Screening for upper gastrointestinal cancer in Hong Kong

Cheng Hui, Jasmine Lam, Kwan Man, Claire Chenwen Zhong, Junjie Huang, Martin C. S. Wong, Hon Chi YipHong 2025 · DOI · OpenAlex

Introduction

Gastric and oesophageal cancers remain significant public health challenges in Hong Kong, characterized by high mortality rates and poor prognoses. In 2021 alone, the region recorded 1,306 new cases of gastric cancer resulting in 631 deaths, and 397 cases of oesophageal cancer resulting in 299 deaths. The mortality-to-incidence ratio—approximately 0.48 for men and 0.44 for women in gastric cancer—underscores a grim reality: most patients are diagnosed at stages too advanced for curative treatment. This editorial argues for a fundamental shift in Hong Kong’s diagnostic strategy, moving away from reactive, symptom-based screening toward a proactive, data-driven risk prediction model.

The Limitations of Current Screening Strategies

Unlike Japan or South Korea, where high incidence rates justify population-wide endoscopic screening, Hong Kong has a lower incidence that makes mass screening economically inefficient. Microsimulation models suggest that universal screening in lower-prevalence regions can cost upwards of US$100,000 per quality-adjusted life year (QALY).

Consequently, Hong Kong currently relies on opportunistic Oesophagogastroduodenoscopy (OGD), typically triggered by "alarm symptoms" such as indigestion, dysphagia, or gastrointestinal bleeding. However, because early-stage upper gastrointestinal (GI) cancers are largely asymptomatic, these clinical signs often indicate advanced disease. By the time a patient presents with weight loss or difficulty swallowing, the window for early intervention has frequently closed.

The "One-Size-Fits-All" Fallacy

While risk prediction models exist to help stratify patients for screening, they are rarely one-size-fits-all. Currently, the only combined gastroesophageal risk prediction model available for general populations was developed and validated in the United Kingdom. Applying this model directly to a Southern Chinese population introduces significant risks of miscalibration due to "domain shift."

Genetic and environmental factors vary significantly across ethnicities. Asian populations, for instance, exhibit different gene-environment interactions regarding gastric cancer compared to Western cohorts. Using hazard ratios derived from a UK population to predict risk in Hong Kong likely results in inaccurate stratifications, potentially missing high-risk individuals or over-screening low-risk ones.

Rethinking Feature Selection: The Alarm Symptom Trap

A critical flaw in many existing predictive frameworks is their reliance on patient-reported alarm symptoms. While these features are highly sensitive for detecting the presence of cancer, they are poor tools for early detection. To move the needle on survival rates, predictive models must prioritize objectively measurable, early-stage factors—such as demographic data, family history, and prior medical records—rather than waiting for late-stage clinical manifestations.

A Path Forward: Localized Data and Combined Models

The authors propose the development of a localized, combined gastroesophageal risk prediction model tailored specifically to the Hong Kong population. Rather than assessing gastric and oesophageal risks in isolation, a combined model would provide a comprehensive risk score to prioritize patients for OGD based on their unique profile.

To achieve this, the authors suggest leveraging the Clinical Data Analysis and Reporting System (CDARS) of the Hong Kong Hospital Authority. This robust database contains de-identified historical records of patients previously diagnosed with upper GI cancers, providing the necessary training data to build a model that accounts for local environmental and genetic risk factors.

Conclusion

Addressing the high mortality of upper GI cancers in Hong Kong requires moving beyond the current reactive healthcare model. By developing a localized, data-driven prediction tool, the medical community can transition from identifying cancer in symptomatic patients to identifying high-risk individuals before symptoms appear. This shift from reactive to proactive care is essential for improving the five-year survival rates of gastric and oesophageal cancer patients in the region.

Summary updated

BibTeX
@inproceedings{hui2025screening,
  author = {Cheng Hui and Jasmine Lam and Kwan Man and Claire Chenwen Zhong and Junjie Huang and Martin C. S. Wong and Hon Chi Yip},
  title = {{Screening for upper gastrointestinal cancer in Hong Kong}},
  booktitle = {Hong},
  year = {2025},
  doi = {10.12809/hkmj255172},
}
Figure