Enhanced risk stratification in AML through integrated mutation and CNV profiling: Insights from Uruguay

Abstract

Abstract Background: Accurate risk stratification in acute myeloid leukemia (AML) increasingly depends on integrated cytogenetic and molecular profiling, as outlined in recent WHO and ELN (2022) classifications. However, data from Latin America remain scarce. While comprehensive genetic testing is standard in high-resource settings, its implementation across the region is limited. This multicentric prospective cohort study aimed to characterize the utility of NGS and cytogenetic testing in risk stratification of adult AML patients in Uruguay and to evaluate how the integration of somatic variant and copy number variations (CNV) findings enhances prognostic classification. Methods: We enrolled 61 adult AML patients from 12 Uruguayan hematology centers (2021–2023). Clinical data and diagnostic bone marrow samples were collected. gDNA was analyzed using a 63-gene myeloid panel (SOPHiA DDM® on Illumina MiSeq) to detect SNVs, indels, and CNVs. Conventional cytogenetics was performed in all cases; FISH in 65%. Results: We included 61 AML patients (median age 56 [39–58] years; 47% male); 38 (62.3%) were fit for intensive chemotherapy and 23 (37.7%) unfit. Among fit patients, ELN 2022 classified 25% (9) as favorable, 47.2% (17) intermediate, and 27.8% (10) high risk. Compared to ELN 2017, 27% were reclassified, mostly up-staged (18.4% from intermediate to high). Cytogenetic abnormalities were found in 31.6% (12), with high-risk features in 15.7%. (6) In contrast, CNVs were detected by NGS in 39.4%, prompting reclassification in 26%—including 5 with normal karyotypes. CNV detection showed 78% sensitivity (95% CI: 54.0–93.8) and 87% specificity (95% CI: 60.4–89.3) vs. conventional methods. Somatic mutations were found in 89.5% (median 3 per patient, IQR 1–4); most frequent were NPM1 (34.2%), FLT3-ITD (28.9%), DNMT3A and NRAS (26.3% each), IDH1/2 (18.4%), TP53 (10.5%, 80% multi-hit), CEBPA BZIP (7.9%), and RUNX1 (7.9%). Targetable mutations (FLT3 or IDH1/2) were present in 52.2%. All received 7+3 chemotherapy; 21% also received midostaurin. CR was achieved in 71% after 1–2 inductions. Median OS was 23 months (95% CI: 17–32). ELN 2022 risk groups stratified survival, and TP53 or myelodysplasia-related mutations were independently associated with worse OS (p = 0.022 and p = 0.021). Among unfit patients (median age 69 [IQR 67–75]), ELN 2024 classified 54.5% (12) as favorable, 13.6% (3) intermediate, and 31.8% (7) high risk. Cytogenetic abnormalities were present in 38.9%, while NGS-detected CNVs appeared in 56.6%, prompting risk reclassification in 26% with normal karyotypes. Somatic mutations were found in 95.7% (median 3 per patient, range 2–4), with STAG2 (34.8%), DNMT3A (30.4%), RUNX1 and TP53 (26.1% each), SRSF2 (21.7%), and IDH1/2 (17.3%) being most common. TP53 mutations were multi-hit in 83% of cases. Treatments included azacitidine alone (43.5%) or with venetoclax (47.8%), achieving CR/CRi in 56.1%. Median OS was 15 months (95% CI: 12–17). ELN 2024 risk categories correlated with outcomes (p = 0.032), and RAS, FLT3-ITD, and TP53 mutations were each significantly associated with inferior OS (p = 0.026, 0.026, and 0.001, respectively). Conclusions: This is the first comprehensive study to evaluate both somatic mutations and CNVs through targeted NGS in Uruguayan AML patients, and one of the few such efforts in Latin America, where available data are limited to small series from Brazil and Mexico. Conducted nationwide in a country with ~100 new AML cases per year, this cohort is representative and contributes valuable regional insight. Incorporating CNV analysis via NGS significantly improved the detection of clinically relevant genetic abnormalities. Additionally, somatic mutation profiling enabled refined prognostic stratification, with 27% of patients reclassified under ELN 2022, leading to potential changes in therapeutic management. The integration of CNV and somatic profiling enhanced diagnostic precision—particularly in cases with technically limited cytogenetics—reducing the size of the intermediate-risk group and supporting more tailored therapeutic decisions. High-risk molecular features, including multi-hit TP53 and myelodysplasia-related mutations, were associated with inferior survival, while targetable mutations were frequent. Overall, our findings support the feasibility and utility of a single-run NGS-based approach to improve diagnostic accuracy and optimize workflows in resource-limited settings.