Thank you very much for inviting me to speak to you today about molecular biomarkers for lung cancer screening. In the next 10 minutes, I'd like to touch on the clinical need for these biomarkers, a few biomarker principles that are important to keep in mind as we review the current status of these tests. First, the clinical need. Despite about a decade of standard of care lung cancer screening in the United States, still only around a quarter of all lung cancers are diagnosed in a localized stage. This is in part related to lung cancer screening eligibility and effectiveness, but also in part related to slow uptake of lung cancer screening, particularly in those who are healthiest and most likely to have the highest net benefit from screening, as well as low adherence to annual screening. In addition, the balance of benefits and harms is influenced by procedures performed on screen detected nodules. And out in practice, there's reports that the rate of these procedures is higher than and leads to more complications than had been reported in controlled trials. So the clinical need for molecular biomarkers includes a need to improve our ability to identify patients at risk, to identify patients with asymptomatic early lung cancer, to improve aspects of screening implementation, and to improve our management of lung nodules. I wanted to touch on a few biomarker principles to help us understand where we are today. Molecular biomarkers move through phases of development, discoveries when you find a single or a pattern of molecules that favors the presence of cancer versus the absence. Analytical validation is when the assay that measures those markers is assessed for its accuracy and repeatability. Clinical validation is when the marker is locked down, tested in the intended use, and you find out how accurate it is. And clinical utility has to do with how the biomarker leads to changes in decisions in practice. There's a hierarchy of evidence that's developed from the discovery through the clinical validation phases. Phase one studies are proof of concept studies. They usually use samples obtained retrospectively, and those samples may not have been collected as you need, they may not have been in the intended use population, they may not have collected metadata that allows you to look at subgroups or control for confounders, but they lead to enough promise to invest in the next phase of study. Phase two studies, clinical grade for lab-developed tests, use a controlled sampling. You enroll only those in the intended use, and you collect enough data to do subgroup analysis and so on. A phase four study, an FDA-level study, is a prospective collection of individuals, in this case those who might be going for lung cancer screening and following them over time to determine the accuracy. You see, I skipped phase three. It's between phase two and four there, where if you get really lucky and there's a retrospective sample set in your intended use, collected just as you needed with all the data, that would be a phase three study. Ideally, after validation, you get an assessment of clinical utility. How does the test affect subsequent clinical decisions and outcomes? A clinical utility study has to include a balance of the benefit from true positives and true negatives and the harms from false positives and false negatives. How you look at that balance and what accuracy leads to a clinically useful test depends a lot on the intended use of the test. So take, for example, if the intended use was to augment uptake and adherence in the currently screen-eligible patient, where we'll say it takes 120 individuals to be screened to find one lung cancer. Now standard of care in that population is a low-dose CT scan. So a positive blood test, other than potentially increasing uptake, won't impact that balance. So the balance is influenced by the negatives. False negatives lead to cancers that are missed. True negatives leads to a reduction in the number you need to screen. We'd love to have a 90-90 test. It doesn't exist. And so there's tradeoffs in development of these markers. And for this intended use, where everyone should be getting a scan, you really wouldn't want to miss a lot of cancers, so you favor high sensitivity over specificity. If instead the intended use was to expand eligibility criteria, and just as an example, say to a lower-risk population where you need to screen 500 individuals to find one lung cancer, then the opposite applies. A negative test leads to standard of care, no CT screening. So the tradeoffs come from the positives. True positives, you're finding cancers you wouldn't have. False positives, you're screening a lot of people that you wouldn't have. Now again, the 90-90 doesn't exist. So when you're looking at a tradeoff, you might say, hey, I'm willing to find 50% of those cancers if I don't need to screen more than is already considered standard of care. The bottom row represents something we can't ignore, and that's the development of multi-cancer early detection tests and what their influence on lung cancer screening may be. Because the standard of care there is no screening, developers are targeting very high specificities. You can see that that should not influence standard of care lung cancer screening, but it might add a little bit to finding cancers in those who aren't currently screen eligible. It's an exciting time to be working on biomarkers in this space. There's all kinds of academic and industry investment. This is just a sampling of some of the companies working on biomarkers, those that I'm most familiar with. And I'm just going to touch on a few of them that have come through the farthest phases of development. The first is a methylation-based biomarker with the intended use of augmenting screening in the currently screen eligible population. It's been reported in abstract form. A phase two study enrolled over 800 individuals. Characteristics of those individuals nicely match those of the NLST in the first million screens, perhaps slightly older. The cancers that were included in the case control study matched the stage distribution, cancer histology, and size of stage one cancers of the NLST. They've reported in abstract form the internal validation of the earliest development of this test. And the overall sensitivity of the test was around 90%. The stage one sensitivity of 87% at a specificity of 53 to 55%. These accuracies could be potentially clinically useful if they hold true through the final phases of development. The second is another phase two study, prospective case control study of a fragmentomics-based assay published in manuscript form with the same intended use in the currently screen eligible population to improve uptake and adherence. This study reported both a training set and a validation set. The characteristics of those included in training and validation matched each other well and matched those of currently screened individuals with the potential exception of a slightly lower percentage of stage one cancers than you'd expect in screening. The training and validation accuracies of the test when stage weighted were around 80% sensitive at 58% specific. Again, accuracies that might augment standard of care screening if they hold true through the final phases of development. I'd also like to touch on the multi-cancerly detection tests that have been published. Two have been published in phase four studies. The first is study of almost 10,000 women. The second, 6,600 individuals who were over 50 years old. You'd see a couple of characteristics of these tests, very high specificities but low sensitivities. In the context of lung cancer screening, certainly sensitivities too low to influence your decision to pursue standard of care lung cancer screening. And in terms of their specificity, a recognition that you will be evaluating some individuals with false positive results hunting for the cancer that they identified. It is possible that these tests can add to the effectiveness of lung cancer screening, but it's still not known. I think it's very important to include these as we're thinking about the role of these biomarkers because more and more of these tests are being developed. And as the technology evolves, the accuracy may as well. These are phase one studies published only in abstract form, but you can see how accurate this is a special methylation type of approach to assay development. Stage one, two cancers with an AUC of 0.94 and lung 0.96. This is a multi-omics approach to multi-cancerly detection test development. And in lung cancer, the sensitivity for stage one was 90% with a specificity of 99.5%. Now this is phase one. We don't know what those results are going to be through the final phases, but certainly promising enough for us to keep in mind for the future of lung cancer screening. So again, I just want to thank you very much for allowing me to speak to you today about this exciting area.

molecular biomarkers

lung cancer screening

early detection

clinical validation

multi-cancer tests