GLOBAL NEWS: OVARIAN CANCER RESEARCH

Researchers from the Chongqing University Cancer Hospital in China have developed a new blood test called the Ovarian Cancer Score (OCS), which uses proteins extracted from small extracellular vesicles (sEVs) to detect ovarian cancer. In a study of over 1,100 patients, the test showed high accuracy, performing better than the commonly used CA125 biomarker and identifying early-stage ovarian cancer more effectively. 

There is currently no early detection test for ovarian cancer, and as a result, many patients are diagnosed at a late stage when it is much harder to treat. Current tests, like CA125 and HE4, are not sensitive or specific enough for early detection or screening, especially in younger women or those with normal CA125 levels. Better, more reliable screening methods are urgently needed to diagnose ovarian cancer earlier and improve patient outcomes. 

The OCS test measures three tumour-related proteins (CA125, HE4, and C5a) found in small extracellular vesicles (sEVs) from blood samples. The test was evaluated in 1,183 women with ovarian masses at four hospitals in China. Key findings include: 

• High accuracy: OCS correctly identified ovarian cancer in 95.5% of cases (sensitivity) and correctly ruled out cancer in 90.2% of cases (specificity), outperforming CA125. 
• Effective for early-stage cancer: OCS detected 89.7% of stage I ovarian cancers and 91.4% of early-stage (I & II) cases, making it a promising tool for early detection. 
• Better performance in younger women: In women under 45, OCS was 92.7% accurate, compared to just 80% for CA125, highlighting its potential for younger patients. 
• Works even when CA125 levels are normal: The test correctly identified 72.7% of ovarian cancer cases where CA125 levels were normal, showing its advantage over current methods.

While the study demonstrates promising results, larger international trials are required to validate its accuracy before OCS can be approved for widespread clinical use. Additional research will also assess its effectiveness across different ovarian cancer subtypes. The timeline for clinical adoption remains unknown, as further studies and regulatory approvals are still needed. 

Researchers based at Iran University of Medical Sciences have conducted a large meta-analysis of 3,460 ovarian cancer patients, and found that artificial intelligence (AI), particularly machine learning (ML) and artificial neural networks (ANN), outperformed traditional statistical methods in predicting survival after ovarian cancer surgery compared to current methods.  

Complete cytoreductive surgery of ovarian cancer is the process of removing all visible tumours during surgery and is a large predictor of survival. However, predicting post-surgical outcomes remains challenging due to multiple factors, including tumour biology, patient health, and surgical expertise. Traditional statistical models have been somewhat limited in their ability to predict survival probability, hospital stays, or potential complications. More accurate prediction tools could help doctors make better treatment decisions and improve patient care. 

AI models predicted overall survival with an accuracy of approximately 70%, regardless of the type of AI model they were using. When predicting whether surgeons could completely remove all visible tumours, AI reached nearly 80% accuracy, significantly better than traditional methods. The accuracy of AI predictions depended on the type and amount of data used, and AI models performed best when they included a wide range of patient-specific data, such as age, BMI, comorbidities, preoperative blood tumour markers (like CA-125), and disease stage. Other AI models have shown to be approximately 90% accurate in predicting post-surgical complications like ICU admission and hospital stay length.

When using AI to assist forecasting of preoperative outcomes, healthcare providers must trust its predictions and apply their own expertise when making decisions. With this foundation in place, AI has the potential to enhance personalised medicine and improve predictive healthcare. In order for AI models to be widely adopted in healthcare, they must first undergo external validation in large, real-world patient populations to ensure their reliability. Further advancements are needed to address technical challenges, along with policy integration and government support to guide implementation. Securing funding will also be essential for successful adoption in clinical practice.