
Researchers at the University of California, San Francisco (UCSF) have identified an epigenetic signature that could help predict recurrence risk in patients with low-grade endometrioid adenocarcinoma whose tumors fall into the copy number low/non-specific molecular profile (CNL/NSMP) molecular subgroup.
The findings, which were presented at the Association for Molecular Pathology’s 2026 annual meeting, reveal previously unrecognized molecular heterogeneity within this common tumor subtype and suggest that DNA methylation profiling could help identify patients at increased risk of recurrence who may benefit from closer surveillance following surgery.
Endometrial cancer is the most common gynecologic cancer in the U.S. and the 4th most common cancer overall in women. The American Cancer Society estimates 68,270 new cases of endometrial cancer will be diagnosed in 2026 with 14,450 women expected to die from the disease.
In 2013, a landmark study within The Cancer Genome Atlas classified endometrioid adenocarcinomas into four molecular subgroups based on the presence of specific mutations or copy number alteration profiles: Ultramutated, hypermutated, copy number-high, and CNL/NSMP.
The most common subgroup is the CNL/NSMP group, which comprises around 40–50% of endometrial cancers. Presenting author Matthew Zunitch, MD, PhD, resident physician in Anatomic and Clinical Pathology at UCSF described this subtype as a “leftover” group, where tumors are placed if they lack the defining alterations of the other three groups.
Although tumors in this molecular category generally follow a favorable clinical course, a subset do recur and gene expression data and tumor mutational profiling has not identified reliable biomarkers to predict this recurrence.
“Methylation profiling has already been used to explore or reinforce differences between the four molecular classifications; therefore, we knew it played a role, but experiments designed for this purpose are less likely to reveal subtleties or subclassifications within each group,” Zunitch told Inside Precision Medicine. “We suspected there may be underappreciated heterogeneity within this ‘leftover’ group and designed our study to look for it.”
Zunitch and team carried out CpG methylation profiling on 219 untreated, uterine primary, CNL/NSMP, FIGO grade 1 to 2 endometrioid adenocarcinomas. They identified around 1500 differentially methylated regions (DMRs) between tumors that ultimately recurred and non-recurrent tumors.
Using the top 700 DMRs, the researchers defined a recurrence-associated methylation signature and built a regression model to score tumors according to this signature. The scoring system splits tumors into those at low-, intermediate-, or high-risk for recurrence.
“In a general sense, the three groups differ in the types of genomic regions that are methylated, and in the direction of those methylation changes,” said Zunitch. “Additionally, we found differences in the rates of a small handful of recurrent mutations and recurrent copy number alterations between the groups.”
He reported that the new three-group system achieved an accuracy of approximately 90% for predicting recurrence, with an area under the receiver operating characteristic curve of approximately 0.99.
“If you look at the rare instances where the model misassigned a tumor, this occurred at the boundaries between intermediate-risk and one of the other groups,” Zunitch remarked. “The model never mistook a low-risk tumor for a high-risk tumor or vice versa, which would be an important feature in the clinical setting.”
The team also found that tumors within the high-risk methylation group were more likely to have CTNNB1 mutations than tumors in the other groups, which Zunitch said is in line with previous work from other researchers who reported CTNNB1 mutations as a risk factor for endometrial cancer recurrence.
The scoring system will now be validated in an external cohort. If the accuracy of risk prediction is proven, DNA methylation testing could easily be added to pathologic and molecular testing workflows for patients with endometrial cancer, as it is a technique that most molecular diagnostics laboratories are familiar with, noted Zunitch.
“Our findings could enable clinicians to identify a subset of patients with elevated risk of recurrence, which in turn could enable personalized surveillance plans after hysterectomy,” he concluded. “Perhaps the simplest example is offering high-risk patients follow-up visits at shorter intervals.”
However, he cautioned that any potential benefit of increased surveillance would need to be weighed against the risks of additional exposures such as additional biopsies or imaging studies.


