Why This Study Matters
Lynch syndrome is associated with increased risks for colorectal, endometrial, ovarian, gastric, small bowel, urothelial, hepatobiliary, pancreatic, brain, and sebaceous skin cancers, and population-based testing studies estimate it affects approximately one in 300 individuals. Gross deletions of the 3′ region of EPCAM cause LS indirectly, via transcriptional readthrough and hypermethylation of the adjacent MSH2 promoter.
The authors hypothesized that deletions encompassing the entire EPCAM gene, which eliminate EPCAM expression, might not carry the same risk, and noted that legacy tests targeting only the 3′ region could misclassify such a deletion as a pathogenic LS variant.
Study Design
Retrospective analysis of individuals who underwent hereditary cancer multigene panel testing (MGPT) at a single clinical diagnostic laboratory (Ambry Genetics, Aliso Viejo, CA) between January 2012 and December 2025. Deletion subtypes were defined by MLPA or chromosomal microarray analysis. A contemporaneous MGPT-negative cohort served as the reference for multivariable logistic regression (adjusted for age at testing, sex at birth except for sex-specific cancers, and race/ethnicity). LS-associated cancer burden was modeled with multivariable Poisson regression using the MSH2 nonsense/frameshift cohort as reference.
662 individuals; 6 molecular cohorts; 1 testing laboratory; testing period 2012–2025; no randomization (retrospective). Funding: supported by Ambry Genetics, a Tempus Company, through institutional resources; no external funding was received.
Patient Population
Overall median age at testing was 48 years (IQR, 35-60 years) and 56.9% (n = 337) were female. Cohorts: isolated EPCAM full-gene deletions (n = 22), isolated EPCAM 3′ deletions (n = 96), contiguous EPCAM full-gene deletions extending into MSH2 (n = 64), contiguous EPCAM 3′ deletions extending into MSH2 (n = 116), MSH2 gross deletions (n = 155), and MSH2 nonsense/frameshift variants (n = 209). Isolated full-gene deletions represented 18.6% (n = 22 of 118) of isolated EPCAM deletions.
Primary Endpoint Results
The study did not designate a formal primary endpoint; the stated objective was to compare LS-associated cancer prevalence, tumor spectrum and PREMM5 scores across cohorts. LS-associated cancer prevalence was 13.6% in isolated full-gene deletions, significantly lower than all other cohorts (Pearson's χ², P < .001).
| Cohort | LS-associated cancer prevalence |
|---|---|
| Isolated EPCAM full-gene deletion | 13.6% |
| Isolated EPCAM 3′ deletion | 54.6% |
| Contiguous EPCAM full-gene–MSH2 | 64.1% |
| Contiguous EPCAM 3′–MSH2 | 69.8% |
| MSH2 cohorts (range) | 56.1%–72.7% |
In the Poisson model relative to MSH2 nonsense/frameshift variants, adjusted cancer burden was 83.2% lower in the isolated EPCAM full deletion cohort (aIRR, 0.187 [95% CI, 0.060 to 0.578]; P = .0036) and 31.7% lower in the isolated EPCAM 3′ deletion cohort (aIRR, 0.632 [95% CI, 0.499 to 0.800]; P = .0001). Neither contiguous EPCAM-MSH2 cohort differed significantly from the reference.
Individuals with isolated EPCAM full-gene deletions did not show evidence of LS-associated cancer risk.
— Kohler et al., JCO Precis Oncol 2026
Subgroup Analyses: Cohort-Specific Associations
Adjusted odds ratios versus the MGPT-negative reference cohort, as reported in the Results text:
| Cohort | n | Colorectal cancer OR (95% CI) | Endometrial cancer OR (95% CI) |
|---|---|---|---|
| Isolated EPCAM full-gene deletion | 22 | 1.62 (0.25–5.91); P = .529 | No cases observed (not estimable) |
| Isolated EPCAM 3′ deletion | 96 | 17.59 (11.35–27.28); P < .001 | 0.76 (0.04–3.50); P = .790 |
| Contiguous EPCAM full-gene–MSH2 deletion | 64 | 10.64 (6.14–18.22); P < .001 | 22.03 (9.69–43.10); P < .001 |
| Contiguous EPCAM 3′–MSH2 deletion | 116 | 12.64 (8.40–18.90); P < .001 | 30.47 (18.14–50.50); P < .001 |
| MSH2 gross deletion | 155 | 8.50 (5.87–12.18); P < .001 | 17.31 (10.59–27.60); P < .001 |
| MSH2 nonsense/frameshift | 209 | 16.03 (11.86–21.65); P < .001 | 22.08 (14.79–32.60); P < .001 |
For any LS-associated cancer, the isolated EPCAM full-gene deletion OR was 0.81 (95% CI, 0.21 to 2.49; P = .731) and the isolated 3′ deletion OR was 5.79 (95% CI, 3.69 to 9.11; P < .001). The source text's any-LS ORs for the four MSH2-involving cohorts are not presented here because the printed values appear internally inconsistent; see the source figure 5.
Family history and risk scores. Family history was available for 594 of 662 probands (89.7%). Among isolated EPCAM 3′ deletion carriers, 88.9% reported a first- or second-degree relative with an LS-associated cancer; 80.0% reported a relative with colorectal cancer versus 5.6% with endometrial cancer, compared with 21.2%-31.4% in MSH2-involving cohorts (overall P < .001). Median PREMM5 was 0.70 (IQR, 0.00-1.88) for full-gene deletions, with 18.1% above the 2.5% threshold, versus 5.55 (IQR, 1.9-16.4) for isolated 3′ deletions.
Safety Profile
Not applicable. This was a retrospective germline testing cohort with no treatment arms, so no adverse event data were collected or reported.
Interpretation and Broader Context
The authors conclude that EPCAM deletion architecture dictates clinical impact, from low risk for isolated full-gene deletions, to tissue-specific colorectal cancer risk for isolated 3′ deletions, to classic high-penetrance MSH2-associated LS for contiguous EPCAM-MSH2 deletions. They suggest providers may consider updated testing in individuals previously identified as having LS because of an EPCAM deletion that excludes MSH2.
Limitations
The authors report limitations including the retrospective design, referral and ascertainment biases, an inability to estimate familial penetrance, limited probe coverage near the MSH2 promoter, and limited tumor data. The isolated full-gene deletion group was small (n = 22), which is reflected in wide confidence intervals. The study therefore could not establish absolute or familial cancer penetrance for any deletion subtype.