Thrombopoietin Level Predicts Response to Treatment with Eltrombopag and Romiplostim in Immune Thrombocytopenia

Introduction: Thrombopoietin receptor agonists (TPO-RAs) are used to treat immune thrombocytopenia (ITP), but predicting clinical response to TPO-RAs before initiation is not possible. As ITP is both a disorder of increased platelet destruction and inadequate platelet production and because endogenous thrombopoietin (TPO) levels are typically normal or only slightly elevated in ITP patients (Makar et al, 2013), we hypothesized an inverse relation between baseline endogenous TPO level and probability of response to TPO-RAs. If a relation exists, obtaining TPO levels prior to TPO-RA therapy would be highly clinically useful in the treatment planning of patients with ITP given that the baseline platelet count, receipt of treatment, and disease duration do not significantly affect TPO level in ITP patients (Mukai et al, 1996, Makar et al, 2013). To determine whether endogenous thrombopoietin (TPO) levels predict treatment response to TPO-RAs we performed a retrospective analysis of ITP patients with known baseline TPO levels who received TPO-RAs.

Methods: TPO-RA treatment and disease-related data were collected for ITP patients with a baseline TPO level treated with eltrombopag or romiplostim. Response fraction (RF) (proportion of platelet counts on therapy that were ≥50×10⁹/L and least 20×10⁹/L higher than pretreatment baseline) was determined for each patient. Multiple logistic regression was used to model the probability of three classes of treatment response [overall (RF>0.0), moderate (RF≥0.5), and superior (RF≥0.8)] based on TPO level. A generalized linear model with logit transformation was used to predict RF based on TPO level. All models controlled for duration of disease, number of prior therapies, and splenectomy status. Receiver operating characteristic (ROC) analysis was performed to identify optimal TPO thresholds for response and correlations between TPO level and various response characteristics were analyzed.

Results: 67 patients (37 receiving eltrombopag and 46 receiving romiplostim; 16 patients had discrete treatment episodes with each drug) were included. Baseline patient characteristics are summarized in Table 1. Logistic regression models demonstrated a significant predictive relation between TPO level and probability of all classes of treatment response as illustrated in Figure 1 and Figure 2. TPO level was significantly inversely correlated with all response classes; correlation coefficients and odds ratios for each class of response (per 10 pg/mL TPO increase) are given in Table 2. Generalized linear modeling demonstrated a significant predictive relationship between TPO level and response fraction for both romiplostim (P<0.001) and eltrombopag (P<0.001) (Figure 3). There was no statistically significant relation between RF or any class of response and any of the other independent variables in any model. TPO level was inversely correlated with both RF [treatment with eltrombopag, r=-0.673 (P<0.001); treatment with romiplostim, r=-0.571 (P<0.001)] and median platelet count [treatment with eltrombopag, r=-0.446 (P=0.020); treatment with romiplostim, r=-0.317 (P<0.068)]. Utilizing Youden's index, receiver operating characteristic (ROC) analysis identified TPO thresholds of ≤136 pg/mL (eltrombopag) and ≤209 pg/mL (romiplostim) as optimally discriminating between responders and non-responders. Most non-responders had high TPO levels but did respond after addition of low-dose (5-15 mg daily) prednisone. Eltrombopag non-responders who responded to combination eltrombopag and prednisone had a median (range) TPO level of 155 (93-223) pg/mL, and romiplostim non-responders who responded to combination romiplostim and prednisone had a median (range) TPO level of 391 (150-1147) pg/mL.

Conclusions: TPO levels predict response to eltrombopag and romiplostim in ITP patients, with lower levels predicting improved probability and magnitude of response. ITP patients with normal TPO levels are likely to clinically respond to either eltrombopag or romiplostim, those with modest elevations of TPO are more likely to respond to romiplostim than eltrombopag, and those with extreme elevations in TPO are unlikely to respond to either agent. Therefore, obtaining a baseline serum TPO level in any ITP patient in whom TPO-RA agonist therapy is anticipated may be clinically useful.

View largeDownload slide

View largeDownload slide

Close modal