Comparative and integrated application of flow cytometry and next-generation sequencing for minimal residual disease monitoring in acute myeloid leukemia

Minimal residual disease (MRD) detection has emerged as a important prognostic biomarker in acute myeloid leukemia (AML), While both flow cytometry (FCM) and next-generation sequencing (NGS) are established MRD detection methods, their comparative performance, concordance, and optimal integration strategy remain incompletely defined in real-world clinical practice. To evaluate the concordance, diagnostic performance, and prognostic value of FCM versus NGS for MRD monitoring in AML patients achieving morphological complete remission. We conducted a retrospective cohort study of 210 AML patients treated between January 2014 and December 2023 who underwent simultaneous FCM and NGS MRD assessment during remission. MRD positivity was defined as ≥ 0.1% leukemia-associated immunophenotype cells by FCM or ≥ 0.01% variant allele frequency by NGS. Primary outcomes included relapse-free survival (RFS) and overall survival (OS). Secondary outcomes included concordance analysis and subgroup-specific performance. Among 210 patients (median age 54 years, 56.2% female), FCM detected MRD in 119 patients (56.7%) while NGS identified MRD in 128 (61.0%), with an overall concordance of 93.81% (197/210; κ = 0.872, 95% CI: 0.812–0.932). Using NGS as reference standard, FCM demonstrated sensitivity of 91.41% (117/128), specificity of 97.56% (80/82), positive predictive value of 98.32% (117/119), and negative predictive value of 87.91% (80/91). Among 13 discordant cases (6.2%), 11 were FCM-negative/NGS-positive, predominantly harboring NPM1 mutations (8/11, 72.7%), with median VAF of 0.35% (range 0.12%-0.98%). During median follow-up of 18 months, 54.5% (6/11) of FCM-negative/NGS-positive patients relapsed versus 50% (1/2) of FCM-positive/NGS-negative patients. Integrated assessment stratified patients into three distinct risk groups: dual-negative (n = 80, 2.5% relapse rate, 11.2% mortality), single-positive (n = 13, 53.8% relapse rate, 23.1% mortality), and dual-positive (n = 117, 95.7% relapse rate, 57.3% mortality) (P < 0.001 for all comparisons). For relapse prediction, the combined approach achieved an area under the curve of 0.965 (95% CI: 0.935–0.989) with 98.4% sensitivity and 97.5% negative predictive value, superior to either method alone. FCM and NGS demonstrate high concordance for AML MRD detection while providing complementary information that enhances risk stratification. The integrated application of both methods improves prognostic assessment, with dual-negative status identifying patients suitable for treatment de-escalation and any MRD positivity warranting intensified monitoring. These findings support a personalized MRD monitoring strategy incorporating both technologies at critical therapeutic decision points.