Using decision support technologies to enhance individual decision-making to forgo or undergo screening for prostate cancer
Decision support tools such as nomograms for screening outcomes and decision analytical models that formally assess the relative advantage and disadvantage of screening are likely to be helpful for patients and physicians in making a screening decision. In the case of screening, there significant trade-offs between potential benefits and harms that are likely to be sensitive to a man's values and preferences.
We endeavor to develop nomograms for outcomes that may occur as a consequence of a man's decision to undergo or forgo screening based on the characteristics and outcomes of men enrolled in the PLCO and Göteborg trials. The reason for including these two trials is that extensive information on baseline features (e.g. age, ethnicity, family history, comorbid conditions, prior PSA testing) are available that were not captured in ERSPC. The outcomes that we will endeavor to predict will include: 1) prostate cancer diagnosis (including high-, intermediate-, and low-risk cancer), 2) metastatic cancer, 3) cancer-specific mortality, and 4) death from competing causes. Separate models will be developed for the screening and control arms of these trials.
We endeavor to develop a Markov-based decision analytical model that considers individualized predictions of outcomes that may occur as a consequence of a man's decision to undergo/forgo screening (from our nomograms) and patient-specific utilities (by standard-gamble) for these outcomes. The decision model will estimate the quality-adjusted life years associated with screening vs. no screening based on a man's unique features and preferences.
In a pilot study in a primary care setting, we will assess the feasibility of applying a formal screening decision model and determine if it enhances knowledge and decision quality compared to providing information on specific outcomes probabilities alone from our nomograms and compare both these methods to standard counseling.
We have conducted a similar study to implement a Markov-based decision analytical model for prostate cancer therapy for men diagnosed with localized prostate cancer. We have utilized probabilities from well-validated nomograms for cancer and quality-of-life outcomes associated with the standard treatments, and patient-specific preferences for these outcomes using standard gamble to estimate the QALY with each intervention. In a randomized trial, we showed that the decision model and a table of nomogram probabilities significantly improved decision quality and decisional regret at 2 years. We also showed a significant increase in the choice of expectant management among men enrolled.
Our pilot study highlights the potential of these two approaches (individualized outcome probabilities from nomograms and formal decison analytical approaches) to influence patient decision-making by encouraging them to pursue interventions that they value and forgoing treatment/testing that they do not value.
1. To develop a decision aid for prostate cancer screening using decision analytical methods, that considers personalized probabilities of outcomes (health states and survival) that may occur as a result of a man’s decision to undergo or forgo screening and patient-specific preferences for these health states. The decision aid will estimate the quality-adjusted life years (QALY) associated with screening versus no screening that reflect a man’s unique features and preferences. Thus, the decision whether or not to undergo screening for prostate cancer will be tailored to a man’s values.
a. We will develop prediction tools (screening nomograms) to estimate the probabilities of important outcomes that may occur as a consequence of screening versus no screening. These five outcomes will include: 1) prostate cancer diagnosis (including high-, intermediate-, and low-risk cancer), 2) metastatic prostate cancer, 3) prostate cancer-specific mortality, and 4) death from competing causes. We will predict these outcomes based on a man’s unique features known at the time of screening (age, ethnicity, comorbidity, family history, etc) using the data from two large prostate cancer screening trials from Göteborg, Sweden and the United States (PLCO).
b. We will develop a computer-based screening decision aid for prostate cancer screening using a Markov-based decision analytical modeling approach that considers individualized predictions for the 5 cancer-related outcomes from the screening nomograms (Aim 1a) and 4 treatment-related side-effects (bladder, bowel, and sexual dysfunction and urinary incontinence) from quality-of-life nomograms we have previously developed for prostatectomy, radiation therapy, and brachytherapy. The decision aid will also consider patient-specific preferences for 6 health states (prostate cancer diagnosis, metastatic prostate cancer, and treatment-related side-effects) obtained using standard gamble to estimate the quality-adjusted life years (QALY) associated with screening vs. no screening.
2. To assess the feasibility of implementing the computer-based screening decision aid to facilitate decision-making for prostate cancer screening in a primary care setting. We will assess the feasibility of implementing the screening decision aid in a primary care setting. In a staged intervention, we will evaluate whether a quantitative approach to patient preferences in our screening decision aid (Aim 1b) enhances knowledge and decision quality compared to providing information on specific outcomes probabilities alone (Aim 1a) and compare both these methods to standard counseling. If we are able to administer the model in a busy clinical practice setting, establish patients willingness to participate in this study and use this additional information to make treatment decisions, and determine our ability to reliably and accurately obtain responses for our outcome measures at specified time points, we intend to test the screening decision aid in a multi-institutional, phase III randomized trial (compared to standard counseling with or without information on specific outcome probabilities) in a future study.
Dr. Sigrid Carlsson, Memorial Sloan-Kettering Cancer Center, New York, NY USA carlssos@mskcc.org
Dr. Hans Lilja, Memorial Sloan-Kettering Cancer Center, New York, NY USA liljah@mskcc.org
Dr. Erik Holmberg, Sahlgrenska Academy at the University of Göteborg, Göteborg, Sweden erik.holmberg@oc.gu.se
Dr. Jonas Hugosson, Sahlgrenska Academy at the University of Göteborg, Göteborg, Sweden jonas@urol.se
Dr. Michael Kattan, Cleveland Clinic, Cleveland, OH USA kattanm@ccf.org
Dr. Li Li, Associate Professor of Family Medicine, Epidemiology and Biostatistics, and Oncology, Case Western Reserve University School of Medicine, Cleveland, OH USA ll134q@rocketmail.com
-
Association of Baseline Prostate-Specific Antigen Level With Long-term Diagnosis of Clinically Significant Prostate Cancer Among Patients Aged 55 to 60 Years: A Secondary Analysis of a Cohort in the Prostate, Lung, Colorectal, and Ovarian (PLCO) Cancer Screening Trial.
Kovac E, Carlsson SV, Lilja H, Hugosson J, Kattan MW, Holmberg E, Stephenson AJ
JAMA Netw Open. 2020 Jan 3; Volume 3 (Issue 1): Pages e1919284 PUBMED