Study Population
SA-PCCOC is a third-party collated, disease-specific, prospective, longitudinal, observational PCa registry and database established in the state of South Australia. The database presently collects approximately 93% of all new histologically confirmed diagnoses of prostate cancer in the state. Men are followed until death or withdrawal from the registry. Commencing in 1998, SA-PCCOC is the longest-running PCa database in Australia and the Southern Hemisphere and at the time of writing has collected approximately 15,000 participants.
Inclusion and Exclusion Criteria
We defined BCR as a post-operative PSA reading of > 0.2 ng/ml which is the most widely accepted consensus definition in the literature.(18)
Men from the SA-PCCOC database were included in this analysis if they had undergone RP between 1998–2018 inclusive. Our study’s censor date was 30th June 2018, to allow at least six months of available follow-up data to the end of 2018.
Of 5157 patients in SA-PCCOC treated with RP during this period, patients were excluded if they: had their surgery after the censor date of 30th of June 2018 (n = 206), did not have a PSA value within three months of surgery (n = 1528), had preoperative treatment with ADT/RTx (n = 50), received secondary treatment with ADT/RTx within six months of surgery (n = 200), had recurrence within six months (n = 62), had a PSA nadir > 0.2 ng/mL (i.e. these patients already had a BCR and thus were not at risk of future BCR) (n = 390), only had one PSA level within the 6 months after surgery (which meant a PSA nadir could not be determined) (n = 810) and had TTN of > 6 months (n = 115). This resulted in a study population of 1796 patients (Fig. S1). Additionally, all selected patients had available data for year of RP, pathological staging, Gleason score, margin status and cause of death.
As SA-PCCOC is a population-based observational registry, there was no standardisation for the timing of post-operative PSA tests which were determined by individual clinician discretion. The PSA nadir was defined as the lowest PSA post-RP where there were at least two PSA results. Time to BCR was defined as time from the date of surgery to the first PSA reading > 0.02 ng/ml.
The definition of undetectable PSA differed slightly in our study. The SA-PCCOC database derives its PSA data from several different pathology laboratories so there were considerable variations in the use of ultrasensitive PSA assays and the level each laboratory defined as an undetectable PSA. PSA assay non-uniformity across different laboratories is a recognised issue when comparing studies involving PSA nadirs.(19) As such, our study defined an undetectable PSA as a range of 0.01–0.1 ng/mL, which was representative of the span of various definitions of undetectable PSA levels across the different pathology companies during the period of observation. Skove defined undetectable PSA nadirs as assays < 0.01 ng/ml.
The SA-PCCOC database does not have complete data on ethnicity or body mass index (BMI). This study also did not analyse clinical staging data, as pathological staging was deemed to be more objective for the evaluation and clinical stage has been reported as having minimal impact on localized prostate cancer treated by RP.
The SA-PCCOC database sources patient death data from the SA Government Births, Deaths and Marriages Registry. If multiple reasons were listed for the cause of death, men were flagged as dying from prostate cancer only if it was recorded as a significant contributing cause of death and not a secondary diagnosis.
Ethics
Ethics approval was obtained from the Southern Adelaide Clinical Human Research Ethics Committee. Access was granted to the SA-PCCOC database by the SA-PCCOC research committee.
Statistical Analysis
The variable of interest was TTN which dichotomised the cohort into two groups: TTN < 3 months and TTN 3–6 months, as described by Skove.(16) Patient characteristics between these groups were compared using chi-square tests for categorical data and rank-sum tests for continuous variables.
Univariable analysis of time to BCR, OS and PCSM was assessed with Kaplan-Meier plots and log-rank tests. Multivariable analysis of TTN and time to BCR, OS and PCSM was performed using a Cox proportional hazards model. The analysis was adjusted for the following potential confounders: age at surgery (continuous), year of surgery (continuous), pre-operative PSA (continuous) and pathological Gleason score (low-risk ≤ 6, intermediate risk 3 + 4, high-risk 4 + 3 to 10). Proportional hazards assumption of the Cox regressions were assessed using the test described by P. Grambsch and T. Therneau (1994).
Associations between time to BCR and PCSM with TTN were also assessed using Fine and Gray competing risk models, which are reported as sub-distribution hazard ratios (sHR) and presented as cumulative incidence curves. In the models where BCR was the outcome, death from any cause was considered a competing risk. Where PCSM was the outcome, death from non-prostate causes was considered a competing risk. The analysis was adjusted for the same covariates used in the Cox model.
A p-value of < 0.05 was considered statistically significant for all analyses. All statistical analyses were performed using RStudio (v1.1.463) with R (version 3.5.2), with the survival survival analysis, survminer drawing survival curves and cmprsk competing risk packages.(20)