Behind the blood-brain barrier: Contemporary screening practice patterns and trends of central nervous system involvement in acute myeloid leukemia treated with intensive regimens and hypomethylating agent/venetoclax

Abstract Background: Patients (pts) with newly diagnosed acute myeloid leukemia (AML) with WBC>40 x103/µL, FLT3-ITD/KD mutations (FLT3mut), monocytic differentiation (MoD), extramedullary disease (EMD) and acute undifferentiated/mixed phenotype acute leukemia (AUL/MPAL) are recommended to undergo a screening lumbar puncture (sLP) in first remission to rule out central nervous system (CNS) involvement. With an evolving treatment landscape, the data on CNS screening patterns, involvement, and outcomes remains limited. Methods: Medical records of adult pts with newly diagnosed AML treated at our institution between March 2015 and May 2025 were reviewed. MoD was designated using morphology and CD markers (≥2: CD11b, CD14, CD56). The Kaplan-Meier method and log-rank tests were used to estimate and compare the median overall survival (mOS). Results: A total of 228 adult newly diagnosed AML pts were included in the analysis. 122 (53.5%) treated with intensive chemotherapy (IC) and 106 (46.5%) with hypomethylating agent/venetoclax (HMA/Ven). Of the newly diagnosed AML pts, 132 (57.9%) met ≥1 sLP criteria and were characterized as high-risk for CNS involvement (CNS-HR), 74 (60.7%) in IC cohort and 58 (54.7%) in the HMA/Ven cohort. Of remaining 96 (42.1%) newly diagnosed AML pts without CNS risk factors (low-risk, LR), 48 (50%) were treated with either IC or HMA/Ven. In the IC-CNS-HR cohort (n=74), the pts were younger than their IC-CNS-LR (n=48) counterparts, 53.9 years (y) vs 59.67y, p=0.05. The HR features were MoD in 53 (71.6%), WBC >40×10³/µL (median 56.9) in 43 (58.1%), EMD in 26 (35.1%), FLT3mut in 15 (20.3%), and MPAL in 1 (1.4%). Nearly half of IC-CNS-HR pts (45.9%) were ELN'22 high risk, while remaining pts were evenly split between low or intermediate risk (27.0% each). Within the IC-CNS-HR cohort, 51 (68.9%) pts underwent sLP with prophylactic intrathecal chemotherapy (IT), with median time to sLP of 28 days(d). CNS involvement was detected in 9 (17.6%) pts; none of them developed CNS disease later in AML course. With a median follow-up of 19.7mo, CNS disease was detected in only 1 (4.3%) of unscreened IC-CNS-HR pts during AML relapse. The unscreened IC-CNS-HR pts had numerically lower WBC count compared to screened pts, 39.4×10³/µL vs 56.2×10³/µL, p=0.4. There were no significant differences in ECOG between these pts. There were no known late CNS involvement in 48 IC-CNS-LR pts. The median age of HMA/Ven-CNS-HR cohort (n=58) was 73.0y, compared to 71.8y in the HMA/Ven-CNS-LR cohort, p=0.22. The HR features were WBC>40×10³/µL (median 85.3) in 18 (31.0%), FLT3mut in 13 (22.4%) and MoD in 44 (75.9%). The majority (78.9%) of pts were ELN'22 high risk, whereas 17.2% and 5.2% were intermediate and favorable risk, respectively. Within the HMA/Ven-CNS-HR cohort, only 9 (15.5%) underwent sLP with prophylactic IT, all were negative for CNS involvement, and none were found to have CNS disease later in AML course. There was no difference in age between pts who underwent sLP and those who did not, 72.0 vs 71.7y, p=0.38. However, higher proportion of pts without the sLP had ECOG ≥3, 22.4% vs 0%, though this was not statistically significant (p=0.18). With a median follow up of 14.3mo for the 49 unscreened HMA/Ven-CNS-HR pts, 3 (6.1%) were found to have CNS disease at the time of systemic relapse (median 3.7mo post-initiation of frontline therapy). Of note, of 48 HMA/Ven-CNS-LR pts only 1 (2.1%) developed known CNS disease later in course. Finally, significantly more IC treated CNS-HR pts (68.9%) underwent sLP compared to HMA/Ven-CNS-HR (15.5%), p<0.05. Despite the low rate of known CNS disease at relapse in the unscreened IC-CNS-HR pts, their mOS was shorter than that of screened pts, 11.4mo vs 37.7mo, p=0.016. However, even with higher rate of CNS disease at relapse in non-screened HMA/Ven-CNS-HR pts, there was no mOS difference between screened and unscreened pts (8 vs 6mo, p=0.56). In a multivariate analysis only ELN'22 intermediate and adverse risk features were associated with mOS detriment across both IC- and HMA/Ven-HR cohorts (intermediate: HR 3.14, p=0.035; adverse: HR 5.13, p<0.001). Conclusions: In this single center retrospective analysis, CNS-HR pts treated with IC were more likely to get sLP than those treated with HMA/Ven. Given high rate of CNS involvement in screened IC-CNS-HR and late CNS involvement in unscreened HMA/Ven-CNS-HR, we recommend adhering to current guidelines and offering sLP to all eligible pts.