Cancer
Approaches to Assessing Prognostic Biomarkers: An example with the Proliferation Marker Ki-67 in Prostate Cancer. Sinead Flanagan* Sinead Flanagan Hannah E. Guard Michelangelo Fiorentino Richard Flavin Zhike Lin Jane B. Vaselkiv Pati Soule Lia DePaula Oliveira Adrianna A Mendes Tamara L. Lotan Massimo Loda Stephen P. Finn Lorelei A. Mucci Konrad H. Stopsack
Background: A subset of localized prostate cancer metastasizes despite curative surgery, underscoring the need for improved prognostication. Using the proliferation marker Ki-67, we evaluated methodological strategies for assessing prognostic utility.
Methods: Prostate cancer patients (N=910 diagnosed 1983-2004) from two prospective cohorts were followed for lethal disease (distant metastases or prostate cancer-specific death) up to 2020. Tumor tissue microarrays were immunostained for Ki-67, quantified as positive nuclei per tumor cell count and categorized as 0%, >0 to <1%, ≥1 to <5% and ≥ 5%. Discrimination was evaluated with time-varying area-under-the-curve (tAUC) using a risks-based (“cumulative/dynamic”) approach and compared to a usual hazards-based (“incident-dynamic”) overall AUC (Harrell’s c).
Results: The majority of tumors (82%) had Ki-67 expression of <1%, while 3% had high Ki-67 expression (≥5%). Among 888 men without metastasis at diagnosis, 97 lethal events occurred over 32 years of follow-up. Compared to no Ki-67 expression, high Ki-67 expression (≥5%) was associated with greater rates of lethality (hazard ratio 3.9, 95% confidence interval (CI) 1.4-11), adjusted for age at diagnosis and Gleason score, a standard-of-care prognostic marker. The tAUC for Ki-67 expression to discriminate the 5-year risk of lethal disease was 0.66 (95% CI 0.56, 0.76), higher than its overall AUC across follow-up of 0.60 (95% CI 0.55, 0.65). Throughout follow-up, from 5 to 25 years, the tAUCs for the combination of Ki-67 positivity and Gleason score for assessing the risk of lethal disease (0.74 to 0.73) were comparable to those of Gleason score alone (0.71 to 0.72).
Conclusions: Risks-based (cumulative/dynamic) tAUC evaluation of the Ki-67 biomarker highlighted its potential to inform prognostication in the early years after prostate cancer surgery, but less so later during follow-up. These time trends would not be apparent with hazards-based overall AUC modeling.