Armando Meier

M.Sc.

University of Basel

Ph.D. Student

PLCO (Learn more about this study)

PLCO-294

Aug 16, 2017

The Impact of Medical Guidelines for PSA Testing on Prostate Cancer Treatment

Medical guidelines are ubiquitous. However, little is known about how strongly doctors adhere to such guidelines. We examine whether doctors adhere to the recommendation of further tests if a patient passes the critical threshold of 4.0ng/ml in a prostate specific antigen (PSA) test. Furthermore, if doctors adhere to this guideline, we examine how that affects the prostate cancer diagnosis and treatment.

Randomized controlled trials are not feasible here: It is unethical to randomly assign PSA values to patients. We therefore leverage previously collected experimental data collected in the PLCO trial in combination with a regression discontinuity design to study the effect of quasi-randomly passing the threshold of 4.0ng/ml on further tests (e.g., biopsies, repeated PSA test), diagnosis, treatments (e.g., prostatectomy), and mortality. This seems particularly important considering the current debate about PSA screening. More generally, our results will show whether doctors adhere to guidelines, and if yes, whether adhering to these guidelines is beneficial for patients.

The regression discontinuity exploits the quasi-random variation in PSA measurements – for instance caused by measurement error – in combination with the medical guideline of further tests at greater PSA value than 4.0ng/ml. The intuition is that individuals with a PSA value of 4.1 are essentially the same as individuals with a PSA value of 4.0 or 3.9. Arguably, the only difference between individuals is measurement error or other incidental circumstances that lead to differing test results. Accordingly, the causal effect of barely passing the threshold on further tests, diagnosis, and treatment can be assessed by a comparison of individuals closely above the threshold (the treatment group) with those closely below (the control group).

We aim at determining whether doctors use a PSA value of >4.0 as the critical value to conduct biopsies or further tests. If yes, we exploit this rule based decision-making to assess the effects of passing the PSA threshold and therefore being more likely to have additional tests, a diagnosis of prostate cancer, treatments and death. Accordingly, our main results will be based on plots of the following probabilities against PSA values, where we are interested in the comparing the values left and right of the 4.0ng/ml threshold:

- Probability of biopsy
- Probability of diagnosis
- Probability of treatments: prostatectomies, chemotherapy, radiation, hormonal therapy
- Probability of death from prostate cancer
- Probability of death