Nationwide impact of CAR-T approval on acute lymphoblastic leukemia mortality in the United States, 1999-2023: Difference-in-differences and synthetic control analyses

W Woo Joo Lee (1AdventHealth Sebring, Internal Medicine Residency, Sebring, United States) M Muhammad Sohaib Asghar S Seon Hye Won (2Dongguk University Ilsan Hospital, Department of Family Medicine, Goyang-si, Korea, Rep. of South) R Robin Park K Khawaja Talha Aziz (AdventHealth, Sebring, Florida, United States) S Sumbal Aziz (1AdventHealth Sebring, Internal Medicine Residency, Sebring, United States) T Thomas Shimshak (Adventhealth Sebring, Sebring, Florida, United States) M Mahender Yellu (4Florida Cancer Specialists, Sebring, United States)

Abstract

Abstract Background Anti-CD19 chimeric antigen receptor T-cell (CAR-T) therapy received its first U.S. approval for relapsed/refractory acute lymphoblastic leukemia (ALL) in late 2017. The extent to which this innovation reduced population-level mortality is unknown. Methods Multiple-cause-of-death data for ALL (ICD-10 C91.0), acute myeloid leukemia (AML, C92.), malignant brain tumors (C71), and malignant bone tumors (C40–C41) were extracted from CDC WONDER (query 10 July 2025). Sex-specific mortality rates per 100 000 were calculated for three age bands (0-14, 15-39, ≥40 years), yielding 12 prespecified strata. A pre-CAR-T era (1999-2016) and post-CAR-T era (2018-2023) were defined; 2017 was treated as the transition year. Post-2018 counterfactuals were generated with a gradient-boosted quantile-regression time-series model (LightGBM; 2.5th/97.5th quantiles). Causal attribution was assessed using (i) pairwise difference-in-differences (DiD) between ALL and each control tumor type and (ii) a synthetic triple-difference estimator that combined AML, brain, and bone mortality with non-negative weights learned from pre-period data (Σw = 1). Ninety-five-percent prediction intervals were reported for the machine-learning model. Results Between 1999 and 2023, the United States recorded 152 983 deaths from ALL (114 013 during the pre-CAR-T era and 38 970 after approval). Mean annual mortality fell in 11 of the 12 predefined age-and-sex strata; the steepest absolute decline occurred in men aged 40 years or older, dropping by 1.37 per 100 000 (an 18 percent reduction). Synthetic triple-difference testing confirmed a significant post-approval decrease across every stratum, with point estimates spanning 0.06 to 1.84 fewer deaths per 100 000. The largest effect again appeared in older men (1.84 per 100 000; donor mix AML 54 percent, bone 46 percent), while even the smallest subgroup, boys aged 0 to 14 years, showed a clear reduction of 0.057 per 100 000 (weights AML 51 percent, brain 21 percent, bone 29 percent). Conclusions U.S. nationwide mortality data show a rapid and clinically meaningful decline in ALL deaths temporally aligned with CAR-T dissemination, greatest among adults ≥ 40 years. Concordant findings from machine-learning forecasts, DiD, and synthetic triple-difference analyses using biologically related malignancies strengthen causal inference and support the real-world effectiveness of CAR-T therapy. Continued surveillance is warranted to evaluate durability and equitable access.

Article Details

Journal Blood
Volume / Issue Vol. 146, Issue Supplement 1
Published November 03, 2025
Pages 4535-4535
ISSN 0006-4971
Publisher Elsevier BV

Journal Info

Blood

Elsevier BV

ISSN: 0006-4971 Health Sciences

Authors (8)

W

Woo Joo Lee

1AdventHealth Sebring, Internal Medicine Residency, Sebring, United States

M

Muhammad Sohaib Asghar

S

Seon Hye Won

2Dongguk University Ilsan Hospital, Department of Family Medicine, Goyang-si, Korea, Rep. of South

R

Robin Park

K

Khawaja Talha Aziz

AdventHealth, Sebring, Florida, United States

S

Sumbal Aziz

1AdventHealth Sebring, Internal Medicine Residency, Sebring, United States

T

Thomas Shimshak

Adventhealth Sebring, Sebring, Florida, United States

M

Mahender Yellu

4Florida Cancer Specialists, Sebring, United States