Presentation Authors: Lisa Newcomb*, Yingye Zheng, Anna Faino, Jeannette Schenk, SEATTLE, WA, James Brooks, Stanford, CA, Matthew Cooperberg, San Francisco, CA, Atreya Dash, Seattle, WA, Martin Gleave, Vancouver, Canada, Michael Liss, San Antonio, TX, Frances Martin, Virginia Beach, VA, Todd Morgan, Ann Arbor, MI, Peter Nelson, SEATTLE, WA, Andrew Wagner, Boston, MA, Ian Thompson, San Antonio, TX, Daniel Lin, Seattle, WA
Introduction: In active surveillance (AS), disease progression as defined by pathology can only be ascertained at follow-up biopsy. However, clinical factors could also influence the timing of biopsy. We evaluated the impact of potential ascertainment bias in in a well-established prospective AS cohort.
Methods: Data were from the multicenter Canary PASS, in which protocol-directed biopsies are performed 6-12 months and 24 months from diagnosis, and then every 2 years, and PSA is measured every 3 months. To evaluate potential bias, we first defined windows, or date ranges, for on-time (compliant) surveillance biopsies. To determine whether factors were associated with the interval of surveillance biopsy, biopsy timing (on-time, early, or late) was regressed on clinical and prior biopsy variables using multinomial regression models. Probabilities for participants having all on-time biopsies were derived from the models. We reanalyzed a published study evaluating the association of PSA kinetics (PSAk) with biopsy reclassification using Cox proportional hazards modeling. Model coefficients were compared in the full cohort and subsets of compliant participants both with and without weighting for the propensity for compliance.
Results: In 851 men with 1750 biopsies, 78% of surveillance biopsies were on-time, 11% were early and 11% were late. 805 men had at least one surveillance biopsy on time, and 562 had all surveillance biopsies on time. After adjustment for prostate size, time since diagnosis, cores with cancer and prior negative biopsies, PSAk was associated with biopsy reclassification (HR=1.61 (95% CI: 1.25-2.09) in the full cohort, but was also found to be associated with biopsy timing. In multivariable models of PSAk in participants with all biopsies on time, both without and with the propensity adjustment, PSAk was still associated with reclassification [HR=2.09 (95% CI: 1.45-3.02) and HR=2.17 (95% CI: 1.45-2.93), respectively].
Conclusions: In the Canary PASS cohort, compliance to the protocol-directed biopsy schedule was high. Applying propensity weights to account for potential ascertainment bias, we demonstrated that despite its association with biopsy timing, PSA kinetics remains independently associated with reclassification. We recommend that studies evaluating risk factors in AS should consider potential ascertainment bias and utilize methods to evaluate how this bias can influence endpoint ascertainment.
Source of Funding: Canary FoundationDoD W81XWH1410595