Data collection methodology has previously been described in Steurer et al. ; we provide a brief summary. Data were collected via chart review from clinical practices in seven European countries (Austria, Belgium, Czech Republic, France, Greece, Portugal, Sweden), with a maximum of 25 subjects enrolled per center. Participating practices comprised a mix of academic and non-academic centers. Each clinical site obtained Institutional Review Board (IRB)/Independent Ethics Committee (IEC) approval as required by the local institution(s). Starting with the first dose of romiplostim, follow-up data, including routine platelet count measurements, were abstracted weekly from medical charts when available. Demographics and prior health history were also abstracted from medical charts before romiplostim initiation. In our study, all analyses were stratified by duration of ITP at romiplostim initiation according to the following categories defined by the International Working Group [7, 8]: newly diagnosed (< 3 months since diagnosis), persistent (3–12 months since diagnosis), and chronic (> 12 months since diagnosis).
As described in Steurer et al. , patients were eligible for the study if they were at least 18 years of age at romiplostim initiation, had primary ITP, and provided appropriate written informed consent. Patients were excluded from the study if, at the time of enrollment, the patient received or planned to receive pegylated recombinant human megakaryocyte growth and development factor, recombinant human thrombopoietin, or other thrombopoietin receptor agonists or related platelet products. Additionally, patients were excluded if they participated in any interventional clinical study or initiated romiplostim therapy before its commercial launch. Patients were enrolled from 2009 to 2012, with follow-up for this study ending in 2013.
The index date was defined as the first date of romiplostim treatment. Patients were followed for 24 weeks or until the end of available data before 24 weeks (e.g., consent withdrawn or lost to follow-up) for descriptive analyses and were not censored for romiplostim discontinuation. For outcome analyses, patients were followed until the earliest of: end of the 24-week follow-up, end of available data, receipt of splenectomy, initiation of eltrombopag, occurrence of platelet transfusion, or lack of certain platelet measurements (Table S1). Additionally, overall platelet response, median overall platelet response, and first platelet response by week 24 were censored at the time of receipt of rescue therapy. Patients were defined as receiving rescue therapy if the therapy was received after (1) an initial platelet response (platelet count ≥ 50 × 109/L) and subsequent loss of platelet response (platelet count ≤ 30 × 109/L); or (2) treatment switch after no platelet response was observed; or (3) if therapies were introduced for active bleeding (Table S1). Rescue therapy of any type, specifically oral steroids as rescue therapy, Rho (D) immune globulin (intravenous anti-D), intravenous immunoglobulin (IVIg), intravenous steroids, or platelet transfusions as rescue therapy were recorded for each patient.
Baseline characteristics were described at romiplostim initiation and were abstracted using all available chart history unless otherwise specified. History of bleeding was defined as any bleeding event occurring in the 6 months before the index date. Active bleeding was defined as a bleeding event within 7 days before or on the index date. Thrombotic events at any point in the available chart history and liver events within 5 years of romiplostim initiation were coded using the Medical Dictionary for Regulatory Activities (MedDRA v.17.1), a clinically validated international medical terminology dictionary–thesaurus used by regulatory authorities and the biopharmaceutical industry for data entry and retrieval, as well as adverse event classification (Tables S2 and S3, respectively). Previously developed sets of embolic and thrombotic event terms from MedDRA were used to identify thrombotic events, including 317 terms related to thrombotic events (including but not limited to myocardial infarction, stroke, and venous thromboembolism; Table S2). Similarly, a previously developed set of 350 terms related to liver events were used, including but not limited to hepatitis, hepatic cirrhosis, and liver transplant (Table S3). The ITP medications that were captured included IV anti-D, autologous hematopoietic stem cell transplantation (AHSCT), azathioprine, alemtuzumab, corticosteroids (oral and intravenous), cyclophosphamide, cyclosporine, danazol, dapsone, eltrombopag, IVIg (a mixture of antibodies prepared from donated blood), mycophenolate, rituximab, and vinca-alkaloids. Medications were considered to be baseline if the start and end date of the medication occurred before the index romiplostim date. Medications were considered concomitant if the index romiplostim date fell between or on either the start or end date of the concomitant medication. Romiplostim discontinuation was documented as the date of the last dose administered in the medical record. Patients were considered to have refractory ITP if they (1) were splenectomized prior to romiplostim initiation, or (2) had three or more ITP therapies prior to romiplostim initiation.
Definitions of each outcome listed below and criteria for censoring are included in Table S1. Platelet-based outcomes included median overall platelet count, overall platelet response, and first platelet response, all of which were censored for rescue therapy. Median overall platelet count was defined as the median of all platelet count measurements during weeks 2–24 after index date. Overall platelet response was defined as the proportion of weekly platelet count measurements ≥ 50 × 109/L during weeks 2–24 after the index date. First platelet response was defined as the first instance of a platelet count ≥ 50 × 109/L during weeks 2–24 after the index date. Patients in these analyses were censored if at least 5 continuous weeks passed with no platelet count measurement during weeks 2–24 after index date (equating to a minimum of four platelet measurements per patient over the follow-up period). Additionally, durable platelet response was defined as ≥ 75% of all recorded platelet count measurements ≥ 50 × 109/L during weeks 14–24 after the index date; rescue therapy was treated as outcome failure.
Bleeding outcomes included any bleeding event, bleeding requiring hospitalization (hospitalized bleeding), bleeding leading to emergent treatment, and specific types of bleeding (i.e., abnormal vaginal bleeding, central nervous system bleeding with neurologic symptoms, hematemesis, hematochezia, hematuria, and/or melena). Documented bleeding events (any), hospitalized bleeding, and specific types of bleeding were assessed from the day after the index date through week 24. Bleeding leading to emergent treatment was defined as having IVIg, IV anti-D, platelet transfusion, or IV steroids as a treatment within 7 days after the start of the bleeding event and was assessed from the day after the index date through week 24.
Discontinuation of concomitant ITP treatment was defined as a patient having no further administration of any type of non-romiplostim ITP medication that was present at romiplostim initiation (including cyclosporin, mycophenolate mofetil, oral steroid, IV steroid, IVIg, IV anti-D, danazol, vinca alkaloids, azathioprine, rituximab, alemtuzumab, dapsone, cyclophosphamide) during weeks 14–24 after the index date. Discontinuation of specific concomitant ITP treatment was similarly defined but focused only on oral steroids, IV steroid, IVIg, and/or IV anti-D.
All adverse drug reactions (ADRs) that occurred from the day after the index date through 24 weeks were evaluated. Thrombotic ADRs were included if they occurred from the day after the index date through 24 weeks after initiation and the event included one of the MedDRA embolic and thrombotic event terms used to identify baseline thrombotic events. Bone marrow ADRs were reported if they occurred from the day after the index date through 24 weeks after initiation and the event contained the term “myelofibrosis.”
We described the overall and stratified (on ITP duration) cohorts at romiplostim initiation. Categorical data were summarized by the number of patients in each category, while continuous data were summarized using the median unless otherwise specified. Additionally, unweighted mean dose and median platelet count were calculated for each patient over the follow-up period. For each time-to-event outcome (bleeding, splenectomy, and ADRs) and binary outcome (durable platelet response and discontinuation of concomitant ITP treatments), we estimated the cumulative incidence of the event for the single treatment group at 24 weeks using the Aalen–Johansen estimator of cumulative incidence accounting for the competing risk of death . We used inverse probability of censoring weights to account for informative censoring [10, 11], and included clinically relevant variables in the censoring models, including severity of thrombocytopenia, number of prior ITP treatments, and history of oral steroid and IVIg use. Patients with missing baseline platelet information (n = 3) were dropped from the outcome analyses. Confidence intervals for all outcomes were obtained by taking the standard deviation of the estimates from 100 nonparametric group-based bootstrap replications. Within each bootstrap replication, the censoring model was re-estimated . The Independent Advarra IRB approved this secondary analysis for NoviSci under NoviSci—01, Pharmacoepidemiology, Health Services, and Outcomes Research Using Limited and De-identified Data (Pro00020889). All analyses were conducted using the open source programming language R v.3.5.2 .