Cancer Research · asia
Why Do HER2-Positive Biliary Tract Cancers Respond Differently? Study Uses AI to Count Tumor Cells
A 40-patient trial found that the proportion of tumor cells strongly expressing HER2 was associated with response rates to a four-drug regimen. This offers a more detailed research lead for pathology assessment, but independent validation is still needed before it can guide treatment selection.
Biliary tract cancers classified as HER2-positive may still differ in how much they shrink after the same treatment. A study published in Nature Medicine sought clues in pathology slides: using AI to calculate the proportion of tumor cells strongly expressing HER2, researchers found that patients with higher proportions had higher response rates to a combination of targeted therapy, immunotherapy, and chemotherapy. However, this association comes from a small exploratory analysis and cannot yet serve as a basis for clinical drug selection.
The phase 1b/2 trial, called HERBOT, enrolled 40 patients with HER2-positive, unresectable advanced biliary tract cancer at 12 centers in South Korea. It evaluated first-line treatment with trastuzumab, nivolumab, and the two chemotherapy drugs gemcitabine and cisplatin. The study was published on September 22; Lunit issued an announcement on October 15 describing the pathology image analysis performed with its Lunit SCOPE platform.
The analysis focused on examining the HER2-positive classification in greater detail. All patients had already been confirmed to meet HER2-positive criteria through conventional pathology testing. Lunit SCOPE HER2 then analyzed whole-slide images, identifying and classifying individual tumor cells by staining intensity and quantifying the proportion with strong positivity, scored as “3+.” This allowed researchers to compare how many cells across an entire slide strongly expressed HER2, even among cases all classified as positive.
In the prespecified exploratory analysis, the 15 patients whose proportion of HER2 3+ tumor cells was at least 10% had an objective response rate of 80%, compared with 40% among the 25 patients below that threshold. The P value for the difference was 0.0138. An objective response means that tumor shrinkage meets predefined imaging assessment criteria; it does not mean a cure. By comparison, conventional HER2 classifications did not show statistically significant differences in response rates among these patients, suggesting that the cell proportion may provide information not captured by the existing classifications.
Tumor shrinkage and longer disease control, however, represent different levels of evidence. Median progression-free survival was 17.4 months in the high-proportion group and 10.5 months in the low-proportion group. The hazard ratio was 0.57, but its 95% confidence interval ranged from 0.23 to 1.34, crossing 1, which represents no difference. This survival result can therefore be viewed only as a numerical trend and does not establish that the high-proportion group had a lower risk of disease progression or death.
For the four-drug regimen overall, the objective response rate among the 40 patients was 55%, the disease control rate was 95%, and the median duration of response among responders was 12.6 months. The efficacy signals were accompanied by a clear treatment burden: grade 3 or higher treatment-related adverse events included neutropenia in 57.5%, anemia in 30%, and thrombocytopenia in 22.5%. These findings need to be weighed alongside efficacy.
HERBOT had no randomized control group, so it cannot establish that the four-drug combination is superior to standard treatment or determine which drug’s benefit is most closely associated with the AI measure. The authors characterized the exploratory biomarker findings, which have not undergone external validation, as hypothesis-generating. Larger, independent studies are still needed to confirm whether the measure can reliably help stratify patients. Lunit also noted that SCOPE products are currently for research use only and must not be used in diagnostic procedures or diagnostic or treatment decisions.