My study that I’m presenting at AACR is looking at ctDNA and how we can actually use that to predict time to progression in patients and also adapt treatment. So we want to see if ctDNA would help with dose de-escalation and escalation in patients.

Oftentimes tumour volume is what is being used widely in the clinic and ctDNA is actually a liquid biopsy biomarker which can be collected much more frequently. We are saying that the results of that, because of the frequent collections, we can actually say that it can actually be a good surrogate for tumour volume by telling us, before even imaging can tell us something, to see if the patient’s tumour would recede, will it grow, and how they are also responding on treatment. We want to see, for instance, if the ctDNA is increasing very fast we know that maybe that patient will have to either dose escalate or if we can see that the ctDNA is responding, with the frequent collections if it’s reducing, we know that that patient would actually be in a dose de-escalation.

So I developed a math model of tumour volume and ctDNA which is actually looking at this tumour volume and then ctDNA interactions to actually predict this time to progression and see which patients would benefit from dose escalation and de-escalation.

What were the results of this study?

The study looked at 32 patients who were HPV positive and they had anal cancer. That’s what the study design is but we only took 26 out of these 32 patients who had good ctDNA and tumour volume. So we built the model around these 26 patients and so far we can see that we can characterise it in 14 non-responders and 12 responders.

The model works and fits these 26 patients and then we can see that the ctDNA is actually rightly following the volume. So what we’ve seen also is that we can also predict the next timepoint for the volume for these ctDNA patients.

So I built a math model which is a three-equation ODE model which uses the ctDNA as the real-time GPS for tracking treatment response in the patients. These patients are HPV-positive anal cancer patients and there were 32 on the trial but we used 26 because they had complete tumour volume and then ctDNA response. So 26 out of these 32 patients, we divided them into 12 responders and then 14 non-responders. The main results that we could see were ctDNA was actually able to correlate with the tumour volume at baseline. So most of the time at baseline is just when treatment begins or the first initial collections. So we realised that we had a good p-value of less than 0.001 and a good variation between the responders and the non-responders.

So from there what we wanted to see was that even though we could see a good signal there, what if we go further? So we went further to look at week 3 and then we saw that we could see a good variation actually between the responders and the non-responders from baseline to week 3. We could also see that the model fits the data properly for the tumour volume. We had good R-squared values of 0.99 for the volume and then 0.91 for the ctDNA.

So, so far the math model works and then it can actually capture the tumour growth and the treatment sensitivity over time and then the ctDNA shedding of the tumour.

What do you think is the importance of these results and what is next for this study?

The guiding philosophy of this work is that we want to create the same cure, less suffering. So that’s how I coined the term. The point is that if we can use ctDNA as a surrogate for what is happening in the tumour in real-time. Normally the turnaround for tumour volume is about 6-8 weeks and sometimes it can take longer for the patient. So ctDNA, because it’s collected much more frequently through blood draws, you can get it much more frequently, so we can potentially adapt treatment earlier before even treatment confirmation.

For patients with rare cancers like anal cancer, with what I’m doing and they’re on treatments like immunotherapy, this could mean that we can catch non-response much more sooner before waiting for the next scan. The whole goal of this is all about making precision-medicine more accessible and responsive much faster.

The next steps are actually very exciting. Because this is a math model and I know we are interested in trying to see how we can translate the math into the clinic, we actually want to go into a deeper dive into the parameters of the model. So every patient is different so we realised that we are trying to build the model for each patient, to make it patient specific. Then it would actually give us a much more informed parsimonious idea of what the model does.

The second goal is that we want to leverage on this frequent ctDNA collection and then try to see if we have a lot of ctDNA points and then maybe we have very little volume points, where are the next volume points? So those are the main two big goals.

Right now ctDNA is not going to become just a biomarker or simple but because of the math involved we can leverage it for real-time adaptation. Maybe in the future we’ll try to generalise it for all cancers as well.