Paper Details

ABSTRACT

Background: Complete response (CR) at 3 months after CD19 chimeric antigen receptor (CAR) T-cell therapy is a strong predictor of long-term disease control in patients with large B-cell lymphoma (LBCL). Nevertheless, approximately one-third of patients achieving CR subsequently relapse. Currently, there are no known biomarkers that can identify patients at risk of relapse despite achieving CR.

Methods: We retrospectively analyzed 153 consecutive patients with LBCL treated with commercial CD19 CAR-T at our institution who achieved a best overall response (BOR) of CR between 05/2018 and 05/2025. Durable CR was defined as ongoing CR without relapse at last follow-up. Serial routine laboratory biomarkers were assessed at pre-lymphodepletion (pre-LD), day 0 (D0), day 7 (D7), day 14 (D14), month 1 (M1), and month 3 (M3). Multiparameter (15-marker) spectral immunophenotyping (Cytek) and CAR transgene quantification by qPCR were performed on available M3 peripheral blood samples.

Results: Median age at CAR-T infusion was 65 years (IQR, 55.5–72.0); 61% were male; tisa-cel, liso-cel, and axi-cel were administered to 45%, 33%, and 22% of patients, respectively. Bendamustine was used for LD in 87%. All patients achieved a BOR of CR at 3 months (median time from CAR-T infusion to best response of 87 days; IQR, 78–94). Median CR duration for the overall cohort was 16.8 months (mo) (IQR, 8.2–34.4). With a median follow-up of 19.8 mo, 93 patients (61%) remained in ongoing (durable) CR (median CR duration, 27.6 mo), whereas 60 (39%) relapsed after a median of 7.8 mo; among these, 35 (58%) relapsed within 12 mo of achieving CR and 25 (42%) after 12 mo.  For the entire cohort, the estimated 12 mo OS was 94.7% (95% CI, 89.8–97.3), and estimated 24 mo OS was 83.9% (95% CI, 76.6–89.2). Although 12 mo OS did not differ significantly between those with and without durable CR (97.8% vs 89.1%, P=.06), 24 mo OS did differ significantly (92.1% vs 72.5%, P=.003). Findings remained consistent in landmark analyses and after adjustment for infusion year.

On univariable analysis of baseline clinical characteristics, tisa-cel treatment (OR, 0.41; 95% CI, 0.21–0.79; P=.008) and ≥3 prior lines of therapy (OR, 0.46; 95% CI, 0.23–0.89; P=.021) were associated with lower odds of durable CR, whereas disease control at infusion (CR/PR/SD vs PD) was associated with higher odds (OR, 2.25; 95% CI, 1.14–4.45; P=.019).  Greater hematopoietic reserve was associated with higher odds of durable CR, including higher pre-LD platelet count (>150 ×10?/L; OR, 2.30; 95% CI, 1.09–4.82; P=.036), higher pre-LD ANC (>1.2 ×10?/L; OR, 4.04; 95% CI, 1.01–16.29; P=.047), and higher monocyte counts (>0.5 ×10?/L) at D7 (OR, 2.70; 95% CI, 1.38–5.33; P=.004) and D14 (OR, 1.97; 95% CI, 1.01–3.88; P=.049). Lower systemic inflammation and tumor burden were also associated with higher odds of durable CR, including lower ferritin (<650 ng/mL) at pre-LD (OR, 5.29; 95% CI, 1.37–20.36; P=.017) and D0 (OR, 3.09; 95% CI, 1.12–8.46; P=.040), and lower D0 LDH (<250 U/L; OR, 4.58; 95% CI, 1.25–21.70; P=.030). Composite indices integrating hematopoietic reserve and systemic inflammation similarly identified patients with higher odds of durable CR, including low-risk CAR-HEMATOTOX (0–1) at pre-LD (OR, 4.81; 95% CI, 1.19–19.41; P=.032) and D0 (OR, 2.30; 95% CI, 1.03–5.08; P=.044), and lower pre-LD mEASIX (≤0.75; OR, 4.56; 95% CI, 1.39–14.97; P=.012). No significant associations between laboratory markers and odds of durable CR were observed at M1 or M3.

M3 spectral immunophenotyping (n=37; 22 durable CR, 15 relapsed) revealed a T-cell exhaustion signature in patients who relapsed, characterized by higher frequencies of CD3?PD-1? T cells (median, 11.2% vs 6.6%; P=.048) and a trend toward increased CD3?TIM-3? T cells (P=.11). In contrast, CAR? T-cell frequency, CD4/CD8 composition, memory phenotype, and CAR transgene levels by qPCR were comparable between groups.

Conclusions: Through this novel analysis of clinicopathologic features of LBCL patients treated with commercial CD19 CAR-T, we demonstrate that durable CR is predicted by a combination of baseline clinical characteristics, hematopoietic reserve, systemic inflammation, and sustained immune surveillance rather than CAR-T persistence alone. Given the association between lack of durable CR and inferior OS, recognition of these prognostic features could inform risk-adapted surveillance and investigational post-CAR-T consolidation strategies.