib. While still a subject of controversy, we included apremilast and tofacitinib in this DMARD category because they are oral small molecules, and the low frequency of use did not allow us to analyze them separately. Janus kinase inhibitors and non-TNFi biologics were not included in this study, as there was an insufficient number of subjects on these medications.

Statistical analysis. We evaluated the relation of each type of treatment (DMARDs, TNFi stratified by NSAID use) to risk of THA or TKA using conditional logistical regression. The outcome of THA or TKA was binary (yes/no) and defined as the first THA/TKA code during follow-up. Arthroplasty codes excluded revisions. After matching subjects by age, sex, disease, diagnosis year, and insurance type, we then performed adjustment for potential confounders. These included age (after matching ± 3 years for more precision, as some subjects were not perfect matches) and history of chronic kidney disease, diabetes, liver disease, peptic ulcer disease, psoriasis (for subjects with PsA only), and RA. We also adjusted for a record of erythrocyte sedimentation rate (ESR) or C-reactive protein (CRP) testing having been performed within 1 year prior to AS/PsA diagnosis, as well as at least 1 rheumatology visit in the year prior to AS/PsA diagnosis (vs no visits). These variables were used as surrogates for disease activity. Among the strata of NSAID users, NSAID use alone was the referent group, and for NSAID nonusers, the absence of any medication of interest was the referent group. Controls were not excluded on the basis of comorbidities. As a way to assess for effect modification by how much risk a subject would incur with a knee/hip arthroplasty surgery, we considered variables that may deem a subject a poor candidate for joint replacement: BMI ≥ 40 kg/m², history of joint infection, any high-risk cancer, end-stage renal disease (ESRD) requiring dialysis, and severe pulmonary disease requiring supplemental oxygen. We looked at the prevalence of these across cases and controls to identify any patterns for descriptive purposes but did not incorporate these variables into the analysis.

Sensitivity analyses. To assess the robustness of our primary results, we performed several sensitivity analyses for AS and PsA individually: models assessing age stratification (< 55 vs ≥ 55 years), THA and TKA separately, history of joint replacement prior to the study period (yes/no), 14-month exposure assessment (rather than our standard 6-month period), and ≥ 2 prescriptions during the exposure assessment (instead of 1). Last, we performed a conditional logistic regression model analysis considering each drug category as a binary variable and assessed the effect of each category after adjustment for other confounders.

Ethical approval. Since this study involved analysis of preexisting, deidentified data, it was exempt from Institutional Review Board approval.