Oral Presentation Clinical Oncology Society of Australia Annual Scientific Meeting 2026

Polygenic risk scores enhance prostate cancer risk stratification: development and external validation of 5-year risk prediction models in UK and Australian cohorts (145202)

Philip Ly 1 , Hamzeh Mesrian Tanha 1 , David Goldsbury 1 , Karen Chiam 1 , Albert Bang 1 , Alan White 1 , Fiona White 1 , Amelia Smit 1 2 , Sibel Saya 3 4 , Jane Crowe 5 , Manish Patel 6 , Vivienne Milch 7 8 , David Smith 1 , Anne Cust 1 2 , Julia Steinberg 1
  1. The Daffodil Centre, The University of Sydney, and Cancer Council NSW, Sydney, Australia
  2. Melanoma Institute Australia, Sydney, Australia
  3. Collaborative Centre for Cancer Genomic Medicine, University of Melbourne, Melbourne, Victoria, Australia
  4. Department of General Practice and Primary Care, Melbourne Medical School, University of Melbourne, Melbourne, Victoria, Australia
  5. General Practitioner, Australian Prostate Centre, North Melbourne, Victoria, Australia
  6. Western Clinical School, Faculty of Medicine and Health, The University of Sydney, Sydney, NSW, Australia
  7. Cancer Australia, Sydney, NSW, Australia
  8. Caring Futures Institute, College of Nursing and Health Sciences, Flinders University, Bedford Park, South Australia, Australia

Aims: Prostate cancer is a leading cause of cancer burden worldwide, yet current early detection strategies are largely one-size-fits-all, overlooking differences in cancer risk between people. Polygenic risk scores (PGS), which aggregate the effects of hundreds of common genetic risk variants, offer an opportunity to enhance population risk stratification and support risk-based early detection. We developed and validated 5-year prostate cancer risk predictions integrating leading PGS with routinely collected health information.

Methods: Model development and internal validation used UK Biobank, a prospective UK cohort (baseline 2006-2010). External validation used the Sax Institute’s 45 and Up Study, a prospective NSW cohort (baseline 2005-2009) with genomic data generated for a case-subcohort sample (DNA collected 2022-2023). The primary outcome was incident prostate cancer within 5 years, ascertained through linked cancer registry data. We included male participants without pre-baseline prostate cancer (UK Biobank: age 40-69, n=192,300, 3640 cases; 45 and Up: age 45-69, n=1233, 481 cases). Cox proportional hazards models were fitted using evidence-based health factors (age, prostate cancer family history, breast cancer family history, vasectomy, prostatitis, aspirin use, statin use, and prior PSA testing); integrative models further included one of three well-established PGS (PGS269, PGS451, or PGS400). Model performance was assessed using Harrell’s C-index and calibration.

Results: Integration of PGS significantly improved risk predictions compared to predictions based on health information alone in both internal and external validation, increasing the C-index by 0.06 to 0.08. In the Australian cohort, the final PGS451-integrated model demonstrated good discrimination (C=0.753, 95%CI:0.726-0.779) and was well-calibrated following adjustment for differences in baseline risk between cohorts.

Conclusions: Integrating PGS with established non-genetic risk factors substantially improves 5-year prostate cancer risk prediction and demonstrates good transportability between UK and Australian populations. These findings support the use of PGS-informed risk stratification to underpin more personalised approaches to prostate cancer early detection.