Variation in the Use of Postoperative Radiotherapy Among High-risk Patients Following Radical Prostatectomy

TM Morgan; SR Hawken; KR Ghani; DC Miller; FY Feng; SM Linsell; JA Salisz; Y Gao; JE Montie; ML Cher


Prostate Cancer Prostatic Dis. 2016;19(2):216-221. 

In This Article

Materials and Methods

Michigan Urological Surgery Improvement Collaborative

MUSIC is a quality improvement collaborative funded by Blue Cross Blue Shield of Michigan (BCBSM). Established in 2011, this initiative tracks all newly diagnosed prostate cancer patients seen at participating practices.[7] There are 42 participating practices, representing a diverse set of academic and community practice settings. Each MUSIC practice obtained an exemption or approval for participation from a local institutional review board.

Study Population and Identification of Postoperative Radiotherapy

Patients with pT3 or greater disease and/or positive surgical margins after undergoing RP at a participating practice from March 2012 through March 2014 were identified. Patients that had <6 months of follow-up after RP or did not reach a nadir PSA ≤0.1 ng ml−1 within 6 months post RP were excluded. Those with node-positive disease were included, as long as they met these criteria. Patients were followed over time, and the administration of postoperative radiotherapy was abstracted from patient records. ART was defined as radiation administered ≤1 year post RP, with all post-nadir PSA levels <0.1 ng ml−1. SRT was defined as radiation administered >1 year post RP and/or after a post-nadir PSA ≥0.1 ng ml−1. MUSIC does not provide recommendations to providers regarding postoperative radiotherapy administration. Trained abstractors at each site enter data into a web-based registry. Abstractors are prompted to review the medical record for postoperative PSA values at 10 weeks post RP and every 6 months from the most recent PSA value abstracted.

Data Validation

As described elsewhere, MUSIC has protocols to ensure data accuracy.[8–10] This includes standard operating procedures and variable definitions, such as the distinction between ART and SRT. In addition, there are regular abstractor training sessions and site visits with data audits. To ensure accurate classification of ART and SRT, we reviewed the timing of each patient's post-RP radiation therapy administration and his corresponding PSA level. In the rare instances with apparent inconsistencies, we reviewed medical records and reclassified patients when indicated. We also externally validated the postoperative radiation treatment data by comparing BCBSM claims with MUSIC registry data. As ART and SRT cannot be reliably distinguished using insurance claims, we grouped these together and assessed the receipt of any postoperative radiotherapy for prostate cancer.

Among men in the study cohort with BCBSM insurance, we obtained all claims data (n=144). On the basis of our prior work,[8,11] we used Current Procedural Terminology and International Classification of Diseases, ninth revision, codes (ICD9-CM) to define a claims-based algorithm for the receipt of postoperative radiation therapy for prostate cancer (Supplementary Table 1). We then determined the Cohen κ statistic to examine the level of agreement between documentation of post-RP radiotherapy in the MUSIC registry and the presence of post-RP radiotherapy claims.

Statistical Analyses

The primary outcome was ART administration, and the secondary outcome was the SRT administration. First, we used chi-square and Fisher's exact tests to compare patient demographic, clinical and pathologic characteristics, as well as practice size based on the receipt of ART. We then fit multivariable logistic regression models to estimate practice-level rates of ART, adjusting for patient age, pathologic stage, Gleason grade and preoperative PSA. For the purpose of model convergence, practices with no ART administered were excluded. Next, we determined the overall use of SRT among patients with a post-nadir PSA ≥0.1 ng ml−1 and no prior ART. For practices with at least four patients meeting these criteria, we determined the SRT rate, adjusting for the same variables as the ART model. Throughout the manuscript, all reported practice-level rates of ART and SRT administration are adjusted values, whereas reported rates of ART and SRT for the entire cohort are only adjusted when specified. The maximum PSA level before SRT administration was determined for all SRT patients. All statistical testing was performed using SAS v.9.3 (SAS Institute, Cary, NC, USA) at the 5% significance level.