Cancer Patients, Prognoses, Oncologists, and Computers

Can a computer outperform an oncologist in making prognostic predictions in cancer patients or in determining the right treatment options for cancer patients? Would you even be interested in finding out?

Such predictions are of enormous value. They can help doctors determine what treatment options might benefit patients the most, which options might hasten their decline, and who might or might not be a good candidate for a clinical trial, among other benefits.

However, these preditions require an overwhelming amount of data – too much for any one doctor to calculate in a week, even presuming he or she had nothing else to do that week.

Dr. Cary Oberije, a postdoctoral researcher at the MAASTRO Clinic, Maastricht University Medical Hospital in The Netherlands, and colleagues, carried out a study of 121 lung cancer patients using a computer model and mathematical analysis of cancer data that takes into account a patient's genetic make-up, cancer stage, tumor type, and other relevant data in order to predict which of the lung cancer patients would:

• still be alive two years later
• experience dyspnea (shortness of breath)
• experience dysphagia (difficulty swallowing)

Oberije reported the findings last week at the Forum of the European Society for Radiotherapy and Oncology (ESTRO).

So how did their computer do against predictions made by the patients' own doctors?

It beat them – kicked their asses, actually – in every category. In fact, the predictions made by the doctors were so poor they were regarded as being little better than chance. Oberije commented:

In our opinion, individualized treatment can only succeed if prediction models are used in clinical practice. We have shown that current models already outperform doctors. Therefore, this study can be used as a strong argument in favor of using prediction models and changing current clinical practice. We know that there are many factors that play a role in the prognosis of patients and prediction models can combine them all ... Our study shows that it is very unlikely that a doctor can outperform a model.

Of course, patients wouldn't be meeting with a computer in the exam room. They wouldn't have a robot performing an exam. Rather, the study was carried out in an effort to show that prognoses in oncology are extraordinarily difficult, and that they require an astonishing amount of information and calculations even to be close to being correct, and that patients and oncologists alike would benefit from having such calculations handy. Nobody is saying a computer would replace a doctor – only that a doctor could use some help.

And let's face it: When we consider, outside of blood cancers, the overall lack of true progress made against most cancers in terms of extended survival and cure rates since Nixon launched the War on Cancer in 1971, doctors need all of the help they can get.