Abstract
INTRODUCTION: Despite advances in diagnostic techniques, many cases of acute myeloid leukemia (AML) remain underdiagnosed in remote centers unequipped with these technologies. We hypothesized that the automated cellular indices with scatter plots and flags may aid in rapid and cost-effective screening of AML. METHODS: Cell population data (CPD) parameters from 100 de novo AML samples were analyzed by Coulter LH 780 automated analyzer and were compared with 100 age-matched controls. Similar parameters were also compared with 100 and 50 reactive cases of neutrophilia and monocytosis, respectively. System-generated flags and scatter plot patterns were also analyzed. RESULTS: Results were compared between AML cases and normal controls; AML FAB M2, M3, M4 vs reactive neutrophilia; AML FAB M4, M5 vs reactive monocytosis. Significant parameters were selected from all comparison groups. Using appropriate statistical tools, we calculated the cutoff values of these parameters and were able to screen out AML cases with 94% sensitivity and 95% specificity. Three statistical equations were generated using two of the most significant parameters which improved the sensitivity to 98% and specificity to 99%. Five hypothetical scatter plot patterns were devised and were classified according to FAB categories of AML. Most common pattern was selected in AML which was seen in 56% of the cases. Output was analyzed combining these patterns and flags with CPD parameters. CONCLUSION: CPD either alone or in the form of statistical equations along with scatter plots and flags can provide rapid and economic tool in preliminary diagnosis of AML in cost-constrained settings.