PRECISE CURATE.AI trial: Dr Dean Ho and Dr Raghav Sundar on pilot study exploring guided dosing with XELOX, XELIRI and single agent capecitabine to treat solid tumours
Cara Yap, 1 March 2021
Study start date: August 2020
The Phase I and II single-arm pilot study with multiple cohorts of patient populations aims to demonstrate the feasibility of applying CURATE.AI. It explores the recommended guided dosing for 10 subjects who will receive palliative-intent chemotherapy with one of 3 regimens: XELOX, XELIRI, and single agent capecitabine. CURATE.AI - a small data, AI-derived technology platform - allows personalised guidance and optimisation of an individual's dose modulations based only on that individual's data. The study is being conducted at the N.1 Institute for Health and Institute for Digital Medicine, National University of Singapore, and National University Hospital, Singapore.
About the principal investigators
Prof Dean Ho (above), PhD, MSc, BSc, Head, Department of Biomedical Engineering; Director, The N.1 Institute for Health (N.1); Director, Institute for Digital Medicine (WisDM), National University of Singapore.
Prof Ho is Provost's Chair Professor at the National University of Singapore. He is the Co-founder of KYAN Therapeutics, Inc and has advised major corporations and venture funds in the areas of digital medicine and personalised healthcare. His research interests involve the use of AI-based platforms for N-of-1 medicine. He was elected to the US National Academy of Inventors in 2018, and is a Fellow of the American Institute for Medical and Biological Engineering.
Asst Prof Raghav Sundar (above), MBBS(S'pore), MRCP(UK), MMed(Int Med), MCI, FAMS (Oncology), Consultant, Department of Haematology-Oncology, National University Cancer Institute, Singapore, Assistant Professor, N.1 The Institute for Health, National University Singapore (NUS)
Asst Prof Raghav Sundar is a Consultant Medical Oncologist specialising in gastrointestinal cancers and early phase clinical trials. His research interests are in developing novel genomic biomarkers and precision oncology.
*Intended readers: healthcare and industry professionals
Tell us more about your ongoing trial. Why is it significant?
Prof Ho: This trial demonstrates the importance of seamlessly building a bridge between innovation and clinical implementation. One of the most promising aspects of this work is how well our team and Raghav’s amazing group of clinicians are designing the study together - not only implementing potentially improved dosing for patients, but trying to understand how we can actually deploy an AI-driven clinical decision support tool in an actual clinical work flow.
At the end of the day, the objective of the trial is to see if we can implement a CURATE.AI-driven dosing optimisation paradigm in the clinical setting. Not only will this allow us to understand the technical and medical implications, but also the implementation implications, which is really exciting.
What sort of patient data is used in calibrating his/her dosing?
Prof Ho: It's quite different from standard approaches that typically use demographics or genetic information, which are also exciting and complementary to what we do. Our calibration approach is done primarily through dosing information - namely, the corresponding longitudinal efficacy and safety measurements at each dosing level. It’s important to note that in the traditional use of data to drive treatment or diagnostics, a lot of the information is derived from conventional dosing, which involves fixed, high doses.
We work closely with clinicians, and the study hinges on their ability to understand how each patient responds to each dose level. The corresponding measurement of outcomes is used as data to drive patient care. The premise of the study is that while we certainly know that all patients are different from one another, individual patients change over time. Thus, it is important to keep this kind of running measurement of how each patient is responding dynamically.
Why is dose optimisation so important in oncology trials?
Asst Prof Raghav: We know that the current way in which Phase 1 clinical trials are designed is to always identify the maximum tolerable dose and recommended Phase 2 dose, without factoring efficacy outcomes in these decisions. While a lot of the newer Phase 1 trials do look at efficacy to see whether the drug works and at lower doses, the final dose that is rolled out to later phase studies is typically the highest dose we can give the patient. We are questioning if there is a need to give the same high dose to every patient, whether we can actually give each patient a dose that is bespoke to them, and what information we can use to guide these decisions. This is such a critical question that Phase 1 clinical trial pioneer Dr Mark Ratain from the United States recently wrote an article discussing why it is important for the government to empower the Food and Drug Administration (FDA) to mandate dose optimisation for all oncological drugs. This is because in the era of targeted therapy and immunotherapy, we are increasingly aware that every patient may not need to receive the highest dose to get the best outcomes.
Tell us a bit more about how this study was designed
Asst Prof Raghav: There are essentially no prior studies that look at differences in efficacy in relation to doses at the individualised patient level, so this led us to design a study that explores changing a patient’s chemotherapy dose or systemic therapy dose with every cycle of treatment. CURATE.AI presented the perfect model that enters clinical data into an algorithm and arrives at an output parameter that we can immediately implement, without involving complicated pharmacokinetic blood draws over multiple time points. Because this is a pilot study, we are looking at making sure that our logistical work flows are easily implemented, and patients are comfortable coming in for blood draws to measure their tumour marker levels, as well as having their doses modulated at every cycle of chemotherapy.
For the first few cycles, we enter different biomarker outputs into the algorithms so they get tuned and once that happens, the algorithms are able to predict a dose that will be optimal for the patient, which we can then use for future cycles. For the first two cycles, we ask the patient to come in for a weekly blood draw, and we only adjust the chemotherapy doses every cycle, as opposed to weekly. The additional blood draws give us data points to feed the algorithm so that it essentially triangulates the best dose for the patient.
What specific outcomes do you hope to achieve from this trial?
Prof Ho: The CURATE.AI implementation process does not use population-wide data. Especially in combination therapy, drug synergy is dose dependent, time dependent and patient specific. If you look at pharmacokinetics, there’s this notion that reaching a blood threshold level is accompanied by threshold efficacy. But as we have seen through our previous n-of-1 studies (where each subject receives a personalised dosing regimen), there are cases where the closer you dose to that threshold level, the worse off the patient is. Our goal is not to low dose everybody, but to understand this process of calibrating each patient’s dose.
One of the most promising outcomes we can aim for is where a patient who doesn’t appear to respond to a standard dose and would typically be moved to another line of treatment, appears to respond to a lower dose instead. This opens up a whole other series of things we could learn about. There’s an opportunity to understand how n-of-1 calibration can lead to n-of-1 profiles that are actionable for patients, then we can potentially capture more n-of-1 responders to treatment where initial treatment parameters would have suggested otherwise. That’s one of our objectives.
How exactly are your various teams collaborating to ensure the study meets its objectives?
Asst Prof Raghav: Because this is the pilot study, one of the main questions we asked is how easy it is for us to get data across to the N.1 team. What sort of data do they need, and how frequently do they need it? We are still trying to hammer out these sort of questions, but the data work flow is getting smoother at this point in time. It is almost like a standard clinical trial, where we de-identify the clinical data, enter it in an electronic CRF and send it over to the N.1 team to analyse. Eventually, this is going to progress to the clinical implementation stage, and it won’t be difficult translating this into a clinical decision support system. Interaction with the N.1 team is absolutely critical right now, because we are trying to figure out things like how to relate information - such as when a patient suffers severe toxicities - over to the technical team. This is to ensure that the dose predicted is not going to be higher than the current one.
Prof Ho, what was the basis for CURATE.AI?
Prof Ho: Back in the day, our team was doing a lot of research on therapeutic delivery with biomaterials. At that time, a lot of it was directed at single agent therapy. Over time, combination therapy - a mainstay for cancer treatment as well as many other diseases - came into play. The notion of therapies evolving synergistically caused us to put new materials on the back burner, because if you put two drugs into a material and lock in the ratio, that often closes the door to studying the interaction of these two agents dynamically. And so we started exploring different things, such as developing new combinations or dosing differently in patients. Using retrospective analysis, we found evidence that different patients could benefit from different doses, which eventually transitioned into pre-clinical studies and a first in human trial for liver transplant patients.
What else should physicians and cancer patients take note of, with regard to this trial?
Asst Prof Raghav: At this point of time, two arms of the trial are open - one enrols patients who are on capecitabine-based regimens - this would use capecitabine as a single agent or a combination of oxaliplatin or irinotecan, with or without bevacizumab. The second arm, which we recently opened, looks at patients with waldenstrom macroglobulinemia who are going to be treated with imatinib. Moving forward, we are likely to open further arms of the study based on modulating chemotherapy doses in disease types with raised tumour markers such as CEA and CEA 125. Because we know that a lot of cancers are not tumour marker secretors, and also that traditional tumour markers are not extremely good parameters of efficacy, we are also starting to measure circulating free DNA or circulating tumour DNA. However, this is in the exploration phase, as that part of the study is expensive to conduct. Otherwise, we are looking at expanding this study to other sites, so we are happy for the latter to reach out to us.
Prof Ho: Compliance and safety are top priority. As innovations are developed, it is really important that all the stakeholders involved - from tech developers to clinicians and medical ethics board - assess all of that thoroughly. We spend a lot of time communicating to all the stakeholders what the tech is, how the safeguards are built, and previous compliance results. Patient safety is number one. What’s important and exciting is that we can learn together through continued communication.
Wish to receive updates on cancer clinical trials? Click here to sign up for our free trial notification service.
*The content in this article is purely educational and written for healthcare professionals. It does not contain forward-looking statements, or those specific to commercial enterprise.