Endometrial cancer (EC) is the sixth most common cancer in women worldwide, with over 417,000 new cases diagnosed in 2020 alone. Its incidence is rising globally, yet early symptoms are nonspecific and diagnosis often requires invasive procedures like endometrial biopsy. This creates a need for better early detection tools rooted in disease biology.
Osteoarthritis (OA) is a degenerative joint disease affecting roughly 350 million people worldwide, causing joint pain and reduced mobility. Although EC and OA affect completely different organ systems, they share striking risk factors: advanced age, obesity, chronic inflammation, and hormonal imbalances - particularly elevated estrogen levels.
One study found that 35% of EC patients also had OA as a comorbidity. This striking overlap prompted researchers to ask whether the two diseases share not just risk factors but actual causative genes - specific molecular drivers that could be targeted with drugs to treat or prevent both conditions simultaneously.
Mendelian randomization (MR) is a powerful statistical technique that uses genetic variants as natural experiments to test whether one factor truly causes another. Because genes are assigned randomly at conception (like a natural lottery), using genetic data sidesteps confounding variables that plague traditional observational studies - for example, the difficulty of separating the effect of obesity from the effect of inflammation when both are present together.
Researchers pulled data from large genome-wide association studies (GWAS) covering over 417,000 individuals for OA and 121,885 individuals for EC. They identified single nucleotide polymorphisms (SNPs) - tiny genetic variations - associated with OA, then tested whether these variants were also statistically linked to EC risk. Multiple MR methods were used including inverse-variance weighted (IVW) analysis, MR-Egger regression, and weighted median estimation to ensure robustness.
Separately, the team analyzed gene expression data from TCGA (554 EC tumor samples) and GEO datasets (19 OA tissue samples) to find genes that were differentially expressed - meaning abnormally high or low - in both diseases. These differentially expressed genes were then tested via a drug-target MR approach using expression quantitative trait loci (eQTLs) to confirm causal relationships with EC.
The main MR analysis confirmed that OA is a statistically significant risk factor for developing EC (odds ratio 1.104, meaning approximately 10% increased risk; P = 0.032). While this effect size is modest, at a population level - given how prevalent OA is among middle-aged women - it represents a meaningful disease burden. Sensitivity analyses confirmed the result was robust and not driven by any single genetic variant.
By overlapping differentially expressed genes from EC and OA datasets, the researchers identified seven common causative genes: CDKN2A, DDA1, LRRC42, POLB, ADCYAP1R1, DNMT3A, and GLRX5. These genes were significantly enriched in pathways related to heterochromatin (the tightly packed, gene-silencing form of DNA), response to alcohol, and cellular aging - all processes with known links to cancer development.
Drug-target MR analysis then narrowed the list to four genes with confirmed causal relationships with EC: CDKN2A, DDA1, LRRC42, and POLB. All four showed increased EC risk when genetically elevated, and Steiger directionality tests confirmed the causal direction was correct (genes driving EC, not the reverse). These four genes were confirmed as having diagnostic value for EC.
Using the four confirmed genes, researchers built an EC prediction model combining LASSO regression (a statistical method that selects the most important predictors) and Random Forest (a machine learning technique that builds many decision trees and averages their outputs). Both methods independently selected the same four genes, increasing confidence in their importance.
A nomogram - a visual prediction tool clinicians can use - was built from these four genes. Its performance was impressive: AUC of 0.974 on the training set and 0.966 on the validation set. AUC (area under the ROC curve) measures how well a model distinguishes cancer from non-cancer, where 1.0 is perfect and 0.5 is random chance. Values above 0.97 indicate excellent diagnostic power.
Calibration curves confirmed the model's predicted probabilities closely matched actual outcomes, and decision curve analysis showed the model would provide genuine clinical benefit in guiding EC diagnosis decisions across a wide range of risk thresholds.
CDKN2A is a well-known cell cycle regulator whose protein products p16 and p14ARF control whether cells divide or halt. In EC, methylation of this gene silences p16, removing the brake on cell proliferation. In OA, accumulation of senescent (aged, damaged) cells triggers CDKN2A expression, driving joint inflammation. The gene also influences copper metabolism, linking it to both cancer progression and cartilage degradation.
DDA1 is a DNA damage repair gene involved in ubiquitination - the cellular tagging system that marks proteins for degradation or modification. Its overexpression in various cancers (lung, breast, colon) promotes tumor growth and invasion. In OA, DDA1 affects ubiquitin ligase activity that regulates joint inflammation. A drug called dihydroartemisinin already shows promise in inhibiting DDA1-driven cancer cell growth.
LRRC42 belongs to the leucine-rich repeat protein family, involved in cell adhesion, immune response, and cancer proliferation. It is overexpressed in lung cancer and hepatocellular carcinoma, and this study found it significantly correlated with macrophage M1 and M2 infiltration in EC - meaning it may help tumors suppress immune responses while promoting blood vessel growth. POLB is a DNA repair polymerase linked to drug resistance in colorectal, gastric, and other cancers. Its dysfunction causes genomic instability and inflammatory responses relevant to both OA and EC.
Beyond direct cancer promotion, the four genes were found to have significant relationships with immune cell infiltration in EC tumors. CDKN2A expression positively correlated with activated natural killer (NK) cells - immune cells that directly kill cancer - while negatively correlating with suppressive memory T cells, suggesting a complex role in shaping antitumor immunity.
LRRC42 positively correlated with M1 macrophages, M2 macrophages, and activated dendritic cells. M2 macrophages are generally considered tumor-promoting because they suppress immune responses and encourage blood vessel formation. LRRC42's strong correlation with multiple macrophage subtypes suggests it may help create an immunosuppressive tumor environment.
DDA1 and POLB showed correlations with T cell subsets and B cells respectively, indicating these genes may modulate the adaptive immune response within tumors. Understanding how each gene influences the immune microenvironment opens doors for combination therapies pairing these molecular targets with immunotherapy.
This study's most actionable finding is that OA is a genetically confirmed risk factor for EC. This suggests that women with OA - a very common condition in middle-aged and older women - may warrant routine screening for endometrial cancer, particularly those carrying high-risk genetic variants in the identified genes. Currently, no such screening protocol exists.
The four shared genes (CDKN2A, DDA1, LRRC42, POLB) represent potential dual-target drug candidates - compounds that might treat or prevent both OA and EC simultaneously. This could be particularly valuable for patients managing both conditions, which is common given their shared risk factors. Existing drugs like dihydroartemisinin (which targets DDA1) could potentially be repurposed.
The researchers acknowledge limitations: the data came from European populations only, and extrapolating to Asian or other populations requires caution. Sample sizes for the OA expression dataset were small (just 19 cases), and functional laboratory experiments to confirm these gene mechanisms have not yet been performed. Future studies need multicohort validation and wet-lab confirmation before clinical translation.