As next-generation antibody-drug conjugates reshape cancer care, digital pathology and artificial intelligence are transforming how HER2 is measured. The advances aim to help clinicians identify low and ultra-low expressors, match patients to the right therapies and make more precise treatment decisions.
The latest generation of antibody drug conjugates (ADCs) has become a focal point in oncology because they can target cancer cells more precisely. But getting the benefit hinges on accurately measuring how much HER2 protein is present in a patient’s tumor to guide treatment.
Traditional lab tests can miss important details, especially for patients with low HER2 expression, but digital pathology and artificial intelligence (AI) are making it much easier to accurately measure HER2 levels. This means more patients can get the most effective ADC treatments for their specific cancer, and doctors can make better decisions about care.
The ADC and HER2 Evolution
The technology surrounding ADCs has existed for more than 20 years, with gradual advancements in three distinct components: the payload, linker and antibody, explained Dr. Rob Monroe, vice president and chief scientific officer of oncology at Danaher Corporation and chief medical officer at Leica Biosystems.
Scientific advances, such as better designed antibodies and more stable linker systems, have allowed for more effective payload release, Monroe said. This in turn has led to therapeutics like Daiichi Sankyo/AstraZeneca’s Enhertu. Enhertu is a targeted antibody- ADC indicated for HER2-positive or HER2-low breast cancer, HER2-positive gastric/gastroesophageal junction cancer, and HER2-mutant non-small cell lung cancer.
HER2 is now having a “rebirth” under the premise of digital and computational pathology and the ability to identify low and ultra-low responders, said Jennifer Faikish, vice president and franchise head of oncology at Danaher Corporation. The advancements offer the ability to dive deeper into cell biology and see small levels of protein expression that can make patients eligible for targeted therapies.
A traditional visual pathologist might have a hard time distinguishing these protein signals with the eye, due to different staining patterns that are difficult to see with lower expressors, Faikish said.
AI however, can be trained on a vast amount of HER2 images and outcomes for subtle pattern detection, outperforming basic image analysis, Monroe said.
Adoption Hurdles
Despite the incredible potential of both digitization and AI, there is still some time before widespread adoption, Monroe and Faikish agreed. Currently, both technologies are an addition to a pathologist’s workflow, Monroe said. Also, relative to radiology images, pathology images are significantly larger: on order of 10 to 20 times. They’re gigabyte file sizes; therefore, they are much more difficult to manipulate as a pathologist orresearcher.
Novel technology implementation and cybersecurity concerns are also among the reasons behind the slow U.S. adoption, Faikish added. Digital or computational pathology adoption is globally quoted at around 25% of labs, led by EU markets and then some U.S academic labs, she said.
Ultimately, digitization will have workflow advantages, such as centralization of imaging in a cloud-based environment, to allow pathologists to share knowledge, she said.
Like many new technologies, the reimbursement certainly lags in the U.S. and other countries, Faikish said. Associations like the Digital Pathology Association and the College of American Pathologists are very focused on driving reimbursement, including both the services that pathologists provide and for oversight to run these advanced technologies.
In terms of the lack of digital standardization and interoperability, that’s a “critical issue that’s facing the field,” Monroe said, noting the many different systems and file formats. There’s a movement to have Digital Imaging and Communications in Medicine, an international standard for storing, transmitting and managing medical imaging data, apply to pathology, he added.
Bright Outlook
In terms of future collaboration among pathologists, data scientists and oncologists evolving over the next five years, Faikish noted the incredible potential for digital and computational pathology in terms of advancing companion diagnostics, therapy predictions and potential diagnosis.
There is also potential for assays to have multi-ADC targets, AI for best match selection and oncologist-led trials for sequencing and combinations, Monroe added.
You can hear more on this week’s Denatured podcast episode.
