Incorporating TET2 Variant Allele Frequency Enhances Prognostication in CMML: Insights from 849 Patients across Multiple Institutions

Chronic myelomonocytic leukemia (CMML) is a myeloid neoplasm with overlapping features of myelodysplasia/myeloproliferation and persistent monocytosis. Prognosis is often poor but varies widely, posing challenges in predicting progression and timing of transplant or systemic therapy. Several prognostic tools exist, yet none provide optimal predictive performance.

TET2 mutations occur in 60% of CMMLs yet their impact on overall survival (OS) and AML-free survival (LFS) remains inconclusive. While previous studies focused on TET2/ASXL1 mutation status, we investigated the prognostic relevance of TET2 variant allele frequency (VAF) in 845 CMML patients from multiple centers. Our UK discovery cohort comprised 645 consecutive CMML patients from North-West England, each with extensive clinical and laboratory metadata. An independent cohort pooled 204 patients from Lithuania, Spain, Taiwan, and USA. In cases of multiple TET2 mutations, that with highest VAF, as determined by targeted NGS, was used for outcome analyses.

Median age of UK patients was 75 years, with male predominance. TET2 was the most frequently mutated gene (67%), followed by SRSF2 and ASXL1. Among 433 patients, 671 TET2 mutations were detected with 218 patients having >1 mutation. Frameshift (46%) and nonsense (33%) mutations were most common. Over median 21 month follow up, 120 patients developed AML and 391 died. Median LFS and OS were 28 and 29 months, respectively.

Treated as a continuous variable by Cox regression, lower TET2 VAF correlated with longer LFS and OS (both HR 0.99, p<0.05). We explored whether grouping patients by VAF could better stratify prognostic relevance. Examining VAF distribution, segmented regression identified inflection points at 32% and 55%, marking significant changes in incidence. Patients with VAF >55% were grouped as TET2^LOH (loss of heterozygosity); those with VAF <32% as TET2^Low; and those between 32-55% were divided at the median (46%) into TET2^Int and TET2^High groups.

TET2^Int and TET2^Low patients exhibited the lowest WBC counts but highest Hb levels among groups. TET2^Int patients had similar LFS and OS compared with TET2^Low (HR 1.1, p=0.653; HR 1.0, p=0.924). Conversely, TET2^High, TET2^LOH and TET2^WT patients had worse outcomes compared with TET2^Low. Thus, we merged TET2^Int and TET2^Low groups into a single TET2^lower category (n=220); and TET2^High, TET2^LOH and TET2^WT into another (n=425). Lower TET2 VAF consistently stratified patients’ LFS and OS across the entire cohort and within subgroups: CMML-1 vs -2; 1 vs ≥2 TET2 mutations; dysplastic vs proliferative; and ASXL1 mutant vs wild type. Multivariable analysis confirmed lower TET2 VAF as independently prognostic, irrespective of age and CPSS-Mol, GFM, and MMM risks (all p≤0.005).

Of 103 patients with NGS data at disease progression (PD) (IWG criteria), the TET2^lower group had significantly longer OS from PD (p=0.002) than others. This prognostic relevance was retained in subgroups with/without prior treatment, highlighting the potential prognostic relevance of TET2 VAF at PD.

We next incorporated higher VAF or wild-type TET2 as adverse prognosticators in existing models. The Cox regression weighted TET2 VAF/status to 1, 2, and 2.5 points for CPSS-Mol, GFM, and MMM models, respectively. After adjusting risk category cutoffs to incorporate TET2 VAF/status scoring, 226 patients were re-staged in the new CPSS-Mol model (5-tier), 344 in the GFM model (4-tier), and 514 in the MMM model (5-tier).

Time-dependent ROC curve analysis showed increased area under curve for the three models compared to the originals in predicting LFS and OS, suggesting that incorporating TET2 VAF enhances predictive performance. In all new models, LFS and OS robustly stratified (all p<0.001) and incorporation of TET2 VAF improved prognostication, as evidenced by improved c-indices. The prognostic significance of TET2 VAF was validated in the extended cohort by multivariable analysis, with consistently higher c-indices for the updated risk models.

This study highlights the prognostic importance of TET2 VAF in CMML, with lower TET2 VAF linked to better outcomes at both diagnosis and PD. Integrating TET2 VAF into existing models significantly improves their predictive accuracy, underscoring the role of molecular data in refining CMML prognostication. Further studies are needed to understand the biological basis of observed VAF thresholds and their implications.