Oral Presentation COSA-IPOS Joint Scientific Meeting 2012

The median informs the message: the accuracy of individualized scenarios for survival time based on oncologists’ estimates (#151)

Belinda E Kiely 1 , Andrew J Martin 1 , Martin HN Tattersall 2 3 , Anna K Nowak 4 5 , David Goldstein 6 , Nicholas R Wilcken 2 7 , David K Wyld 8 , Ehtesham A Abdi 9 , Amanda Glasgow 10 , Philip J Beale 2 3 , Michael Jefford 11 12 , Paul A Glare 13 , Martin R Stockler 1 2 3
  1. NHMRC Clinical Trials Centre, Camperdown, NSW, Australia
  2. Sydney Medical School, University of Sydney, Sydney, NSW, Australia
  3. Medical Oncology, Sydney Cancer Centre, Sydney, NSW, Australia
  4. School of Medicine and Pharmacology, University of Western Australia, Perth, WA, Australia
  5. Medical Oncology, Sir Charles Gardner Hospital, Perth, WA, Australia
  6. Prince of Wales Hospital Clinical School, University of NSW, Sydney, NSW, Australia
  7. Medical Oncology, Westmead Hospital, Sydney, NSW, Australia
  8. Medical Oncology, Royal Brisbane and Women's Hospital, Brisbane, QLD, Australia
  9. Medical Oncology, The Tweed Hospital, Tweed Heads, NSW, Australia
  10. Medical Oncology, Wollongong Hospital, Wollongong, NSW, Australia
  11. Medical Oncology, Peter MacCallum Cancer Centre, Melbourne, VIC, Australia
  12. Faculty of Medicine, Dentistry and Health Sciences, University of Melbourne, Melbourne, VIC, Australia
  13. Memorial Sloan-Kettering Cancer Center, New York, NY, USA

Purpose: We sought the accuracy of oncologists’ estimates of survival time in individual patients with advanced cancer.
Methods: Medical oncologists estimated survival time for patients with advanced cancer as the “median survival of a group of identical patients.” Accuracy was defined by proportions of patients with observed survival time bounded by pre-specified multiples of their estimated survival time. We expected 50% to live longer (or shorter) than their oncologist’s estimate (calibration); 20-30% to live within 0.67-1.33 times their oncologist’s estimate (precision); 50% to live from half to double their estimate (typical scenario); 5-10% to live ≤¼ of their estimate (worst-case scenario) and 5-10% to live ≥3 times their estimate (best-case scenario). Discriminative value was assessed with Harrell’s C-statistic and prognostic significance with Cox proportional-hazards regression.
Results: Median survival time was 11 months after 68 deaths in 114 subjects. Oncologists’ estimates were well-calibrated (54% shorter than observed), imprecise (27% from 0.67-1.33 times observed), and had moderate discriminative value (Harrell’s C-statistic 0.62, p=0.001). The proportion of patients with an observed survival: half to double their oncologist’s estimate was 62%; ≤1/4 of their oncologist’s estimate was 6%; and ≥3 times their oncologist’s estimate was 9%. Independent predictors of observed survival were oncologists’ estimate (HR=0.92, p=0.004), dry mouth (HR=5.1, p<0.0001), alkaline phosphatase >101U/L (HR=2.8, p=0.0002), Karnofsky performance status ≤70 (HR=2.3, p=0.007), prostate primary (HR=0.23, p=0.002), and steroid use (HR=2.4, p=0.02).
Conclusion: Oncologists’ estimates of survival time were well-calibrated, imprecise, moderately discriminative, independently associated with observed survival, and accurate for estimating worst-case, typical and best-case scenarios for survival.