Association Between Glucocorticoid Exposure and Healthcare Expenditures for Potential Glucocorticoid-related Adverse Events in Patients with Rheumatoid Arthritis

Data source

We used US administrative healthcare claims data from the Truven Health MarketScan Commercial Claims and Medicare Supplemental Databases from January 1, 2012, to December 31, 2013. The Commercial Database contains the healthcare experience of privately insured individuals covered under a variety of fee-for-service, fully capitated, and partially capitated healthcare plans. The Medicare Supplemental Database contains the healthcare experience of retirees whose Medicare supplemental insurance is paid by employers. Medical claims are linked to prescription claims and person-level enrollment data. Data were previously collected, statistically de-identified, and fully compliant with the Health Insurance Portability and Accountability Act of 1996; therefore, approval from an institutional review board was not required in accordance with the policy of Truven Health.

Variables were measured using International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) diagnosis and procedure codes, Healthcare Common Procedure Coding System codes, and National Drug Codes (NDC), as appropriate. Diagnoses were searched for in all diagnosis positions in the administrative claims. Rule out claims, such as laboratory tests and radiology procedures, were not evaluated.

Sample selection and time periods

We identified patients with ≥ 2 nondiagnostic medical claims with an ICD-9-CM diagnosis code for RA (714.0x) from January 1 through December 31, 2012. At least 2 claims with an RA diagnosis were required to increase the likelihood of selecting true patients with RA. Patients were required to be ≥ 18 years of age, and to be continuously enrolled with pharmacy benefits from January 1, 2012, through December 31, 2013. We excluded patients with ≥ 1 nondiagnostic claim with a diagnosis of any other autoimmune disease (i.e., ankylosing spondylitis, Crohn disease, juvenile idiopathic arthritis, polyarteritis nodosa, psoriatic arthritis, ulcerative colitis, Wegener granulomatosis/granulomatosis with polyangiitis, or systemic lupus erythematosus) from January 1, 2012, through December 31, 2013. The final sample consisted of newly diagnosed and prevalent RA cases, and nonusers, new initiators, and continuing users of OGC. The study period was divided into a 12-month baseline period (January 1 to December 31, 2012) and a 12-month evaluation period (January 1 to December 31, 2013). January 1, 2013, represented the fixed index date for all patients. The index date was not based on any clinical event.

Exposures

Cumulative dose of OGC based on prescriptions during the 12-month baseline period was the exposure of interest. Intravenous glucocorticoid use was not included because the dose could not be accurately determined based on medical claims. Cumulative OGC dose was expressed as prednisone-equivalent mg using standard conversions⁹. A daily dose for a prescription claim was determined using the drug strength associated with the NDC, the daily supply and metric quantity fields from the claim, and the prednisone conversion factor. If a patient filled the prescription for the same OGC prior to the end of the previous prescription, the remaining supply from the first prescription was added to the total supply for the next prescription. If the prescription was for a different OGC, the remaining supply of the prior drug was truncated when the new prescription was filled. Daily doses < 1 mg were set to 1 mg, and daily doses > 60 mg were set to 60 mg, based on review of outliers and clinical input.

We categorized patients’ cumulative OGC doses during the baseline period into dosing levels. One level comprised patients with no OGC. The incidence of AE among this group could be considered the background rate of these conditions, unrelated to OGC exposure. Four nonzero levels were determined based on the quantiles observed in cumulative dosing distribution.

Presence of potential OGC-related AE

Binary variables were created to indicate the presence of each potential OGC-related AE based on diagnosis codes in any position (primary or nonprimary) on medical claims during the evaluation period (Supplementary Table 1, available with the online version of this article). Three bone-related AE were examined individually and as an aggregate outcome: osteoporosis, vertebral and nonvertebral fractures, and aseptic necrosis of the bone. Patients with ≥ 1 claims related to osteoporosis, Paget disease, or osteomalacia, or with a prescription claim for any bone disease-related medication (bisphosphonates, raloxifene, denosumab, teriparatide, or calcitonin) in the baseline period were not evaluated for osteoporosis during the evaluation period because osteoporosis preceded the index date. Fractures as a result of motor vehicle accidents (identified by an E code on the same claim) and open fractures (usually caused by trauma) were excluded.

Additional OGC-related AE were hospitalization for pneumonia or opportunistic infections (such as tuberculosis, candidiasis, or herpes zoster), hospitalization for myocardial infarction (MI) or stroke, type 2 diabetes, and gastrointestinal (GI) ulcer disease or GI bleed. Cataracts (excluding traumatic and congenital cataracts) were included in the secondary analysis. Patients with ≥ 1 claims for type 2 diabetes or with a pharmacy claim for any antidiabetes medication in the baseline period were not evaluated for the type 2 diabetes outcome, and patients with a pharmacy claim for medication to treat ulcers in the baseline period were not evaluated for ulcer/GI bleed because both would be considered “preexisting conditions.”

Expenditures for potential OGC-related AE

Payments on medical claims with a relevant diagnosis code for a potential AE in any position and pharmacy claims for AE-related medications during the evaluation period were considered potential AE-related expenditures. Payments comprised insurers’ payments, as well as patients’ cost-sharing. Expenditures were estimated overall and individually for each AE, and were expressed in 2013 constant dollars, adjusted using the medical care component of the Consumer Price Index.

Patient demographics and clinical characteristics

Demographic characteristics were measured at the patient’s index date: age, sex, geographic region, payer (commercial vs Medicare), and insurance plan type. We identified the use of biologic and nonbiologic DMARD in the baseline period. Medical claims for potential OGC-related AE and for AE-related medications in the baseline period were identified. Comorbidities assessed during baseline were hypothyroidism, ischemic heart disease, cancer, and asthma/chronic obstructive pulmonary disease (COPD). Additional health indices included the Deyo-Charlson Comorbidity Index (DCCI) score¹⁰, the number of 3-digit ICD-9-CM diagnosis codes in any position on nondiagnostic claims and the number of unique NDC, which measure overall health, and baseline healthcare expenditures.

Statistical analyses

We estimated the OR of experiencing potential OGC-related AE across cumulative dose categories by fitting adjusted logistic regression models. No OGC exposure was the reference. In addition to those reported here, we fit several versions of our models, which is standard practice. Assuming a false-positive rate of 10%, we arrived at an alpha of 0.0035 and compared the p values for our final models to this conservative alpha. Correspondingly, p < 0.0035 were considered statistically significant and 99.65% CI were calculated. In secondary analyses, the odds of occurrence of a cataract was estimated separately for age < 45, age 45–65, and age 65+ because ICD-9-CM diagnosis codes do not distinguish between age-related cataracts, such as nuclear and cortical cataract, versus cataracts that are more likely to be caused by OGC exposure, such as posterior subcapsular cataract¹¹.

Among patients who experienced a potential OGC-related AE, we used generalized linear models with a log link and gamma error distribution to estimate the adjusted incremental costs and 99.65% CI per patient with AE for each cumulative dose category. Patients with AE and no OGC use served as the reference group. Again, p < 0.0035 were considered statistically significant. Covariates used for adjustment in multivariable models were the following: age, sex, geographic region, insurance plan type, population density, payer, biologic and nonbiologic DMARD use during the baseline period, claim for any AE or AE-related medication during the baseline period, diagnosis of hypothyroidism, ischemic heart disease, cancer, asthma/COPD during the baseline period, DCCI, number of 3-digit ICD-9-CM diagnosis codes, number of NDC during the baseline period, and baseline healthcare expenditures.