Patients and study design
The PSOLAR patient populations and study design have been previously reported [18]. Briefly, PSOLAR enrolled adult patients (> 18 years) receiving, or eligible to receive, treatment with biologics and/or conventional systemic agents for psoriasis [18, 19]. As of August 23, 2015, PSOLAR was fully enrolled with 12,090 patients.
In the current report, all patients had psoriasis, including a subset of which self-reported a diagnosis of PsA (N = 4315) [20]. Of the patients self-reporting PsA, a subset self-reported that their healthcare provider established a diagnosis of PsA (N = 1719), but with no further confirmation by the investigator [20]. Patients were enrolled at international dermatology sites including academic centers, hospitals, and community practices [18]. Demographic and psoriasis disease characteristics, medical, social, and family histories, and previous medication use were collected at each site using electronic case report forms (eCRF). Data were collected at site visits at baseline and every-6-months, except medical, social, and family histories, which were only collected at baseline.
PSOLAR is conducted in accordance with the International Conference on Harmonizing guidelines on Good Clinical Practices and the Declaration of Helsinki. An institutional review board or ethics committee (Goodwyn Institutional Review Board and Ontario Institutional Review Board) approved the registry protocol. All patients provided written informed consent. This study adheres to the STROBE guidelines for observational studies.
In this report, patients were stratified by exposure to specific biologics or non-biologics. Specifically, patients were evaluated using the following treatment cohorts: ustekinumab, infliximab, adalimumab, etanercept, all other biologics, non-biologic/methotrexate (MTX) and non-biologic/non-MTX. The specific tumor necrosis factor (TNF) inhibitors evaluated in this analysis (infliximab, adalimumab, and etanercept) were selected because they were the most prevalently used in the PSOLAR registry. The “all other biologics” cohort included patients receiving a biologic indicated for PsA (e.g., golimumab, certolizumab, secukinumab), that was being studied in PsA but not yet approved at the time (e.g., abatacept, brodalumab, ixekizumab), or that is no longer commercially available (e.g., efalizumab, alefacept). The “all other biologics” cohort has been removed from all tables presented in this report due to low numbers, but is still included in the total. Patients had received one of these agents when registered in PSOLAR or began treatment after PSOLAR registration. These patients may also have been exposed in the past to a different biologic, topical psoriasis therapy, or phototherapy. Previous use or current exposure was allowed for MTX, but not for other systemic immunomodulators. Non-biologic/non-MTX could have included, but were not limited to, cyclosporine, tacrolimus, mycophenolate mofetil, azathioprine, oral corticosteroids, and psoralen plus ultraviolet A, or ultraviolet B phototherapy. Patients in cohorts other than the non-biologic/non-MTX group were excluded if oral corticosteroids were used. This was done to allow for more fair comparisons and to prevent confounding the SI based on the risk of the biologic itself. For each comparison, the MTX-exposed patient cohort served as the reference group.
The cohorts described above were analyzed in three populations as follows. The overall population included prevalent (previous) users of biologics (receiving a biologic at the time of registry enrollment) and incident (new) users of biologics (starting a biologic at the time of, or after, enrollment). The incident biologic user subpopulation included only new users of a biologic at or after enrollment as defined above. Prior usage of any biologic other than the newly initiated biologic was allowed among incident users. The bionaive subpopulation included users who had never been exposed to any biologic. The primary analysis for this study was based on the overall population.
Index date (initiation of follow-up to record person years) for the cohorts was the date of registration into PSOLAR for those on current therapy OR the start date of therapy after registration for new users. Incident rates of SI were calculated based upon person years of exposure. Follow-up began on the index date until: 1) death; 2) withdrawal from the registry; 3) last data cut (August 2016); 4) 91 days after discontinuation of treatment; or 5) switching of treatment. Patients could not qualify for other treatment cohorts if switching of therapy occurred.
The exposure period for the new biologic users (incident user and bionaive subpopulations) started at the date of the first cohort-defining biologic dose on registry and ended at the earlier of 90 days after the last cohort-defining biologic dose, the date of switching to a different biologic, the date of starting other systemic immunomodulators, discontinuation from the registry, registry data cut (August 23, 2016), or death. The exposure period for prevalent biologic users started at registry enrollment and ended as described for the new biologic users. Only the first cohort-defining biologic was considered; biologic usage after switching was excluded.
Both non-biologic/MTX and non-biologic/non-MTX cohorts were defined the same for both prevalent and incident users. Exposure for the non-biologic/MTX was calculated from the later date of the enrollment date and the first MTX dose date and ended as the earlier of 90 days after the last MTX dose, the date of switching to other systemic immunomodulators, discontinuation from the registry, registry data cut (August 23, 2016), or death.
Exposure for the non-biologic/non-MTX starts from the later time of enrollment date and the start date of other immunomodulators on registry and ended at the earlier of 90 days after the last cohort defining drug dose, the date of switching to MTX, discontinuation from the registry, registry data cut (August 23, 2016), or death.
Outcome measures
The outcome included the number of SI events, defined as serious adverse events (SAEs) classified as “Infections and Infestations” based on the Medical Dictionary for Regulatory Activities (MedDRA) coding system. SAEs are defined as any undesirable experience associated with the use of a medical product in a patient that may result in death, be life-threatening, lead to hospitalization or prolonged hospitalization, disability or permanent damage, congenital anomaly or birth defect, or require intervention to prevent permanent impairment or damage. All AEs were identified by the site investigator, collected through eCRF, and reviewed. Verbatim terms provided by sites are coded to MedDRA terms and are subject to verification by site monitors. All SAEs regardless of exposure are subject to verification by study personnel. Incomplete reports are queried for context to verify the event. This is performed by the study nurse and physician initially and cross-checked for completeness by the sponsor’s medical monitor. Queries are generated if the rationale/context of the reported event term is unclear. Despite monitoring of serious event documentation, source documents may not necessarily be obtainable for all diagnoses. Nonetheless, reporting, categorization and definition of serious infections is designed to be non-differential across all treatment cohorts since the study support team performs uniform follow-up on all cases. History of infections was defined as infections requiring a prescription medication within 3 years of enrollment. Rates of multiple infections and modeled analyses were based on time to first infection.
Statistical analysis
All patients from the PSOLAR registry at the time of this analysis (August 23, 2016) who self-reported having PsA were categorized into a treatment cohort based on the definitions mentioned above while excluding patients violating the rules. Baseline demographics and disease characteristics were summarized for each cohort group without imputation for missing values. Cumulative incidence rates for SI were calculated based on 100 patient-years. Cox regression models were used to identify predictors to first SI event and to estimate adjusted hazard ratios and 95% confidence intervals (CI) of SI event. Pre-determined covariates at baseline (defined as the last non-missing value that is closest and prior to/on the cohort start date) were included in the multivariate model (Fig. 1). In case of missing continuous covariates, they were imputed with population mean and the imputed values were used in the model.