3054 Background: Early circulating tumor DNA (ctDNA) kinetics may be instrumental in identifying patients with DLBCL who can safely discontinue therapy after three cycles, as well as those who are less likely to benefit from completing the full six cycles. However, the optimal timing for ctDNA measurement, added value of early cycle data, and influence of assay sensitivity remain uncertain. Through population tumor kinetics (pop-TK) modeling of Pola-R-CHP (Pomeroy et al., Blood Cancer Discovery, 2025), we evaluated how adding data from cycles 1 and 2 enhances risk stratification beyond the assessment conducted after cycle 3. Methods: A 5-drug pop-TK model calibrated to Pola-R-CHP single-agent sensitivity distributions and clinical outcomes simulated tumor cell populations and patient-level parameters. Predicted residual disease after each cycle was mapped to ctDNA titers under assays of varying detection thresholds, representing conventional ctDNA, high-sensitivity ctDNA, and PET. We evaluated (1) ctDNA negativity at cycle three only, (2) adding cycles 1 and 2 for prediction, and (3) using cycles 1 and 2 to identify non-responders. For detection thresholds, representative values were selected to reflect overall trends. Results: Among ctDNA-negative patients at cycle 3 (n=557), stopping therapy at cycle 3 led to a recurrence rate of 0.172, while completing six cycles eliminated all recurrences. Adding ctDNA from cycles 1 and 2 separated cycle 3-negative patients into a high-risk group with much higher recurrence (90/121; 0.744) and a low-risk group with very low relapse risk (6/436; 0.014). Across assay sensitivities, high-sensitivity ctDNA reduced recurrence after cycle 3 cessation to 0.093 (43/464), whereas PET-level sensitivity showed a higher recurrence rate (0.288; 171/593), and adding data from cycles 1 and 2 consistently improved sensitivity. Early-cycle ctDNA (C1, C2) stratified patients into a high-risk non-responder group (5-year PFS 0.009), who can benefit from transferring treatment plans, and a low-risk group with favorable outcomes (5-year PFS 0.961). Conclusions: Pop-TK modeling suggests that although most patients with cycle 3 ctDNA negativity do not relapse when treatment is continued through cycle 6, some patients may still experience relapse that might not have occurred if treatment had been completed through cycle 6. Adding early cycles produces a large low-risk subgroup, supporting safer de-escalation, while still identifying patients who will consistently relapse if therapy is stopped early. Higher assay sensitivity reduces false negatives, but early-cycle ctDNA remains very informative. Furthermore, early cycle ctDNA enables identification of non-responding patients before completion of therapy. These findings support the addition of early cycle ctDNA into response-adapted de-escalation and early treatment switching in trials for DLBCL treated with Pola-R-CHP.
Kim et al. (Wed,) studied this question.