Prostate cancer screening using PSA and DRE: A decision analytic model. Free

A50

Objective: To estimate the effect of one-time screening for prostate cancer using Prostate Specific Antigen (PSA) and DRE (digital rectal exam) on life expectancy (LE) and Quality Adjusted Life Expectancy (QALE) in the context of current diagnostic and treatment practice. Background: PSA-based screening has been controversial since its introduction in the late 1980's. The incidence and aggressive treatment of well-differentiated organ confined cancers has increased, leading to concerns about overdiagnosis and -treatment. Our model estimates the effects of screening on LE and QALE incorporating recent data. Methods: A semi-Markov state transition simulation was built describing relevant health states. Two strategies were compared: 1) Screening - a single screening PSA and DRE; 2) No Screening - patients diagnosed (dx) after developing symptoms. Transition probabilities and utility weights were derived from a review of the literature and expert opinion. Sensitivity analyses were performed on all parameters. A PSA threshold of 4 ng/mL and age 65 was used for the base-case. The model has been validated against SEER and published literature. The model was developed using TreeAge™ software. Results: For our base case, screening conferred a LE benefit of 0.37 y (15.86 vs. 15.49 y) and a QALE benefit of 0.22 QALY (15.65 vs. 15.43 QALY). The PPV of +PSA/+DRE was 55%. Predicted 5 y cancer specific survival (CSS) for patients dx after screening was 98.5% (vs. SEER 99.1% +/- 0.2 SE); 10 y CSS was 95.7% (vs. SEER 97.7% +/-0.4 SE). Predicted 5 y CSS for patients dx by DRE screening was 98.3% (vs. SEER 5 y CSS 76% for patients dx 1984-86), reflecting modern treatment and screening practices. The model also predicted stage at dx: 8.8% of screened patients had metastatic disease vs. 5% in SEER. In our unscreened population, the rate of metastatic disease was 18/100,000 vs. 15/100,000 in SEER. Sensitivity analyses (SA) Increasing the incidence of organ-confined cancers in PSA+/DRE- men from 43% to 100%, a surrogate for increasing lead time, marginally increased LE (15.86 to 15.88 y) and QALE (15.65 to 15.66 QALY). Increasing the probability that PSA+/DRE- men with well-differentiated (Gleason<7) organ-confined cancer underwent treatment from 71% to 100%, a surrogate for increasing overtreatment, slightly decreased LE (15.86 to 15.85 y) but QALE was unaffected. SA of utilities revealed the model was most sensitive to varying the disutility of positive PSA at screening: even so, a 50% increase was required to decrease the QALE advantage of screening by 3 weeks. Conclusion: Our validated model demonstrates a modest benefit to one-time screening for prostate cancer. Our results support the importance of lead time and the overtreatment of clinically insignificant disease but suggest that the effect may be smaller than previously thought in men aged 65. The impact of varying lead time, PSA threshold, serial screening and cost effectiveness on LE and QALE is being analyzed.