Using deep learning to guide the selection of chemotherapy after surgery for colorectal cancer patients

Among patients with colorectal cancer spread to lymph nodes, a deep learning marker can identify one-third with excellent prognosis and consequently no need for adjuvant chemotherapy. Double-agent chemotherapy is the current standard of care for these patients and may cause severe side effects and even death.

New research shows how deep learning can be used to guide the selection of chemotherapy after surgery for colorectal cancer patients. In particular, many patients currently treated with adjuvant chemotherapy can safely avoid such treatment.

Applying the method in clinical practice may thence reduce morbidity and costs associated with the treatment of colorectal cancer and perhaps also prevent deaths by allowing chemotherapy to be offered earlier to patients who actually need it."

Professor Håvard E. Greger Danielsen

Danielsen is the leader of the research study, the director of the Institute for Cancer Genetics and Informatics (ICGI) at Oslo University Hospital, a Professor at the Department of Informatics (IFI) at the University of Oslo, and a Visiting Professor at the University of Oxford.

No benefit from treatment

Colorectal cancer is one of the most common cancer types, particularly in western countries. The tumor is typically resected, and patients at increased risk of recurrence and cancer death are usually offered adjuvant chemotherapy.

- Because current methods cannot precisely predict which patients need chemotherapy, the current standard of care is to offer adjuvant chemotherapy to large patient groups that include many patients who will not benefit from the treatment, Danielsen says.

For patients where cancer has spread to lymph nodes, the current standard of care is double-agent chemotherapy because it has been shown to benefit more patients than less aggressive treatments, but it is recognized that about half the patients would not have needed adjuvant chemotherapy at all. Neuropathy is a common side effect of double-agent chemotherapy and can sometimes be both severe and prevalent years after treatment is ended. In some cases, the chemotherapy might even cause death.

Utilizing deep learning

Two years ago, a study in The Lancet demonstrated that deep learning can be used to predict which patients are more likely to die from colorectal cancer after surgery. Building on this finding, the research recently published in The Lancet Oncology shows how to integrate the deep learning marker with the markers currently used in clinical practice and that the combination can be used to administer adjuvant chemotherapy better by tailoring the treatment to patients who actually need it.

Among patients with colorectal cancer spread to lymph nodes, the new method for selecting adjuvant chemotherapy can identify one-third who should not receive adjuvant chemotherapy even though the current standard of care is double-agent chemotherapy. This is because these patients have excellent prognoses, similar to those without spread to lymph nodes that are not treated with adjuvant chemotherapy according to the current standard of care. Actually, treating this one-third with double-agent chemotherapy is expected to cause a similar number of treatment-related deaths as the number of cancer deaths prevented by the treatment.

Dr Andreas Kleppe notes that not treating these patients with adjuvant chemotherapy will allow them to recover sooner from the cancer treatment, and they will not have to experience as significant side effects. Kleppe is first author of the new research paper, a researcher at ICGI, and an Associate Professor at IFI.

More individualized treatment

The new method is also designed to indicate how aggressive the adjuvant chemotherapy needs to be and to guide treatment decisions for colorectal cancer patients without spread to lymph nodes. Some patients without spread are recommended adjuvant chemotherapy according to the current standard of care, but only a fraction of these will benefit from the treatment because most do not need any adjuvant treatment, while a few could perhaps benefit from a more aggressive adjuvant chemotherapy such as double-agent chemotherapy.

- A more accurate determination of which patients need adjuvant chemotherapy, and how aggressive the treatment should be, may result in improved quality of life for many patients and possibly prevent some cancer deaths, Kleppe explains.
Chemotherapy entails substantial financial costs related to drugs, medical personnel, and sick leaves. Applying the new method to enable a more individualized selection of adjuvant chemotherapy will therefore improve the cost-benefit ratio of colorectal cancer treatment.

The new method was developed and validated as a part of an IKTPLUSS Lighthouse program funded by The Research Council of Norway, DoMore!, which aimed at applying artificial intelligence to improve the current treatment of cancer. The method is ready to be applied in clinical practice and has recently been CE marked for use in European countries.

Journal reference:

Kleppe, A., et al. (2022) A clinical decision support system optimising adjuvant chemotherapy for colorectal cancers by integrating deep learning and pathological staging markers: a development and validation study. The Lancet Oncology.


The opinions expressed here are the views of the writer and do not necessarily reflect the views and opinions of News Medical.
Post a new comment
You might also like...
Potential breakthrough in breast cancer: Five proteins identified as risk markers