Switch rates and total cost associated with apremilast and biologics in biologic-naive patients with psoriatic arthritis

We conducted a retrospective analysis using the IBM MarketScan^® Commercial and Medicare Supplemental databases (IBM Watson Health, MA, USA) to identify two cohorts of biologic-naive patients with PsA: patients who initiated apremilast and patients who initiated a biologic agent for the treatment of PsA between 1 January 2015 and 31 December 2016. A similar retrospective study was carried out for patients with psoriasis but without PsA who received treatment [20].

Adult patients, ≥18 years of age on the index date (defined as the date of the first claim for the treatment of interest), were eligible for inclusion if they initiated a new treatment with apremilast or a biologic agent (i.e., adalimumab, certolizumab, etanercept, golimumab, infliximab, ixekizumab, secukinumab or ustekinumab) for treatment of PsA with or without psoriasis between 1 January 2015 and 31 December 2016. Patients were required to have at least two International Classification of Diseases, Ninth or Tenth Revision, Clinical Modification (ICD-9-CM/ICD-10-CM) diagnosis codes for PsA with or without a claim for psoriasis during the 12 months before the index date. For PsA, the ICD-9-CM code was 696.0 and the ICD-10-CM codes included L40.50, L40.51, L40.52, L40.53, L40.54 and L40.59. For psoriasis, the ICD-9-CM code was 696.1 and the ICD-10-CM codes included L40.0, L40.1, L40.2, L40.3, L40.4, L40.8 and L40.9. The start date of a new treatment episode was considered to be the index date. A minimum 12-month baseline period that was free of any index agent was required to identify the baseline patient demographics and clinical characteristics, as was a minimum follow-up period of 12 months continuous enrollment to assess switch rates and costs. Patients were included if they did not have a history of biologic use in the pre-index period; however, pre-index conventional synthetic DMARDs or other agents used in PsA were permitted. All available data from the entire pre-index period were used to identify previous biologic- and psoriasis-related therapy exposure. Key exclusion criteria included other biologic-indicated autoimmune conditions (e.g., ulcerative colitis, Crohn’s disease, rheumatoid arthritis and other inflammatory polyarthropathies, ankylosing spondylitis or juvenile idiopathic arthritis), or cancer at any time during the pre-index or postindex period.

Switch rate was defined as the proportion of patients who switched to or added a new systemic treatment, either apremilast or biologic therapy (categorized into either TNFi or ILi regimens) after the initiation of the index treatment. Time to switch was defined as the time from the date of initiation of the index treatment to the date of initiation of a new treatment for those patients who did switch within the 12-month follow-up period. Adherence to index treatment was calculated utilizing the medication possession ratio (MPR) and the proportion days covered (PDC). MPR is a ratio of the number of drug doses taken to the number of doses prescribed over a certain time period, whereas PDC is the proportion of days that are covered by prescriptions [21]. Healthcare costs, which comprised both inpatient (i.e., hospital room and bed, medical professional care and procedures/tests conducted during the hospital stay) and outpatient costs (i.e., emergency room admissions, office visits, laboratory tests, diagnostics, infusions, other outpatient and outpatient pharmacy, including psoriasis/PsA-related treatments – both office-administered and outpatient pharmacy prescriptions for apremilast and biologics – and concomitant psoriasis and PsA-related medications), were based on the paid amounts of adjudicated claims, including insurer and health plan payments, as well as patient cost sharing in the form of co-payment, deductible and co-insurance. Healthcare costs reflect actual paid costs based on patient adherence to treatment and treatment switching. Total healthcare costs were defined as the total sum of healthcare costs from the initiation of treatment. Per-patient per-month (PPPM) total healthcare costs were defined as the average total monthly healthcare costs while patients remained on the index treatment and were reported for the 12-month post-index period in subgroups that switched and did not switch, and in the pre-switch and postswitch periods among the subgroup that switched.

Baseline patient demographics and clinical characteristics between the apremilast and biologic cohorts were compared using a t-test for continuous variables and a chi-square test for categorical variables. A 1:2 propensity score matching was used to adjust for possible selection bias and to maximize the number of patients available for analysis. Logistic regression was used to estimate the propensity score for individual patients in the 12-month pre-index period with the following variables: age, gender, region, payer, plan type, index year, pre-index days, Charlson Comorbidity Index score, pre-index cost and a limited number of clinical characteristics, including number of prior systemic agents, previous usage of nonsteroidal anti-inflammatory drugs or COX-2 inhibitors, previous use of corticosteroids or phototherapy, and total healthcare cost in the 12 months prior to the index period. To examine the quality of the propensity score match, standardized mean differences in matching factors between patients receiving apremilast and patients receiving other psoriasis/PsA-related biologic treatments were calculated before and after matching. A p-value ≤0.05 was considered statistically significant. The Wilcoxon rank-sum test was used to assess the cost differences between patients initiating apremilast and a biologic.

Between 1 January 2015 and 31 December 2016, a total of 88,025 patients in the IBM database initiated treatment with apremilast or a biologic agent. After the inclusion and exclusion criteria and propensity score matching were applied, 471 biologic-naive patients were included in the apremilast group and 890 biologic-naive patients were included in the biologic group; the biologic group was further divided into patients treated with a TNFi (n = 804) or an ILi (n = 86; Figure 1). With the exception of gender, the baseline characteristics were well balanced between the cohorts after propensity score matching. The mean age was similar across the three cohorts (apremilast: 50.2 years; TNFi: 50.0 years; and ILi: 50.4 years). The duration of available pre-index data was also similar across the three cohorts (apremilast: 1360 days; TNFi: 1345 days; and ILi: 1325 days). The mean Charlson Comorbidity Index scores were also similar for the three treatment cohorts (apremilast: 0.58; TNFi: 0.55; and ILi: 0.56). After propensity score matching, the ILi cohort had a lower proportion of females (41.9%; p = 0.02) compared with the apremilast cohort (55.0%) and the TNFi cohort (54.6%; Table 1). Most patients treated with a TNFi had an index treatment with either adalimumab (64.4%) or etanercept (29.0%), whereas nearly all patients treated with an ILi had an index treatment of either ustekinumab (77.9%) or secukinumab (20.9%).

Biologic-naive patients receiving apremilast had significantly lower switch rates at 12 months compared with biologic-naive patients receiving a TNFi (15.5% vs 26.6%; p < 0.0001); no significant differences were observed in the 12-month switch rates between the apremilast and ILi cohorts (15.5% vs 14.0%; p = 0.71; Figure 2). For patients who switched by month 12, those initially treated with apremilast most often switched to adalimumab (49.1%) or ustekinumab (19.3%). Patients treated with a TNFi most often switched to apremilast (23.8%), followed by etanercept (19.3%). Patients treated with an ILi most often switched to adalimumab (30.0%), followed by apremilast (20.0%; Supplementary Table 1). No significant differences were observed between treatment cohorts in days to switch at 12 months.

The mean (standard deviation) 12-month adherence rates while on the index treatment for apremilast initiators was 0.81 (0.20), compared with 0.85 (0.16) for TNFi initiators (p = 0.0176) and 0.76 (0.20) for ILi initiators (p = 0.0083) as measured by PDC. Using the MPR, the mean (standard deviation) 12-month adherence rates while on the index treatment for apremilast initiators was 0.84 (0.20), compared with 0.88 (0.16) for TNFi initiators (p = 0.0008) and 0.78 (0.20) for ILi initiators (p = 0.0074).

Switch rates were also calculated for patients with data at 24 months postindex; rates increased over time in all treatment groups and were significantly higher at all time points for patients receiving TNFi compared with patients receiving apremilast (Figure 2).

Total 12-month healthcare costs were significantly lower for biologic-naive patients initiating apremilast treatment (US$39,854) compared with a TNFi and an ILi (US$57,243 and US$65,867, respectively; both p < 0.001 vs apremilast; Table 2). The majority of total healthcare costs across all treatment groups were due to outpatient pharmacy costs, although the pharmacy costs were significantly lower for patients treated with apremilast (US$29,130) compared with TNFi and ILi (US$48,540 and US$57,038, respectively; both p < 0.0001 vs apremilast). No differences were observed in outpatient costs across the three treatment groups. Among those who did switch treatment by month 12, total healthcare costs throughout the postindex period were higher for patients receiving apremilast or a TNFi as their index treatment compared with the overall cohort (switchers plus nonswitchers) but were lower for those initiating an ILi (Figure 3). Mean total healthcare costs at 12 months among patients who switched from their index treatment were lowest for patients initiating apremilast (US$48,160) compared with a TNFi (US$65,096; p < 0.0001) or an ILi (US$60,825; p = 0.0202). Patients who did not switch at 12 months had lower total healthcare costs with apremilast (US$38,330) compared with a TNFi and an ILi (US$54,395 and US$66,685, respectively; both p < 0.0001 vs apremilast; Figure 3).

PPPM healthcare costs were significantly lower over the 12-month follow-up period for patients initiating apremilast (US$3319) versus a TNFi and an ILi (US$4768 and US$5486, respectively; both p < 0.0001 vs apremilast), and outpatient pharmacy costs were the largest driver of these differences. Patients who did not switch index treatments had significantly lower PPPM healthcare costs at 12 months when they were treated with apremilast (US$3192) compared with a TNFi and an ILi (US$4530 and US$5554, respectively; both p < 0.0001 vs apremilast). Mean PPPM healthcare costs in patients who switched treatments at 12 months were also lower with apremilast (US$4011) compared with a TNFi (US$5422; p < 0.0001) and an ILi (US$5066; p = 0.0202). Before switching treatments, patients treated with apremilast had significantly lower PPPM healthcare costs (US$2562) compared with patients treated with a TNFi and an ILi (US$4999 and US$5272, respectively; both p ≤ 0.0001 vs apremilast). However, after switching, no significant differences were observed between treatment groups in mean total PPPM healthcare costs with apremilast (US$7140) compared with a TNFi (US$9183; p = 0.8497) and an ILi (US$5782; p = 0.5873). Among patients who switched treatments, PPPM costs increased after switching for patients with apremilast or a TNFi as their index treatment, while PPPM costs were relatively unchanged for patients with an ILi as their index treatment (Supplementary Figure 1).

Several important limitations need to be noted, including those inherent to retrospective claims data analyses such as data coding limitations, data entry error and lack of clinical detail. Although our study used a large claims data source in the USA, findings are only generalizable to individuals with commercial or private Medicare supplemental health coverage. Because the study relied on administrative claims data, clinical details that may impact treatment patterns such as disease activity/severity, location of psoriatic disease or specific PsA subtype were not available for analysis, and these unknown differences between the treatment groups may have influenced the results observed in this study. This is not a limitation unique to MarketScan; information regarding disease activity/severity is unlikely to exist in any claims database. We attempted to mitigate the lack of data on disease activity/severity by matching the cohorts based on proxies for potential disease activity/severity, including the Charlson Comorbidity Index (i.e., studies have shown more comorbidity burden is associated with higher disease activity and worse disability/function [23,24]), pre-index cost, prior treatment experience (nonsteroidal anti-inflammatory drugs, steroids, phototherapy) and presence of psoriasis. Residual confounding after the propensity score matching cannot be ruled out; we could only match apremilast-treated patients with biologic-treated patients and could not match apremilast-treated patients with TNFi-treated patients and with ILi-treated patients separately because the sample size for the latter group was too small. In addition, treatment switching and adherence measures based on claims for filled prescriptions may not accurately reflect whether patients actually took the medication as prescribed. Since this database compiles claims from various different health plans, a limitation for using switch rates as a proxy for treatment success is the possibility that heterogeneity among medical policies and cost sharing may influence switch rates. These possible reasons for switches are independent of treatment effectiveness and patient satisfaction and may distort data interpretation. However, reasons for treatment switches could not be determined in this claims analysis. Finally, we could not identify prescriber specialty, unfortunately, because prescription claims in MarketScan do not provide such information.