Day Two: ISOPT Annual Meeting

In stage 1A lung cancer, chemotherapy adds nothing, yet about 20 percent of these patients still recur. Professor Wenhua Liang shows how a gene assay changes that picture: enrich for the high-risk patients, give adjuvant EGFR TKI, and recurrence falls below 5 percent. Drawing on a Chinese registry of roughly 7 million inpatient cases from 2016 to 2021, he lays out where comprehensive treatment is heading and where access still fails.

Liang covers who actually needs adjuvant therapy, how to handle multiple primary tumors, when chemotherapy can be dropped in EGFR-mutant disease, and how to reach patients in regions where molecular testing is slow, costly, or unavailable. The recurring theme is precision applied to the parts of the system that usually get left out.

Key points for clinicians:

• In a retrospective cohort, high-risk stage 1A patients identified by a gene assay had about 30 percent recurrence under surveillance, dropping below 5 percent with adjuvant EGFR TKI. The FORWARD registry study of adjuvant furmonertinib, a Chinese third-generation TKI, was the first to include high-risk stage 1A.

• Multiple primary lung cancer accounts for about 15 percent of early-stage cases in China and East Asia. A radiomic AI model predicts EGFR status without biopsying every lesion, and EGFR TKI in predicted-positive cases reached about 60 percent overall response, sparing many patients repeat surgery.

• In EGFR-mutant disease, a reanalysis of ADAURA showed chemotherapy before EGFR TKI added no benefit. Liang argues for dropping chemotherapy and prolonging the TKI, since most patients recur after withdrawal.

• Neoadjuvant IO cycles cluster around three to four as optimal in real-world data, possibly more in multicenter work. In stage 3, patients who chose surgery over radiotherapy had better overall survival.

• Lymph node positive patients had roughly one-eighth the pathologic response rate of node-negative patients. Liang makes the case for scoring nodal pathologic response too, using a cutoff below 10 percent viable cells, since combining primary and nodal response improves prognostic prediction.

• For access, a pathological AI model predicts mutation status to triage patients toward cheap confirmatory PCR or IHC where full testing is unavailable. And for late-stage tissue, excisional surgical biopsy reached 100 percent NGS success versus about 40 percent insufficient sampling with needle biopsy, protecting the chance at targeted therapy.

Screen only smokers and people with a family history, and you miss almost half of the lung cancers you could still cure. Dr. Wenhua Liang built his talk around that gap, and around the tools his team in Guangzhou is developing to close it. His argument is that lung cancer is a chronic disease with a long window for cure, and that finding it earlier does more for mortality than any drug aimed at late-stage disease.

The talk walks through the full early-management chain: who we should be screening, how to screen them without drowning in benign nodules, how to decide which nodules actually need surgery, and how to treat the earliest lesions without cutting. It closes with the part that is still early but hard to ignore, a neoantigen vaccine aimed at the mutation that drives much of the disease in China.

Key points for clinicians:

• The current high-risk criteria leave cancers on the table. In the Guangzhou screening project, detection was about 2 percent in the high-risk population and about 1.6 percent in the non-risk population. Skipping the non-risk group would miss close to half of the cancers, and stage 1 curable disease makes up a larger share of what turns up there.
  
  

• CT is sensitive but not the right screening tool on its own, because it surfaces many benign and indolent lesions. Liang's team built a cell-free DNA fragmentomics assay on low-depth whole-genome sequencing, which is cheaper. In simulation it matched LDCT on sensitivity with higher specificity, and a prospective randomized trial against LDCT is underway.
  
  

• Overdiagnosis at the surgical step is real: in a survey of top Chinese hospitals, roughly one in four resected nodules was benign on final pathology. A ctDNA methylation assay of 100 loci reached about 85 percent AUC for benign versus malignant, and adding radiomic AI pushed the multi-modality model to about 91 percent, holding accuracy even for nodules under 1 centimeter. In real-world use it cut overtreatment by nearly 90 percent and treatment delay by 73 percent.
  
  

• For pre-invasive ground-glass lesions, a single-agent PD-1 antibody produced immune-cell infiltration and complete responses, with lesions disappearing. Liang framed this as a first demonstration that reversing immune evasion can clear an early lesion.
  
  

• Using AI to flag EGFR-mutated lesions on imaging alone, the team gave two months of a third-generation EGFR TKI with no biopsy, and saw about a 60 percent response rate. But roughly 80 percent of lesions recurred after the drug was withdrawn, which is why consolidation matters.
  
  

• To consolidate, they built what Liang described as the first EGFR-mutant neoantigen mRNA vaccine, covering the major mutation types that account for about half of Chinese patients. It was safe with no rash, 80 percent of patients showed immunogenicity on ELISpot, and in heavily resistant patients PD-1 plus the vaccine reached 5.7 months progression-free survival with durable responses. A mouse model showed a prevention signal, and an in-human prevention study is now running.