Mathematical algorithms with demonstrable clinical utility – the spark for personalized prostate cancer treatment?
Jonathan James | | Quick Read
Prostate cancer is a major malady – in fact, 8 percent of the worldwide cancer burden (1). Primarily a disease of old age, the prognosis for patients is variable; some tumors remain indolent for the rest of a person’s life, whereas others progress relentlessly. Characterizing this diverse disease continues to provide considerable challenges – and David Thurtle, Academic Clinical Fellow in the University of Cambridge’s School of Clinical Medicine, believes the resulting uncertainty hinders patient care. “We wanted to create something objective and standardized, capable of helping men in this difficult position,” he says. Now, in a paper published in PLoS Medicine, Thurtle and his team present a new model – PREDICT – intended to provide a better theoretical basis for clinical decision-making (2).
PREDICT is not Thurtle’s first prognostic model. Previously, in collaboration with Professor of Cancer Epidemiology Paul Pharoah, he developed effective tools for breast cancer decision-making (3,4) – experience that gave him confidence in PREDICT’s success. “We had the clinical expertise and, coupled with the clear clinical need, the goal was straightforward – to merge the two spheres of knowledge,” says Thurtle.
Of course, prostate cancer prognosis is no simple equation to crack. Patient medical history, comorbidities, and tumor-specific markers, such as PSA and the Gleason score, are already in routine clinical use. Individually, these indicators aren’t typically sufficient to make an informed decision about patient care – but attempts to create systems capable of providing a more comprehensive overview have failed. Thurtle believes the focus on the underlying mathematical principles of such approaches, with little or no consideration for ease of use in the clinic, has been a major contributing factor. “A lot of people have made such tools in the past, but these often only remain useful on paper – which, as you can imagine, is not very clinically helpful,” he says.
To overcome this hurdle, the team collaborated with the Winston Centre for Risk and Evidence Communication (WCREC) (5) in Cambridge – an organization specializing in the conveyance of risk and statistics to patients and the public. “Working with the WCREC, we were able to create what is, hopefully, a very user-friendly website (6),” says Thurtle.
The tool is also freely available to anyone online – so some may worry that its results could easily be misinterpreted by patients. That’s why Thurtle is adamant that medical professionals will always remain a crucial component of the decision-making process. “The intention of the tool is to empower patients and clinicians to think more broadly, rather than jumping towards a treatment modality prematurely,” he says. “The hope is that the tool would be used in the clinic, either at the point of diagnosis or shortly thereafter.”
To reach that stage, Thurtle acknowledges there’s still some way to go. “We need to gain a little more confidence that the tool is accurate,” he says. That will require further testing in independent cohorts. He hopes such an approach will also incorporate different ethnicities, too. “So far, we’ve focused on Caucasian groups, so it would be a good idea to test it out in other ethnic groups,” says Thurtle. “Then, over time, as the data matures, we’d like to start plugging additional variables into our models to assess their effect on prognostic power.”
Perhaps the most significant aspect of the model is that it assesses overall survival, rather than cancer-specific survival. “It’s all very well and good curing someone of prostate cancer,” Thurtle says, “but if they are going to die of something else – whether it be their age or other comorbidities – then it is a bit misleading to tell them they can be cured of prostate cancer when it’s essentially irrelevant to their overall health.”
- World Cancer Research Fund – American Institute for Cancer Research, “Worldwide Cancer Data: Global Cancer Statistics for the most common Cancers” (2019). Available at: bit.ly/2S1ZynX. Accessed May 17, 2019.
- DR Thurtle et al., “Individual prognosis at diagnosis in nonmetastatic prostate cancer: Development and external validation of the PREDICT Prostate multivariable model,” PLoS Med, 16, e1002758 (eCollection 2019 Mar) (2019). PMID: 30860997.
- PD Pharoah et al., “Polygenes, risk prediction, and targeted prevention of breast cancer,” N Engl J Med, 26, 2796-2803 (2008). PMID: 18579814.
- DF Easton et al., “Gene-panel sequencing and the prediction of breast-cancer risk,” N Engl J Med, 4, 2243-2257 (2015). PMID: 26014596.
- University of Cambridge, “Winton Centre for Risk and Evidence Communication” (2019). Available at: bit.ly/2JMyXa3. Accessed May 16, 2019.
- Predict Prostate – NHS, “Predict Prostate: An individualized prognostic model for men newly diagnosed with non-metastatic prostate cancer,” (2019) Available at: bit.ly/30uGCjl. Accessed May 17, 2019.