It is very likely that on your last flight the turbofan engines were controlled by full authority digital engine controls – FADECs for short. FADECs have played a significant role in keeping airline ticket prices low (except during holidays) by continually adjusting engine parameters so that the engine operates with maximum fuel efficiency and within operational limits, allowing pilots to focus on other tasks. As the FAA notes (not the FDA for once), “Occasionally, pilots have run engines beyond operational limits in order to get out of tight situations. That can’t happen with FADEC.” Also, if the FADEC fails, the engine fails, and, “The hardest fact for most to get accustomed to is that the system provides no reversion to manual control.”
As the FDA notes, “A companion diagnostic is a medical device, often an in vitro device, which provides information that is essential for the safe and effective use of a corresponding drug or biological product.” Companion diagnostics are commonly qualitative yes/no tests that usually indicate whether or not a patient would have been included in the phase III trial that led to the approval of the companion diagnostic. Since the safety and efficacy of the corresponding therapeutic product was only established on those patients who were selected with the companion diagnostic, the safety and efficacy on other patients may be unknown. In the practice of medicine, the physician is often given authority to treat patients in the most appropriate manner, but insurers have authority on whether they cover the treatment. When an indication for a therapeutic product calls for a particular result by a companion diagnostic, insurers sometimes treat the companion diagnostic as having full authority on whether the treatment with the drug can be authorized. In such a case, one might refer to it as a full authority companion diagnostic (FACD).
The problem with this is that many biomarkers are not present/absent in a patient, but exist on a continuum. Additionally, responses to a therapeutic product may be on a continuum as well and not just dependent on a single biomarker. For a phase III clinical trial, it is likely that biomarker cutoffs were chosen that took the likelihood of having a successful clinical trial into account. Thus, a FACD may not be possible when biomarkers are on a continuum and there is no clear a priori yes/no answer as to whether a therapeutic product will be effective. In fact, biomarkers such as PD-L1 are on a continuum in terms of expression level and the number of tumor cells expressing them. In immuno-oncology, it is known that PD-1/PD-L1 inhibitors are often more likely to be effective in the treatment of patients with PD-L1-expressing tumors, which is why some indications call for the assessment of PD-L1 expression on tumor cells by immunohistochemistry. However, some patients with little to no PD-L1 expression on tumor cells have also appeared to benefit, which could present a problem if this imposes limits on treatment options.
One possibility is to omit any requirement for a particular companion diagnostic result from the indications – if the FDA (or applicable regulatory body) allows for it, or even asks for it. Presently, this appears to be the case for IMFINZI (durvalumab) where in a clinical trial 26% of PD-L1 high patients with urothelial carcinoma had a response and 4% of PD-L1 low/negative also had a response. In the words of the FDA reviewers, “Use of the PD-L1 (SP263) IHC assay, as proposed for this BLA, may identify patients who may have a modestly higher chance to respond to durvalumab. Nevertheless, the observed durable responses in patients with low or no PD-L1 expression (as defined by <25% staining of both tumor and immune cells) in their tumor specimens agues against mandatory use of the PD-L1 assay for selection of patients. Given the limited treatment options available for such patients and based on the reviewers’ best clinical judgments, it is important to make durvalumab clinically available to all patients who may benefit from this second-line treatment.” At the same time, Aetna would still like patients to get tested by the PD-L1 (SP263) assay and IMFINZI – like many other things – does require preauthorization. This raises the question of whether the FDA-approved Ventana PD-L1 (SP263) Assay – which is technically not a companion diagnostic – is considered by insurers to be a FACD.
Another issue with biomarkers on a continuum is that the mechanism of action of a therapeutic product may not be understood sufficiently well. For example, the immune system is complicated and PD-1/PD-L1 interactions are just one of many that influence T cell responses. About two decades ago, I first observed PD-1, LAG-3, CD6, and other markers appearing (or disappearing) on naïve mouse gamma delta T cells that were responding to antigen recognition with the goal of then identifying gamma delta T cells that had recognized antigens in order to determine what those antigens are. Upon identifying the markers, my thesis committee then tasked me with trying to figure out how all those changes in gene expression affect immune responses; that took a while and could only be addressed partially. At last year’s AACR meeting, there was a talk on human T cells that concluded with where I was about 20 years ago – albeit in studies of human immunology and with the more common alpha beta T cells. It’s a very hard question.
Another example is the flu shot. Unlike some cancers, the flu is quite common and, every fall, there is this question of which vaccine to get (there are quite a few). At my wife’s medical practice – just like pilots who are not supposed to eat the same meal – half the staff gets one vaccine and the other half gets another. Sometimes your insurer or employer has a say in this; but you can always pay out of pocket if you disagree. There is also the question of when to get vaccinated; early, or just before the flu becomes widespread. In that regard, a look at the CDC’s FluView shows that the dominant influenza A strain this winter has been (H1N1)pdm09 – formerly referred to as the swine flu. During the 2009-2010 flu season, you either got vaccinated against this, came down with the flu, or both. It took a while (about a decade), but it’s back. This could indicate that protection lasts for many years until the flu mutates sufficiently, which could also indicate that getting vaccinated early is effective. On the other hand, it might not. However, even with many people getting vaccinated and many people getting the flu, many things are still not understood sufficiently well. There is also a question of whether knowing the mechanism of action always matters. One of my immunology professors pointed out that when treating patients, you want to make them better and sometimes you just want to find something that works – even if it is unclear why it works.
The FDA recognizes the importance of mechanism (or not knowing the mechanism fully) in its recent draft guidance related to expanding the indications of companion diagnostics to a group or class of oncology therapeutic products. “FDA recognizes that as science evolves, our understanding of the mechanism of action of therapeutic products and of the interaction between therapeutic products and biomarkers will evolve, which may impact how specific groups or classes of oncology therapeutic products are defined … Having a detailed understanding of the mechanism of action for the therapeutic is critical to support broader labeling identifying the specific class of therapeutics for which a companion diagnostic could be safely and effectively used.” In aviation, there is an excellent understanding of how planes work and how to engineer them to work, and it is possible to take critical aspects of keeping an airplane flying completely out of the hands of a pilot. At the same time, should diagnostics that measure biomarkers on a continuum, or where the link to the mechanism of action of a therapeutic product is not well understood, be given full authority over whether that therapeutic product can be used?
Finally, with all diagnostics, accuracy and precision are critically important. If a small difference in a biomarker is going to determine whether someone may or may not be eligible to be treated with a therapeutic product, then the performance of a diagnostic that measures that biomarker needs to be appropriate. This is an important reason for why we continue to develop reference standards that allow for the design and evaluation of diagnostic devices as new biomarkers emerge.
If you’re looking for guidance on validating your NGS-based companion diagnostic assay, check out our free eBook. It’s full of key considerations and guidelines to help you accelerate NGS assay validation for less money and greater confidence in results.