Predictive Value of Mutation Analysis in the Diagnostic Approach to Patients with Unexplained Cytopenia

The diagnostic approach to unexplained cytopenia is hampered by the poor specificity of dysplastic changes that may complicate the distinction between myeloid neoplasms (MN) and non-malignant cytopenias. In the last years, several somatic mutations were identified in MN; however, the diagnostic value of mutation analysis needs to be defined. In this study, we performed a mutation screening in a prospective cohort of patients with unexplained cytopenia undergoing a comprehensive diagnostic work-up, with the aim to estimate the predictive value of somatic mutations.

This study included two cohorts: a learning cohort that consisted of 683 consecutive patients investigated for unexplained cytopenia at the Policlinico San Matteo & University of Pavia, Italy, between 2003 and 2015; and a validation cohort, including 190 patients referred as second opinion for suspected MDS. A set of 42 genes was analyzed on DNA from peripheral blood granulocytes using Illumina HiSeq (Illumina Inc., CA, USA). The diagnosis of patients in the learning cohort was MN in 409 cases (233 MDS, 86 MDS/MPN, 35 MPN; 55 AML), other cytopenia in 120 cases, whereas in 154 patients a provisional diagnosis of Idiopathic Cytopenia of Undetermined Significance (ICUS) was adopted. After a median follow-up of 22 months (range 3-136), 38 patients in this category developed a MN (ICUS-MN).

The most frequently mutated genes were TET2 (171/683, 25%), ASXL1 (15%), SRSF2 (14%), SF3B1 (11%), DNMT3A (10%), RUNX1 (9%). Significantly higher number of mutations per subject and variant allele frequency (VAF) were observed in MN (n=2, range 0-9; VAF=0.39, 0.03-0.57) compared with ICUS (n=0, 0-7; VAF=0.31, 0.03-0.51) or other cytopenia (n=0, 0-2; VAF=0.06, 0.03-0.44) (P<0.001). Fifty-seven of 409 (14%) MN patients did not carry any mutation in the set of analyzed genes. Among these, 6 had MDS del(5q), while 18 received a diagnosis of MDS based on mild dysplasia with normal karyotype, and showed a 5-year probability of progression of 0%, suggesting that a diagnosis of MDS might have been not appropriate despite WHO criteria. When accounting for these cases, an unmutated status had a negative predictive value for MN of 0.83. Conversely, having two or more mutations had a positive predictive value (PPV) for MN of 0.94. We then performed ROC analyses to explore the cut-off value of VAF with the highest predictive value for MN, and found that a VAF equal or higher than 0.10 had a PPV for having or developing MN of 0.92.

Next, we calculated the predictive value for diagnosis of MN of the most frequently mutated genes. In multivariable logistic regression, having two or more mutations (OR=4.07, P<0.001) or carrying SF3B1 mutation (OR=3.56, P=0.016) were independent predictors for MDS or other MN. SF3B1 mutation showed a specificity for myelodysplasia of 0.99, suggesting that this lesion may provide presumptive evidence of MDS even in the absence of definitive morphological features. RNA splicing genes (SF3B1, SRSF2, U2AF1) and RUNX1 had the highest predictive value for MN irrespective of co-occurring mutations, PPVs ranging from 0.88 to 0.93. PPVs increased to 0.97-0.99 when accounting for ICUS-MN as true positive cases. Conversely, PPVs of mutations in genes of DNA methylation (TET2, DNMT3A) and chromatin (ASXL1) as isolated lesions ranged from 0.39 to 0.69, while having one or more co-mutated genes significantly increased PPVs to 0.88-0.96.

Among patients with a diagnosis of ICUS, 57 of 154 (37%) carried one or more mutations (Clonal Cytopenia of Undetermined Significance, CCUS). Patients with CCUS showed a significantly higher probability of developing MDS compared with those without evidence of clonality (HR=7.48, 10-year cumulative probabilities of progression: 96% vs 15% respectively, P<.001).

Finally, the predictive values of mutation analysis were tested in the independent validation cohort of patients referred to our institution for suspected MDS, and were fully confirmed.

In conclusion, selected mutated genes or co-mutation patterns may identify patients with high likelihood of having MDS or other MN. The definition of a category of CCUS allows to recognize with high sensitivity patients who do not fulfill diagnostic criteria but are at high risk of developing MDS. Taken together, these data suggest that mutation analysis on peripheral blood cells may significantly improve current diagnostic approach to patients with unexplained cytopenia.