In this study, we used research data created by the Health Insurance Review & Assessment Service (HIRA) of the Republic of Korea (M20180306928). The views expressed are those of the authors and do not necessarily reflect those of the HIRA and the Ministry of Health and Welfare. The HIRA database consists of administrative claims data submitted by healthcare service providers to the HIRA for reimbursement for a service provided to patients. As almost all South Korean citizens (approximately 50 million) are covered by this insurance system, it could be representative of the South Korean population [13,14,15].
The database contains not only demographic information and healthcare utilization information, including surgical, medical, pharmaceutical services, and clinical diagnostic codes [International Classification of Disease, 10th Revision (ICD-10)] of each patient but also the medical costs of each service availed [13, 16]. Thus, this database is useful for estimating the treatment patterns and changes in healthcare costs. In a previous validation study, which compared diagnoses derived from the database of medical records of patients, the overall positive predictive value of the diagnostic codes was approximately 76.5–83.6% .
In this study, we investigated the treatment patterns among patients with AS administered various TNFis. We defined the index date for each study subject as the date of the first occurrence of claim for the administration of any TNFi of interest between July 1, 2013, and June 30, 2015, which was considered the identification period. The index TNFi was defined as the TNFi prescribed on the index date. We defined a 2-year period after the index date as the follow-up period, and a 1-year period before the index date as the pre-index period. The baseline characteristics of the study population were assessed in the pre-index period; concomitant medication was assessed in the follow-up period (Fig. 1).
This study involved patients who (1) had at least one insurance claim with both an AS diagnosis code (ICD-10: M45) as a primary or secondary diagnosis code (up to 5) and the codes for Rare and Intractable Diseases registration (RID) program for AS (V140) during the identification period; (2) were prescribed the TNFi of interest (etanercept, adalimumab, golimumab, or infliximab) within the same claim; and (3) were aged ≥ 18 years at the index date. Patients could be assessed using the RID program only if they met the modified New York criteria, which is a system that offers special cost–benefit to patients with a chronic illness. Thus, both codes (M45 and V140) can ensure that the included patients had AS . To ensure the inclusion of TNFi-naive patients, patients prescribed any biologics (adalimumab, etanercept, golimumab, and infliximab, including infliximab biosimilars) within the pre-index period were excluded. By excluding these patients, we could avoid the influence of the use of previous biologics. Similarly, the patients who had at least one insurance claim including a diagnosis code for other indications of the index TNFi (rheumatoid arthritis: M05, M06; psoriatic arthritis: M07.0, M07.1, M07.2, M07.3, M09.0, L40.5; Crohn’s disease: K50; psoriasis: L40; ulcerative colitis: K51; Behçet's disease: M35.2, N77.8; hidradenitis suppurativa: L73.2) during the preceding year were excluded from the study. Patients who could not be followed up for 2 years were excluded from this study. For infliximab, the follow-up period started from the fourth infusion of infliximab, because an induction period during the first 6 weeks (weeks 0, 2, and 6) is required for the administration of infliximab . Patients were classified into each TNFi group according to their index TNFi.
This study was performed in accordance with the Helsinki Declaration of 1964 and its later amendments. The study protocol was approved by the institutional review boards of the Inje University Seoul Paik Hospital (2018-07-011), Sungkyunkwan University (SKKU 2018-02-006), and Hanyang University (2017-11-020). Informed consent was waived by the institutional review boards because this was a retrospective study that used data from insurance claims.
The primary outcomes were the treatment patterns of TNFis during the follow-up periods in each treatment group. The treatment patterns were classified into two large groups: continuation and discontinuation. Discontinuation was further subdivided into three categories: switching, restarting, and stopping. Switching was defined as a change from index TNFi to another TNFi within 90 days from the last exposure period of the index TNFi. The TNFi exposure period was assumed to be 7 days for each administration of 50 mg etanercept, 14 days for each administration of 40 mg adalimumab, 28 days for each administration of 50 mg golimumab, and 56 days for each administration of infliximab . Restarting was defined as prescribing index TNFi or other TNFi after more than 90 days from the last exposure period. Stopping was defined as no prescription of the index TNFi after the last exposure period without switching or restarting. Patients who had been continuously prescribed the index TNFi without any defined discontinuation event during the 2-year follow-up period were assigned to the continuation group.
A dose reduction of TNFi was evaluated in patients listed in the continuation group in each TNFi group. To estimate dose reduction, the weekly dose was calculated as the prescribed dosage divided by the time gap between consecutive prescriptions. The patients who were prescribed a weekly dose of 50% or lower of the recommended weekly dose of the index TNFi in a series of ≥ 2 claims during the follow-up period were classified into the dose reduction group, whereas the others were classified into the maintenance group . We referred to the drug approval documents of the Ministry of Food and Drug Safety to define the recommended weekly doses (adalimumab: 40 mg every 14 days; etanercept: 50 mg every 7 days; golimumab: 50 mg every 28 days; infliximab: 5 mg/kg every 56 days). The recommended weekly dose of infliximab was estimated from the average weight for men (71 kg) and women (57 kg) in South Korea . The amount of reduced dose was calculated from the amount of reduction in the weekly dose, which was obtained by subtracting the first reduced weekly dose from the recommended weekly dose. The reduced percentage was estimated as the amount of reduced dose divided by the recommended weekly dose. Time to dose reduction was calculated as the number of days from the index date to the date of dose reduction (Table 1). A sensitivity analysis was performed by changing the definition of dose reduction to the following: a weekly dose of 30% or lower of the recommended weekly dose of the index TNFi in a series of ≥ 2 claims during the follow-up period.
To compare AS-related healthcare costs between the dose reduction and dose maintenance groups for each index TNFi, we estimated the costs from the patients’ claims that had an AS diagnosis code (M45) and a specialized diagnosis code (V140) during the follow-up period. The AS-related healthcare cost was subdivided into TNFi costs, other drug costs, and other medical costs. Other drug costs represented the costs of drugs except for TNFis, and other medical costs represented the AS-related healthcare costs, except for the costs of the drugs. All costs are expressed in U.S. dollars at the May 2019 exchange rate (1206 won = 1 US dollar).
The patient baseline characteristics and treatment patterns are presented as number and percentage of patients in each TNFi group. Chi-square test or Fisher’s exact test was used to test for significant differences between groups.
In the dose reduction analysis via index TNFi, dose reduction occurrences are presented as the number and percentage of patients. The amount of dose reduction (mg) and time to dose reduction (days) are presented as mean (standard deviation) and median (interquartile range). For the results presented as mean and standard deviation, analysis of variance was used to compare the differences among the TNFi groups. Kruskal–Wallis test was used to compare the differences presented as median and interquartile range.
Healthcare costs are presented as regression-adjusted predicted least-square mean cost from the generalized linear model (GLM) with gamma distribution and log link function. The GLM was used to estimate healthcare costs because healthcare cost data usually have a skewed distribution . Demographic covariates including sex, age, and Charlson comorbidity index were adjusted by the model.
All statistical tests were two-sided with an alpha level of 0.05. The data were analyzed using SAS software (version 9.4, SAS Institute, Cary, NC, USA).