Clinical Utility of Hematopathologist-Triaged NGS Testing When Investigating Patients with Suspected MDS

The recognition of MDS is challenging in early stages, where diagnosis may rely solely upon morphological criteria for dysplasia, a non-specific finding prone to inter-observer variation. Patients with equivocal bone marrow (BM) findings may be discharged from Hematology clinics and lost to follow up, or subjected to serial, invasive BM investigations and diagnostic delays. We therefore aimed to demonstrate the importance of hematopathologist-triaged, targeted NGS in identifying clonal cytopenias of undetermined significance (CCUS) in cases where MDS diagnostic criteria are not met based on morphology or cytogenetic analysis. We explored this using three REB-approved cohorts.

Our first cohort was retrospective with BM samples ranging from 2010-14, involving cases that were previously suspicious for but non-MDS diagnostic. This included 70 patients from Sunnybrook (SHSC) and Kingston Health Sciences Centres (KHSC): 16 age-matched controls (8 negative lymphoma staging, 8 non-MDS cytopenias); 18 suspicious for MDS; 20 MDS; and 16 MDS/MPN. DNA was extracted and NGS was performed using our custom 48-gene Ion Torrent AmpliSeq myeloid panel (ThermoFisher).

We identified suspected mutations in 2/16 (13%) controls (i.e. CHIP), 12/18 (67%) suspicious cases, 17/20 (85%) MDS cases, and 16/16 (100%) MDS/MPN cases. The mean and median number of mutations per suspicious patient (respectively 0.89 and 1; most commonly in SF3B1, TET2, RUNX1, and ASXL1) were lower than MDS (1.85 and 2; p=0.011) and MDS/MPN (3.13 and 3; p<0.0001). There was a significant difference in the average variant allele frequency (VAF) per patient (those with ≥1 mutation) between control and suspicious groups (p=0.022), however, there were no significant differences in the average VAF between suspicious, MDS, and MDS/MPN cases. Furthermore, of the 16 patients with BM suspicious for MDS, 7 went on to get MDS. 4 of these patients had at least 1 clinically relevant somatic variant, while 3 had none. Of those with at least 1 variant, 3 had IPSS-level cytopenias at the time, indicating that had their mutational status been known at the time of their assessment, they would have been diagnosed with the provisional CCUS entity (while the rest would be classified as CHIP).

To supplement these findings, we are amassing a prospective cohort involving cases at SHSC where patients have either idiopathic cytopenias (ICUS), or confirmed MDS diagnoses with one or more previously non-diagnostic BM. To date, we have performed sequencing for 36 of these patients, including 23 ICUS and 13 diagnosed MDS cases. Of the ICUS cases, 10 (44%) had at least 1 variant (mean # variant/patient = 1, mean variant allele frequency (VAF) = 34.0%) consistent with CCUS, while 12/13 (92%) of MDS patients had at least 1 variant (mean # variants/patient = 2, mean VAF = 42.3%). These findings are consistent with CCUS being common in suspicious MDS cases, with similar clonal size but lesser mutational burden than diagnosed MDS. In addition to these preliminary findings, 15/36 patients have serial samples that we are currently processing for NGS (among other cases we are accruing to present at the ASH meeting). By exploring serial cases with molecular results pre- and post- MDS diagnosis, we aim to further elucidate which features of CCUS may predict progression to MDS.

Finally, we assessed clonality in cases suspicious for myeloid malignancy in our existing prospective myeloid NGS cohort at KHSC (Ferrone et al, JMD 2021). In this cohort of 168 patients, when focusing on cytopenias yet to be diagnosed, 71 patients had suspected MDS, MPN, or MDS/MPN prior to NGS (completed using the Oncomine Myeloid Assay; ThermoFisher). 36/71 (51%) were found to have variants that indicate clonality. This facilitated diagnoses of either myeloid malignancies or pre-malignant states, with nine cases in total of ICUS resulting in the identification of variants that were non-diagnostic of MDS (mainly in TET2), but indicative of CHIP (n=2) or CCUS (n=7). Furthermore, for the limited number with available follow up data, we found no significant difference in survival between individuals with low-grade MDS (n=10) and CCUS (n=6) (p=0.457).

This evidence is in keeping with recent findings that the clinical features of CCUS may be consistent with low-risk MDS, emphasizing the importance of closely monitoring these patients, and even the possibility of assessing and treating them similarly to those with low-risk MDS.