Personal Health Train for Radiation Oncology in India and The Netherlands

"Big data analytics in cancer care holds immense potential to unlock valuable clinical insights from an abundance of patient medical records, aided by sophisticated statistical models, that will lead to improved population-based outcomes and deeper personalization of cancer treatment. However, the clinical data (which includes medical images, clinical examinations and laboratory results) has been locked away in disconnected "silos" within every clinic. Additionally, patient information is exceedingly sensitive to privacy issues and confidentiality breaches.

The investigators have pioneered the innovative Personal Health Train approach, whereby support for choosing the best treatment (i.e. decision support) is accessible without any patient records ever leaving the clinic of origin. This extends our current work on an extensible data architecture to learn from quantitative imaging data in India and The Netherlands (without images being taken out of the clinic) - NWO/DeITy BIONIC. The investigators have now developed numerous models of clinical outcome after treatment, including those for undesirable side-effects of treatment. The investigators continue to lead big data integration work within multicenter clinical decision support projects such as KWF-ProTraIT and Horizon2020-BD2DECIDE.

The overall aim of the TRAIN project is to combine big data (including images, laboratory tests and clinical examinations) to improve the outcomes for head & neck cancer patients in both India and The Netherlands. The investigators will do this by creating data-driven Decision Support Systems to predict which treatment gives the best outcome given individual patient characteristics, and local diagnostic and treatment capabilities. Cancer specialists in both countries will lead the design and clinical evaluation of this decision support system, which could be deployed in multiple clinics across all of the settings encountered in India and The Netherlands.

Head and neck cancer is a relatively rare condition in the Netherlands, such that the data volume available to learn from is much smaller than in India. Conversely, Indian patients typically present at a more advanced stage of cancer compared to Dutch patients. These differences in patients and treatments can be leveraged by machine learning algorithms to learn better predictive models. Decision support systems are essential, since guideline deviations in both countries are common due to individual patient characteristics, patient preferences and uneven distribution of treatment capacity outside major urban centers.

To achieve the above, The investigators first deploy the ICT infrastructure (in collaboration with Philips India) to connect local hospital information systems so that clinical, imaging and outcome data on head & neck cancer patients becomes findable, accessible, interoperable and reusable (FAIR) big data. The investigators then deploy learning algorithms that traverse the big data repositories of each participating hospital, using the privacy-preserving Personal Health Train approach, to develop a decision support system. Cancer specialists in India and the Netherlands will jointly evaluate the clinical utility of the decision support system by means of a prospective randomized clinical trial."

Inclusion criteria

• Histologically confirmed diagnosis of head and neck squamous cell carcinoma (HNSCC)
• Head and Neck primary tumor site: oral cavity, oropharynx, larynx or hypopharynx
• Treated between 01-2008 and 12-2017
• Clinical stage III and IV (a, b) according to TNM 7th edition
• No distant metastases (M0)
• Treated with curative intent: primary definitive radiation therapy with or without systemic treatment
• Availability of baseline imaging:

• Planning CT scan of the HN region in treatment position, with RT-structures available, performed with contiguous cuts of 2-3 mm or less in slice thickness with i.v. contrast
• (if available) FDG-PET in treatment position

Exclusion criteria

• Any previous HNC
• Patients with previous malignancies in the last 5 years before treatment for HNC, with the exception of surgically cured carcinoma in situ of the cervix, in situ breast cancer, incidental finding of stage T1a or T1b prostate cancer, and basal/squamous cell carcinoma of the skin
• Any previous malignancy that was treated with surgery and/or radiation of the head and neck region
• Histological type other than HNSCC
• Cancers originating in the oral cavity, nasopharynx, salivary glands or sinonasal area
• Postoperative radiation treatment setting