The IQVIA longitudinal claims database used in this study drew on Canadian private and public prescription claims. Specifically, the study used prescription claims from three databases: IQVIA Private Drug Plan database (PDP), Ontario Drug Benefit database (ODB), and Régie de l'Assurance Maladie du Québec (RAMQ) [6, 13, 14]. Patients from all three sources were combined for analysis. IQVIA’s PDP database comprised drug benefit claims paid by a host of private insurers including the top ten private insurance carriers, third-party administrators, and benefit plan managers and represented approximately 83% of the total private (direct-pay) business in Canada. The ODB and RAMQ are public drug plans administered in Ontario and Quebec, respectively. The ODB database contains 100% of the fully adjudicated dispensed prescription claims from the ODB program at the anonymized patient level. IQVIA’s RAMQ dataset contains a sample of dispensed prescription claims collected at the anonymized patient level. Overall, this database represents most patients on private plans in Canada.

All sources of data were actively managed and quality controlled, and captured patient demographic characteristics, specific drugs dispensed, dosage, quantity dispensed, number of days’ supply, service date, pharmacy location, cost, payer, and prescribing physician information. Additionally, all three sources had full capture of prescriptions for a patient if the patient stayed in the drug plan. This study was performed in accordance with the Helsinki Declaration of 1964 and its later amendments. In accordance with our institutional policies and those of Canada, ethics approval and informed consent were not required since this is a prescription claims-level study using anonymized data.

It is important to note that these administrative claims data do not include diagnosis. Therefore, consistent with methodologies previously adopted, indication was inferred using an established rules-based algorithm [6, 13, 15]. As shown in Table 1, the study included seven target drugs of interest for IA indications: five biologic DMARDs administered by subcutaneous injection (abatacept, golimumab, secukinumab, tocilizumab, and ustekinumab) and two oral targeted-synthetic DMARDs (apremilast and tofacitinib). A selection period from 1 January 2015 to 28 February 2019, where all these drugs were listed in public formularies, was utilized for patient selection based on eligibility criteria. The target drug that a patient initiated during the selection period was defined as the “index drug.” Similarly, the date a patient first started the index drug treatment during the selection period was defined as the “index date.” A 12-month look-back period from index date was applied to ensure patients were naïve to their respective target drug. The look-back period was also used to assess patients’ history of the target drugs and other demographic characteristics. Patients in each group were analyzed for up to 12 months from index date or until they discontinued or switched to other in-market drugs. This “analysis period” was used to analyze adherence, persistence, dosing interval, and dose optimization and concomitant medication use. Owing to the consideration of the loading phase in some drugs’ treatment regimens, the first 28 days post-index were excluded from the adherence and dosing analysis for all index drugs. Additionally, a look-forward period of 3 months following the analysis period was applied to ensure that patients were still active in the drug plan and that discontinuation was not due to missing data. For each included patient, the study period was a combination of look-back period, analysis period, and look-forward period.

Patients were included in the study if they met all entry criteria: (1) inferred indication of IA based on IQVIA’s proprietary indication algorithm; (2) naïve to the target drug during the look-back period; (3) had at least one claim of one of the target drugs within the selection period; (4) were ≥ 18 years of age on date of index claim; (5) active in the drug plan within the 12-month look-back period, 12-month analysis period, and 3-month look-forward period. Patients were excluded from the study if they met at least one of the following criteria: (1) had any claim of the index drug in the 12-month look-back period; (2) patients 18–24 years of age in Ontario with index date or analysis period between 1 January 2018 and 31 March 2019 as patients were transitioning from a pediatric to an early adult program (OHIP+). Patients did not need to be continuously acquiring the index drug throughout the 12-month analysis period to be included. A patient could be indexed on only one target drug.

The following endpoints based on each study objective were measured for each target drug separately: the proportion of patients (%) adherent to each index drug within the 12-month analysis period and the proportion of patients (%) persistent at 12 months; the dosing interval of each index drug was also analyzed in the following way to explore the frequency of dose escalation and the proportion of patients (%) whose index drug dose was increased (i.e., dose escalated) within the 12-month analysis period was assessed. Additionally, the mean and median escalated dose, dose difference, and time to dose escalation within the 12-month analysis period were assessed for all dose-escalated patients. For concomitant medication use, the presence, mean number of claims per month, and total days’ supply per month of OCS use for adherent and nonadherent groups within the 12-month analysis period was assessed, as well as the presence, mean number of claims per month, and total days’ supply per month of OCS and cDMARDs (azathioprine, cyclosporine, hydroxychloroquine, leflunomide, methotrexate, and sulfasalazine) use within each target drug group pre-index and post-index.

Biologic experience was defined as the presence of any in-market biologics in the look-back period. The proposed prior biologic experience groupings were biologic-experienced and biologic-naïve, where biologic-naïve was defined as having no claims for in-market biologics in the look-back period. Biologic-experienced patients were further categorized by biologic tier on the basis of whether one, two or three or more different biologic claims were reported in the 12-month look-back period. Polypharmacy was defined as the number of concomitant medication classes [number of EphATC (level 2) classes] patients took in the look-back period and reported in numerical categories (0–3, 4–6, 7–9, 10–12, 13+).

Adherence, commonly defined as the extent to which patients take their medications as prescribed [16], was measured using the medication possession ratio (MPR). MPR was defined as the proportion of days’ supply obtained during one episode of medication use, calculated by dividing the aggregated number of days’ supply obtained during the episode by the length of the episode, excluding the last prescription fill [6, 17]. MPR was calculated for the treatment period in which patients are continuously taking the index drug with no evidence of switching to other in-market products.

$$\mathrm=\frac}\text }}$$

Days’ supply was calculated from the number of units in a prescription and length of treatment per unit listed in the Health Canada product monograph. Patients who scored ≥ 80% MPR were considered adherent as per the commonly accepted operational definition of adherence [6, 17, 18].

Patients were considered persistent until they discontinued the index drug. Discontinuation was defined as the end of treatment of the index drug. A patient is flagged as discontinued on the index drug if (1) there is a treatment gap of more than 90 days between the end of supply of a claim and the start of a subsequent claim during the analysis period and look-forward period or (2) the patient switches to other in-market products, whichever occurred first [17].

Dosing interval was defined as the average days between units for all target drugs with SC injections (Table 1). It was estimated by taking total days on therapy and dividing by the number of units the patients received.

For the oral drugs, tofacitinib and apremilast, dosing interval was defined as average number of tablets per week.

$$\text=7 \times \frac}\left(}\right)}$$

Dose escalation was defined as having an average weekly dose that was at least 20% higher than the recommended dose [13, 19]. This could be achieved by either increasing the dose and/or shortening the dosing interval. Recommended dose(s) were based on Health Canada product monographs (Table 1). Owing to the consideration of the loading phase in some drugs’ treatment regimens, the first 28 days post-index were excluded from the analysis for all index drugs.

$$\text=7 \times \frac \times \text}}$$

Continuous variables were reported using mean, standard deviation, median, and interquartile range. Categorical variables were reported using counts and proportions. All descriptive endpoints were reported separately for each target drug. For each of the endpoints, statistical modeling was used to assess each target drug in a single model. Any patients that switched between formulations of their index drug were considered censored at their first switch of formulations. All analyses were independently performed by IQVIA using SAS 9.4 (SAS Institute, Cary, NC, USA). There was no imputation of missing data given that patients were not expected to be lost to follow-up as the 3-month look-forward ensured patients were still active in the drug plan. Where the assumptions underlying any statistical modeling were violated, appropriate alternatives were employed.

Adherence across target drugs was analyzed using a multivariate logistic regression treating adherence as a binary variable (adherent, nonadherent). Expected predictor variables in the model included target drug, age, gender, province (or group of provinces), prior biologic experience, polypharmacy, and time on treatment. Persistence was analyzed using a multivariate Cox proportional hazards model for the time to target drug treatment discontinuation. Expected predictor variables in the model included target drug, age, gender, province (or group of provinces), prior biologic experience, and polypharmacy. Patients were not expected to be lost to follow-up owing to activity requirements in the look-forward period. The proportion of patients with escalated dose of index drugs were compared across target drugs using a chi-square test. Since it is not possible to collect disease activity measures in these claims databases, post-index usage of OCS was used as a surrogate for disease flare (poor disease outcomes), and discontinuation of DMARDs and OCS was used as surrogate marker of positive disease outcomes. Concomitant OCS use was compared between adherent and nonadherent groups for each target drug. No comparisons were made across target drugs for this endpoint. The presence or absence of OCS use was assessed using a chi-square test. The number of OCS claims per patient per month and days’ supply of OCS per month were compared across all target drugs using Kruskal–Wallis test. OCS and DMARD use were compared 12 months pre-index and up to 12 months post-index for each target drug. No comparisons were made across target drugs for this endpoint. The presence or absence of OCS and DMARD use was assessed using McNemar’s test. The number of OCS and DMARD claims per patient per month and days’ supply of OCS and DMARD per month were compared across all target drugs using Wilcoxon signed-rank test.