High Predicted Thiotepa Exposure is Associated with Increased Risk of Thrombotic Microangiopathy and Non-Relapse Mortality in Pediatric Patients Undergoing Alpha-Beta T-Cell Depleted Haploidentical Transplantation

Thiotepa is commonly included in combination with other agents to prevent rejection and relapse following alpha-beta-T-cell/CD19-depleted (AB-TCD) haploidentical hematopoietic cell transplant (HCT) for pediatric patients with hematologic malignancies. A standard regimen of 5 mg/kg for two doses was developed in adults and has been extrapolated to the pediatric population. We hypothesized that outlier (low or high) thiotepa exposures would be associated with inferior event-free- (EFS) and overall- (OS) survival due to increased rejection and non-relapse mortality (NRM) in pediatric patients undergoing AB-TCD haploidentical HCT. We also hypothesized that there would be an increased incidence of transplant-associated thrombotic microangiopathy (TA-TMA) in patients with high thiotepa exposure. Utilizing a validated pharmacokinetic (PK) model, we retrospectively predicted total exposure of thiotepa and its active metabolite TEPA combined as a cumulative area under the curve (cAUCtotal) in 203 patients with hematologic malignancy undergoing AB-TCD haploidentical HCT at nine centers on two prospective trials between 2015 and 2025. No actual PK samples were available. The median thiotepa dose was 5.0 mg/kg (range, 3.4 to 5.3) administered in two doses 12 h apart, resulting in a median predicted cAUCtotal of 50.3 mg·h/L (range, 36.8 to 64.7). We first modeled time-to-event (rejection, relapse, or death) as a function of thiotepa cAUCtotal using fitted "b-splines" under Cox regression, with complexity limited to control over-fitting. We identified the cAUCtotal level that maximized the 3-yr EFS probability as approximately 45 mg·h/L (95% confidence interval [CI], 43 to 52). We next established conservative cut-points to identify an optimal range of exposure to account for residual variability in the PK model, classifying patients as having low (<42 mg·h/L; n = 21), medium (42 to 50 mg·h/L; n = 78), or high (>50 mg·h/L; n = 104) exposure. Low thiotepa exposure was significantly associated with a higher 1-yr cumulative incidence of rejection (23.8% versus 4.5% for ≥42 mg·h/L; P < .001), and high thiotepa exposure was significantly associated with a higher 3-yr NRM (20% versus 6.8% for <50 mg·h/L; P = .007). Furthermore, a thiotepa cAUCtotal of ≥53 mg·h/L was associated with development of TA-TMA; the 1-yr cumulative incidence was significantly higher in those with thiotepa exposure ≥53 mg·h/L (25.6% versus 7.5% for <53 mg·h/L; P < .001). No statistically significant associations were found between thiotepa exposure and Day 100 sinusoidal obstruction syndrome (P = .337), 3-yr relapse (P = .449), or 3-yr EFS (P = .212). However, 3-yr OS was significantly lower in those with high thiotepa exposure (68.9% versus 85.8% for <50 mg·h/L; P = .007). We then sought to evaluate whether key thiotepa exposure-outcome associations were explained by potential confounders (patient/donor age, antithymocyte globulin exposure, etc.). To contrast outcomes across groups defined by high relative to medium/low exposure, propensity scores (PSs) were used to re-weight the dataset to balance confounders across groups. The PS-weighted hazard ratio (HR) for NRM was 3.19 (high relative to medium/low exposure; 95% CI, 1.14 to 19.89). The data did not support PS reweighting to contrast low relative to medium/high thiotepa exposure, and we instead used adjusted Cox regression. The adjusted HR for rejection was 6.81 (low relative to medium/high exposure; 95% CI, 1.55 to 29.83). The associations between thiotepa exposure and key outcomes such as NRM and rejection did not appear to be explained by measured confounders. For pediatric patients undergoing AB-TCD haploidentical HCT for treatment of hematologic malignancies, we found predicted cAUCtotal of thiotepa was associated with increased risk of rejection (low exposure), TA-TMA and NRM (high exposure), as well as overall mortality (high exposure). The identified optimal cAUCtotal of 45 mg·h/L was well below the median of 50.3, suggesting that the conventional regimen of 5 mg/kg × 2 doses may be supra-therapeutic for many patients. Model-based dosing of thiotepa to achieve optimal exposure may lower toxicity and improve outcomes and should be tested in prospective trials.