Real-world evidence on clinical outcomes in immune thrombocytopenia treated with thrombopoietin receptor agonists

Of 671,909 patients (≥18 years) diagnosed with ITP in the EHR network on 3 April 2018, 1617 patients were on eltrombopag (with no prior romiplostim) and 1140 patients were on romiplostim (with no prior eltrombopag) (Table 1). The patient flow is displayed in Figure 1. To analyze patient outcomes from matched samples, we excluded patients with the following co-morbidities/characteristics as there were significant differences (p < 0.01) in prevalence between the two groups: acute hepatitis C, prothrombin gene mutation, systemic lupus erythematosus, essential (hemorrhagic) thrombocythemia, myelodysplastic syndrome and splenectomy post treatment. Following these exclusions, there were 1281 patients treated with eltrombopag and 930 patients treated with romiplostim available for analysis. Mean ages of patients included in the subsequent analysis of clinical outcomes and AEs were: eltrombopag, 50 years (standard deviation: 23) and romiplostim, 55 years (standard deviation: 23). There were 48% males in the eltrombopag group and 51% in the romiplostim group, and mean platelet volumes were 10.24 ± 1.94 fL (eltrombopag) and 10.43 ± 1.84 fL (romiplostim). Differences in distribution between the two groups for these three parameters were nonsignificant.

The use of TPO-RAs for treatment of ITP has expanded over the past decade and improved therapeutic options for patients in whom first-line treatment fails to provide clinical improvement, namely reductions in bleeding risk based on improved platelet counts [14,35]. The original indications for these treatments were limited to adult patients who had relapsed after splenectomy (and those in whom splenectomy was contraindicated) and those in whom first-line treatments were ineffective [36]. The RCT data that initially led to approval of both the treatments included patient populations in which 34–67% had previously undergone splenectomy [37–41]. One RCT included no splenectomized patients [42] and one RCT examined safety and efficacy of romiplostim versus placebo separately in splenectomized and nonsplenectomized patients [23]. Both eltrombopag and romiplostim are now FDA approved in the USA for ITP patients who have had an insufficient response to corticosteroids, immunoglobulins or splenectomy (i.e., patients who may or may not have had a splenectomy) [30,31]. These improved responses reported in clinical trials are known to translate into reduced bleeding and decreased necessity of providing concomitant or rescue medication [32,35].

We sought to quantify the responses – clinical outcomes and AEs – to TPO-RAs in a real-world setting utilizing de-identified aggregated patient data from a federated EHR network with access to multiple US HCOs. In this study of 2211 ITP patients, eltrombopag’s use was associated with subsequent significantly lower rates of BREs (including severe BREs), decreased use of rescue medications (including IVIG or corticosteroid injections) and platelet transfusions compared with romiplostim. We sought to avoid potential sources of bias that might affect our results; notably, the patient populations included here were mutually exclusive and comparable with respect to TPO-RA treatment. Patients on romiplostim at the time of data capture were excluded if they had previously been on eltrombopag, and vice-versa. Second, we assessed 20 co-morbidities for imbalances between the two patient groups. Five of these, acute hepatitis C with hepatic coma, prothrombin gene mutation, systemic lupus erythematosus, essential (hemorrhagic) thrombocythemia and myelodysplastic syndrome, had significantly different (p < 0.01) prevalence rates between the two treatment groups. Consequently, all patients with these co-morbidities were excluded. Third, although a minority of patients were reported to have undergone splenectomy, either before or after treatment initiation, we found that higher rates of patients treated with romiplostim subsequently underwent splenectomy (4%) compared with eltrombopag-treated patients (2%; p = 0.0039). To prevent bias, patients from both the groups were excluded if they underwent splenectomy following TPO-RA initiation. As for exclusion of patients who previously underwent splenectomy, this was considered not necessary, given the indications for both the treatments and given that the majority of clinical trials for TPO-RA enrolled patients both with and without prior splenectomy. Data from clinical trials are not uniformly consistent with respect to splenectomy prior to treatment. For example, multivariate analysis from one study that separately analyzed patients with and without splenectomy showed baseline weight <70 kg (p = 0.0106) and no splenectomy (p = 0.0306) were the only two variables significantly associated with increased rates of durable response [23]. In a separate study, previously splenectomized patients responded equally well to romiplostim compared with non-splenectomized patients [38].

The results presented here are comparable with a recently published meta-analysis of 13 RCTs that analyzed data on any or severe bleeding events from ten of the included studies. Pooled analysis of TPO-RAs indicated a significant reduction in incidence of any bleeding events compared with placebo [28]. Subgroup analysis indicated that eltrombopag, but not romiplostim, significantly reduced any or severe bleeding events compared with placebo [28].

In this real-world evidence study, we also captured data on selected nonbleeding AEs in both the patient groups. AEs, which are arthralgia, fatigue, headache and diarrhea, with >10% prevalence were the same in both the groups. In this study, the latter two AEs were present at a significantly lower rate in eltrombopag-treated patients compared with romiplostim. These AE results are in contrast with results obtained from a French Pharmacovigilance database indicating higher risk of gastrointestinal and hematologic AEs with romiplostim versus eltrombopag [50]. The reasons for differences in AE patterns between these two studies are not immediately clear but may involve a susceptibility to under-reporting in such databases, along with lower power to detect differences in AE signals between treatments [50].

While current US guidelines include TPO-RAs as appropriate second-line treatment options for ITP, it is recognized that not all patients will have a clinically meaningful response to TPO-RA treatment, and some patients discontinue TPO-RA because of the lack of response [15,43]. For example, in the EXTEND study of eltrombopag treatment, of the 302 patients enrolled, 14% withdrew from the study due to AEs, 13% due to patient decision, 11% due to lack of efficacy and 17% due to other reasons [44]. In one long-term romiplostim study, 8.5% of the 292 enrolled patients discontinued because of withdrawn consent, 3.8% due to AEs or alternative therapy, and 24% for other reasons [45].

In this respect, it is quite interesting that the recognized benefits of TPO-RAs have led to studies where switching from one to the other has proven beneficial following inadequate response to initial romiplostim or eltrombopag [46–49]. A retrospective pooled analysis of 18 switching studies (including 401 patients) estimated a relatively high response rate (improved platelet counts) of 65% when patients switched due to lack of efficacy with the first TPO-RA [49]. These differential effects of TPO-RAs in different patients demonstrating a lack of cross-resistance are consistent with known differences in activation of downstream signaling pathways between eltrombopag and romiplostim [35]. These studies raise the theoretical possibility that molecular and/or genetic markers may eventually be identified to personalize treatment decisions to select the most appropriate TPO-RA option for each patient.

Differential effects between TPO-RAs were also noted in a recent study that compared response rates (defined as platelet counts ≥50 × 10⁹/l and ≥20 × 10⁹/l higher than the pretreatment baseline) as a function of baseline TPO levels [51]. This retrospective chart review of patients on eltrombopag (n = 37) or romiplostim (n = 46) concluded that patients with normal (or lower) baseline TPO were equally likely to benefit from either TPO-RA, and those with moderate TPO levels were more likely to benefit from romiplostim than eltrombopag, whereas those with more substantial elevated baseline TPO were unlikely to respond to TPO-RA [51].

Improving platelet count is the most practical clinical end point in randomized trials, but the other key goal of treatment is to prevent bleeding, as this feature of ITP is the most worrisome for patients and clinicians [52,53], and associated costs for treatment of BREs are high. Factoring in fear of bleeding, fatigue, restriction in daily activities, altered immune responses and treatment-related side effects, there is the potential for diminished HRQoL in ITP patients [16]. Though our results did not include formal HRQoL assessments, our real-world findings suggest that patients may have fewer disruptions on eltrombopag compared with romiplostim, based on the potential reduction in platelet transfusions, fewer BREs and fewer severe BREs, along with lower rates of diarrhea and headache. Recent analyses directly link higher rates of BREs to increased healthcare resource use and elevated costs [33,34]. Reducing BREs is critical to the cost–effectiveness of ITP treatment [32].

In this respect, the type of real-world evidence presented in this study is directly relevant to potential cost savings and potential improved QoL attributable to TPO-RAs in general, and further research is warranted to investigate these issues in ITP patients on TPO-RA treatments.

This study utilized a federated EHR network (TriNetX) incorporating data from various HCOs providing statistical results, that is, aggregated data and frequency calculations, on variables of interest. Without access to individual patient-level data, regression-aided matching of cohorts cannot be carried out, and patient characteristics cannot be identified with date or time. Regression is preferred when a researcher wants to make inferences about causal relationships and predictions (e.g., using the regression coefficients to predict how changes in the covariates affect the outcome of interest); therefore, this is an important limitation. The imbalances and potential confounding factors introduced by the nonrandomized nature of the study were thus not fully accounted for. However, we believe the methods employed in this analysis are reasonable for the goals of the analysis.

While our inclusion criteria used codes for primary ITP, and we excluded patients with codes for secondary ITP, we cannot exclude the possibility that some patients had ITP secondary to malignancy, infection or other conditions. The co-morbidities (Table 1) suggest that there was a fraction of patients receiving either of the drugs who had co-morbid malignancies or infections.

As romiplostim is administered as a weekly injection and eltrombopag is an oral suspension or tablet, the difference in therapeutic setting may have contributed to differences in documentation. This may especially be applicable to the AEs of diarrhea and headache as typical outpatient complaints; likewise, minor bleeding events that do not require additional follow-up may be more likely to be recorded for patients who have more frequent clinic visits. Unlike adjudicated or audited claims, data from EHR are subject to change as patients progress through episodes of care, and this process introduces an opportunity for error as new data and older conflicting data should ideally be reconciled. This cross-sectional method does not capture duration of disease, duration of therapy, or various other clinical details that may provide additional insights. This study examined events from treatment initiation and the ensuing 6 months, while perhaps a longer duration of follow-up would identify more clinical outcomes. However, since bleeding episodes are severe occurrences, these incidences are mostly captured in the hospital setting, suggesting that there may be minimal impact to the underestimation of these clinical outcomes.

This is the first direct (real-world evidence based) comparison of clinical outcomes and AEs in patients with ITP treated with eltrombopag or romiplostim that suggests there may be a potential advantage for the use of eltrombopag, based on lower rates of BREs (normal and severe), lower utilization of rescue medications and fewer AEs versus romiplostim in the practice settings studied. Further research is warranted in more patients, and in a variety of different settings to arrive at more definitive understanding of the real-world relative risks and benefits of the two TPOs.

To view the supplementary data that accompany this paper please visit the journal website at: Supplementary Material. Some preliminary results from this study were presented at the American Society of Hematology 2017 Congress. The associated abstract is available at: http://www.bloodjournal.org/content/130/Suppl_1/3409?sso-checked=true

A Forsythe and A Roy made substantial contributions to study conception/design, data acquisition and critically revising the article for important intellectual content. CS Kwon, J Schneider and A Allepuz substantially contributed to data analyzed and critical article revision. T Pham, M Bhor, Q Said and MSO Portella were substantial contributors to study design and data analysis, as well as to the critical review of the article. All authors approved the final version to be published and agreed to be accountable for all aspects of the work.

This study was conducted with support from Novartis. A Allepuz, T Pham, M Bhor, Q Said, MSO Portella and A Roy are employees of Novartis. The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.

The authors would like to acknowledge that this paper was written and edited with assistance from N Merritt and C Crotty of Purple Squirrel Economics, with support from Novartis.