Full Blood Count Trends for Colorectal Cancer Detection in Primary Care: Development and Validation of a Dynamic...

Area of Science:

• Oncology
• Hematology
• Biostatistics

Background:

• Colorectal cancer (CRC) diagnosis at later stages is associated with poor survival rates.
• Early detection of CRC is crucial for improving patient outcomes.
• Full blood count (FBC) is a routine blood test frequently performed in primary care settings.

Purpose of the Study:

• To develop and validate dynamic prediction models for early colorectal cancer detection.
• To investigate the utility of trends in repeated FBC parameters for predicting CRC risk.
• To assess the performance of these models in identifying individuals at higher risk for CRC within a two-year timeframe.

Main Methods:

• A large cohort study involving over 250,000 males and 246,000 females in the development cohort and over 312,000 males and 462,000 females in the validation cohort.
• Sex-stratified multivariate joint models were employed, incorporating baseline age and simultaneous trends in haemoglobin, mean corpuscular volume (MCV), and platelet counts.
• Models predicted the two-year risk of CRC diagnosis based on historical FBC data up to the most recent FBC measurement.

Main Results:

• Patient-level trends showing declines in haemoglobin and MCV, alongside a rise in platelets, were associated with an increased risk of CRC diagnosis within two years.
• The developed models demonstrated strong predictive performance with a c-statistic of 0.751 for males and 0.763 for females.
• Calibration slopes of 1.06 for males and 1.05 for females indicated good calibration with low miscalibration.

Conclusions:

• Dynamic prediction models utilizing trends in FBC parameters (haemoglobin, MCV, platelets) show significant potential for earlier colorectal cancer detection.
• Incorporating these longitudinal blood count trends could lead to earlier diagnoses, potentially improving survival rates.
• External validation of these models is recommended to confirm their generalizability and clinical utility.