Yes, we can omit TBI for ALL!

In this issue of Blood Advances, Ben Hassine et al¹ report on an elegant post hoc pharmacokinetic-pharmacodynamic (PKPD) analyses focused on a subgroup of patients treated on the FORUM trial: a randomized controlled trial studying the effect of total body irradiation (TBI)-free (chemotherapy-based) regimens vs "standard of care" TBI/VP16 regimens for pediatric patients with acute lymphoblastic leukemia (ALL).² The trial was terminated prematurely when a futility stopping rule was applied because patients receiving chemotherapy-based conditioning with fludarabine, thiotepa, and either busulfan or treosulfan, appeared to have inferior overall survival (OS) compared to those receiving TBI/VP16, driven by higher relapse rates.

This, unfortunately belated, post hoc PKPD analysis is nevertheless a crucial one in the effort to identify TBI-free options for ALL. The data suggest that for patients who received optimally exposed busulfan, the 5-year leukemia-free survival/OS is the same as for patients who received TBI/VP16 conditioning and the finding of inferior survival in the busulfan-treated patients was driven by the patients who received nonoptimal exposures. Furthermore, TBI is associated with significantly more devastating late effects, especially in younger children, including high incidences of second malignancies, cataract, endocrine abnormalities, cognitive functioning, etc.³ Therefore, if a long-term (10-20 year) analysis of a composite event-free survival were to be performed, we believe that a TBI-regimen would prove to be inferior to optimally-dosed busulfan. Therefore, avoiding TBI-based myeloablation should be on top of the list for all pediatric oncologists; this approach is being actively studied in the Pediatric Transplant and Cellular Therapy Consortium’s EndRAD Trial (ClinicalTrials.gov identifier: NCT03509961).

In 2016, Bartelink et al demonstrated, in a very large (including 674 patients) international multicenter retrospective busulfan PKPD analyses, that estimated exposures derived from an individual center compares very poorly with centrally re-estimated exposures using a validate population-PK model.⁴ Up to fourfold differences in cumulative exposure were found in that analysis, toward both under- and overexposure. The analysis by Ben Hassine et al, once again underscores the importance of standardizing the busulfan-exposure estimation using a validated population-PK (popPK) model. The identified optimal exposure in this post hoc analyses by Ben Hassine et al (73-98 mg·h/L) was very similar to the optimal exposure described by Bartelink et al (78-101 md·h/L). In both PKPD analyses, underexposure was found to be associated with relapse, whereas overexposure was associated with increased toxicity (and associated mortality). Although PK-guided dosing was recommended on the FORUM trial, it lacked guidance on what model (popPK vs in-house or noncompartmental methods) to use to attain the protocol recommended cumulative exposure of 66 to 90 mg·h/L. Re-estimated exposures in the post hoc analysis ranged between 50 to 134 mg·h/L (with over 35% of patients being outside the optimal exposure range and associated with >20% lower survival). The Dutch Childhood Oncology Group reported that using a similar busulfan-based regimen (including clofarabine instead of thiotepa), targeting a cumulative exposure of 90 (±5) mg·h/L showed similar leukemia survival as the TBI arm on the FORUM.⁵ Ben Hassine’s analyses go further, because it analyzed patients, who received either TBI/VP16 or the busulfan-based regimen on same randomized trial.

In line with others, they also nicely cross validated the analytical methods for busulfan quantification in each local PK laboratory (to take potential variations in that out of the equation). They subsequently re-estimated the real busulfan exposures using a validated popPK model. PopPK models (unlike other PK models) are now considered gold-standard by the US Food and Drug Administration/European Medicines Agency. Over the last decade several validated popPK models for busulfan have been developed. They compare very similarly, with some behaving a little better for the younger age range and others better for the older age range.^6,7 More recently, by having added more PK data from patients at all ages, including neonates to feed existing validated PK models, an optimized popPK model for busulfan was validated. This optimized popPK model⁶ compared better at all ages and can therefore be considered a further refinement of the various Bayesian popPK models. Historical limitations like there being no consensus about what kind of model to use or easy access to platforms to interface with these validated popPK models, should no longer be a limitation to current practice and future trials.

In addition to busulfan there is more to improve with PKPD modeling. Others have shown that high variability in rabbit anti-thymocyte globuline (rATG; thymoglobulin) and fludarabine exposure can also impact the outcomes. Both agents were also part of the conditioning used in the FORUM trial. With further optimization of these agents the outcomes can potentially further improve, which was demonstrated in both the prospective trials using model-based dosing of rATG, as well as real-world updates using it.^8,9 We should take advantage of great leaps in technology and population PKPD modeling to use these model-based dosing strategies for all our patients undergoing transplant. Future trials should focus on this rather than trying to utilize one-size-fit-all regimens based solely on milligram per kilogram or milligram per meter square dosing. An upcoming bone marrow transplantation-CTN trial (2502) will open in the summer of 2026, where all agents in conditioning will be dosed, based on the model, to previously identified optimal targets for thymoglobulin, fludarabine, and busulfan. This is an example which shows that it is possible to study this critical question in multicenter settings.

In summary, the post hoc analyses by Ben Hassine et al, is an extremely important one that helps us identify feasible alternatives for TBI. It also underscores the importance of applying more stringent trial designs, including mandating certain elements, such as drug exposure targets and the model to be used for achieving this. We have learned, not limited to busulfan, that one size does not fit all. The data presented by Ben Hassine et al, combined with that presented earlier by the Dutch group, leaves only 1 option for a FORUM 2 trial: to re-run the trial and randomize model-based dosing of busulfan with therapeutic-dose monitoring, to achieve an optimal exposure compared to TBI/VP16. In such a trial, we would also suggest including the targeting of optimal exposures of rATG and fludarabine. If the bone marrow transplantation community is serious about this, this is the opportunity to finally eliminate the devastating potential late effects of TBI.

Conflict-of-interest disclosure: J.J.B. reports compensation for consulting from Sanofi, Sobi, Merck, Rocket Pharma, Medexus, Beam Therapeutics, Papillon Therapeutics; is a data monitoring committee chair or member with Advanced Therapies and CTI; reports honoraria from CTI and Advanced Clinical; and research funding from Sanofi (all unrelated to the current study). C.C.D. reports consulting for Alexion Inc and Medexus Pharma (both unrelated to the current study); and support from the National Institutes of Health/National Cancer Institute Cancer Center.