Diagnostic performance and heterogeneity of validated flow cytometry scores: a systematic review

Most flow cytometry diagnostic scores are derived from single-center, retrospective studies, and the reproducibility of their reported performance in validation cohorts remains uncertain. We searched MEDLINE for FC diagnostic scores with at least one validation cohort published in the past 10 years, ultimately focusing on B lymphoid disorders (B-LPD) and myelodysplastic syndrome (MDS) scores, where most validation efforts have been undertaken. Forty-four publications were included: 29 on MDS (10 scores) and 15 on B-LPD (10 scores). Most scores were validated once or twice. For scores with both derivation and validation cohorts, sensitivity and specificity differences ranged from -0.25 to +0.29 (validation minus derivation). Among the three scores with more than five validated cohorts, I2 values were consistently high for sensitivity and, in two cases, also for specificity, indicating high heterogeneity. To explain this variability, we systematically reviewed study methods and identified 31 potential sources of heterogeneity, spanning study design, pre-analytical, analytical, and post-analytical factors. In conclusion, few flow cytometry scores have been validated, most only in limited cohorts, and their reported diagnostic performance varies widely. While numerous potential sources of variability can be identified, the small number of available studies prevents reliable quantification of their impact. Published accuracy estimates should therefore not be accepted at face value, and proposed scores should be validated in-house before being adopted into diagnostic practice.