Senior Medical Director, Brainomix
Consultant Pulmonologist and ILD Clinical Lead,
Royal Brompton Hospital
Spirometry has been the backbone of fibrotic lung disease monitoring for decades. But its 10 to 15 percent day-to-day variability, and its inability to capture early structural change, means clinicians have long been managing a condition they cannot fully see. Brainomix is making the case that AI-powered quantitative CT can fill that gap, and the evidence is accumulating fast. The company’s e-Lung platform is now FDA-cleared and CE-marked, has been validated across multiple datasets including the landmark INBUILD trial in partnership with Boehringer Ingelheim, and is about to be embedded into 20 U.S. centers through a prospective outcomes study called PROGRESS PPF. At the center of that evidence-building effort is Dr. Peter George, consultant pulmonologist at Royal Brompton Hospital and Senior Medical Director at Brainomix, who has been shaping the clinical case for e-Lung since the platform was still just a concept. We spoke with him about what the data actually shows, where the technology sits today, and what has to be true for quantitative CT to become standard of care.
How did a specific patient case first convince you that quantitative CT was capturing something spirometry was missing?
Peter George: I’ve been looking after patients with fibrotic lung disease for over 15 years, and the challenge we have with these conditions is that our previous gold standard tools have always been CT with visual assessment and lung function, which comprises spirometry and gas transfer measurement.
We’ve known for many years that spirometry carries variability of around 10 percent day-to-day and lab-to-lab, and gas transfer, the DLCO, is affected by about 15 percent variability. So spirometry is insensitive to small changes in disease progression, and it’s a volitional test. If patients are having a bad day, it can underestimate the true physiology.
What clinched it for me was a specific case where a patient had stable lung function tests. The forced vital capacity hadn’t changed in six months. But they were feeling more breathless, and the serial CT scans showed extensive disease which the radiologist reported as being stable. When disease is extensive on a scan, it’s very difficult for radiologists to identify progressive fibrosis. We then applied the e-Lung technology to that case, and what we found was that the quantitative CT biomarkers had in fact risen over that period, at the exact time the patient was feeling more breathless, despite the spirometry being stable.
You could ask, well, how do you know which is more accurate? When we asked our radiologists to re-evaluate the scan with the benefit of the quantitative AI-powered technology, and when they did so, they revised the report and agreed that they could see that progression. What’s special about the tool is that it’s very explainable. When the patient has a scan, the imaging biomarkers light up on the scan in different colors, and the radiologist can go back to that area of abnormality and evaluate whether they can now see the areas of fibrosis progression which correspond to the colour and can then determine whether they believe that thereis true progression. That, I think, is when it really hit me that this technology is going to make a difference for patients in the long term.
“The quantitative CT biomarkers had in fact risen over that period, at the exact time the patient was feeling more breathless, despite the spirometry being stable.”
Why did the taladegib CT analysis focus on lung volume and fibrosis extent rather than anchor to FVC as the primary reference?
Peter George: The original taladegib study has already been published in the Lancet Respiratory Medicine. It was a phase 2a study, so primarily a safety and tolerability study, but it did have an efficacy endpoint included within it, and that study showed that taladegib was associated with a significant reduction in FVC change over 12 weeks compared to placebo. So we already knew there was a signal with FVC.
But what FVC doesn’t tell you is about drug mechanism of action, and about different structural elements of interstitial lung disease. So we then looked at the CT scans, which were performed as part of that study at baseline and at the end of the study (week 12). What we were able to show is that in line with the deceleration of FVC decline, we could find changes in different anatomical lung compartments that represented these abnormalities. We found a significant rise in lung volume, a significant reduction in total interstitial lung disease extent, a significant reduction in fibrosis extent. So this shows us that we were able to not just identify the change that had already been published using FVC, but to identify different pathological processes that might explain that spirometric finding.
Why did the INBUILD post-hoc analysis structure its prediction around six-month CT changes rather than later timepoints?
Peter George: The INBUILD study was a landmark study, the first to show that an antifibrotic drug could be used in patients with non-IPF progressive pulmonary fibrosis. As part of that study, a subset of patients had a baseline CT, a CT at six months, and an end of study CT at 12 months. Brainomix, through its close relationship with Boehringer Ingelheim, had privileged access to those CT scans, and that allowed us to work collaboratively to see what further information we could extract from this pivotal study.
What’s particularly exciting is that this is one example of the growing number of collaborations we’ve built with Life Sciences companies over the years. These partnerships allow us to bring our quantitative CT technology into clinical development programmes to accelerate the delivery of new drugs to the clinic and to extract additional mechanistic insights from imaging data that might otherwise be missed.
What we found in this INBUILD post-hoc analysis was that the change in quantitative CT metrics at 24 weeks could predict FVC decline at 52 weeks. What’s important about that is it shows us that e-Lung quantitative CT biomarkers measure changes in fibrosis predicting lung function decline 6 months earlier than a trial readout. It means that we may be in a position to advance decision-making in clinical trials from 12 months to six months in some settings. It might help us in terms of thinking about go, no-go decisions in earlier phase clinical trials.
How do you guard against the possibility that the model is fitting to historical datasets rather than capturing true disease biology?
Peter George: What is unique about e-Lung and the e-Lung biomarkers is that they are FDA-cleared and CE-marked. They are now locked andas a consequence, when we apply them to clinical trial datasets or to routine clinical practice, the fact that they cannot be trained against the additional data they are being tested upon guards against that possibility of overfitting. These thresholds and levels have been validated extensively across a number of publications over the past few years.
So whereas that might have been a reasonable assertion in our very early work, the fact that these biomarkers continue to be effective in identifying progressive disease and demonstrate a treatment effect in clinical trials some years on shows us the validity and the strength of their potential.
How do you see quantitative CT biomarkers sitting in the ILD diagnostic pathway five years from now, and what would have to be true for that to happen?
Peter George: That really comes down to implementation science and we are now in the implementation phase. We have extensive scientific validation of the tool and the biomarkers. We’ve shown they can identify progressive pulmonary fibrosis more readily than visual assessment or lung function change. We’ve shown that change in e-Lung measures of fibrosis and interstitial lung disease are predictive of mortality. And we have shown that e-Lung biomarkers predict patients at risk of future mortality and lung function decline. Brainomix has a strong track record in widespread clinical adoption through its stroke activity and this is now being replicated across interstitial lung disease networks – The tool is cleared for clinical practice in both Europe and the US.
We’ve also recently had a study accepted for publication, called the REVISE PPF study, where we were able to show that had e-Lung been embedded in any one of three centers we worked with;University of Chicago, University of Alabama, and Weill Cornell Medicine, patients would have been diagnosed with progressive pulmonary fibrosis up to two years earlier. The tool is cleared for clinical practice in both Europe and the US.
We are now embarking on an ambitious multi-center study in the US called PROGRESS PPF, where we will embed e-Lung into 20 US centers and will evaluate the clinical impact on patient outcomes. I think this will be really important work that we reference when we look to then scale the technology and move towards more widespread clinical adoption. I think we are on the cusp of a sea change in how these interstitial lung diseases are managed and the routine integration of AI-powered tools to improve diagnostic efficacy and eventually patient outcomes.
How does a clinician actually navigate the conversation with an IPF patient who presents with both genetic risk signals and a high CT fibrosis burden today?
Peter George: What we are now doing building multi-modal predictive models. We want to gather important prognostic information from an individual patient to provide a personalized risk stratification approach for that individual.
Most centers at this stage do not routinely measure telomere lengths, and most centers at this stage do not routinely use e-Lung as a quantitative CT tool. But I could envisage in the next 5 to 10 years a situation where we are integrating e-Lung quantitative CT data with multi-omic biomarkers, genetic information and telomere lengths with the patient at the very heart to provide a much more nuanced approach to individualized treatment plans, risk stratification for outcomes and assessment of treatment efficacy. Although I don’t think we’re there yet, I do think that’s where the future may lie.
“I could envisage in the next 5 to 10 years a situation where we are integrating e-Lung quantitative CT data with multi-omic biomarkers, genetic information and telomere lengths with the patient at the very heart.”
Prof Peter George is a consultant pulmonologist and clinical lead for the Interstitial Lung Disease service at Royal Brompton Hospital, and Senior Medical Director at Brainomix.
Moe Alsumidaie is Chief Editor of The Clinical Trial Vanguard. Moe holds decades of experience in the clinical trials industry. Moe also serves as Head of Research at CliniBiz and Chief Data Scientist at Annex Clinical Corporation.
