The field of personalized medicine is a new and evolving field with new combinations of products and new terms to understand. A lot of work is being done on what are called companion diagnostics (CDx) in the IVD world. This blog explores what companion diagnostics are and how they differ from other diagnostics including some points to consider if tasked with developing one.
In the US and in the European Union, CDx development is approached in the same manner. However, it is important to note that the definition of what is considered to be a companion diagnostic is different in the US and EU.
FDA defines a companion diagnostic (CDx) as follows:
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. The test helps a health care professional determine whether a particular therapeutic product’s benefits to patients will outweigh any potential serious side effects or risks.
The IVD Regulation 2017/746/EU, Article 2(7) defines a CDx as follows:
Companion diagnostic' means a device which is essential for the safe and effective use of a corresponding medicinal product to:
a) identify, before and/or during treatment, patients who are most likely to benefit from the corresponding medicinal product; or
b) identify, before and/or during treatment, patients likely to be at increased risk of serious adverse reactions as a result of treatment with the corresponding medicinal product
FDA further clarifies by pointing out that a companion diagnostics can:
- identify patients who are most likely to benefit from a particular therapeutic product;
- identify patients likely to be at increased risk for serious side effects as a result of treatment with a particular therapeutic product; or
- monitor response to treatment with a particular therapeutic product for the purpose of adjusting treatment to achieve improved safety or effectiveness.
If the diagnostic test is inaccurate, then the treatment decision based on that test may not be optimal.
CDx assays are typically described in terms of:
- the clinical condition of the subject (e.g. breast cancer)
- the drug to treat the clinical condition (e.g. Herceptin or trastuzumab)
- the biomarker that is evaluated in the assay (e.g. ERBB2 or HER2)
- details on the biomarker (e.g. HER2 protein overexpression)
Excerpt from the FDA website on “Cleared or Approved Companion Diagnostics.”
What is the mystery around CDx IVDs? Are they developed any differently than the run of the mill, lowly every day IVD (Dx)? The simple answer is no and yes. Okay, the real answer is that it depends on how you look at things.
The mechanics of development of a CDx assay versus a “regular IVD” are the same. During development we need to make sure that the “analytical studies”: limit of detection (LOD), precision, interfering substances, and stability (just to name a few) are met for the assay no matter if it’s Dx or CDx. The reason we evaluate these metrics is so that when it comes time to make sure we are calling an apple an apple (or in our example here, an ERBB2 amplification) is that these performance metrics are tied to “accuracy” or what I call “allowable misclassification.” In other words, how often must we call a “positive” (that is a subject that has the clinical condition) correct and how often must we call a “negative” (that is subject that does not have the clinical condition) correct? Which is more important given the intended use of the product; correctly identifying positives or negatives?
For most Dx assays this means comparing to another method as the association of the analyte with the clinical condition has been established. If one company’s assay (ACME Diagnostics) is comparing to another company’s assay (Wild E Coyote Diagnostics) you are doing a method comparison and would report positive and negative percent agreement. You want to know that your apple is the same as someone else’s apple.
But what about CDx? Is accuracy about identifying those subjects with the biomarker or would it be about identifying whether the population you’ve identified will or won’t respond to the drug in question? Ultimately, it’s about the latter, finding the subjects that will respond to the drug versus the ones that won’t respond to the drug as these are the clinical conditions. Depending on how the analyte works with the mechanism of action of the drug these could be positive or negative (i.e. if the subject has the mutation, drug will work, or if the subject doesn’t have the mutation, the drug will work).
This is the biggest difference between the CDx and Dx. A Dx can be more often compared to another Dx method since the association of that analyte has been established with a clinical condition. With a CDx, we don’t have the same level of information on the association of the analyte with the clinical condition (respond to the drug) as we often don’t have much if any information on how the analyte is related to the drug’s mechanism of action.
We can make the best assay to detect ERBB2 amplification as compared to any other method, but we can’t do anything in the development of the CDx to ensure that the ERBB2 biomarker detected will be indicative of the population that will or will not respond to the drug. This is something that we rely on the pharma partner in terms of due diligence. How well the analyte is associated with the mechanism of action of the drug has a bigger impact than anything else in terms of the success of a CDx. Asking questions about this and understanding this prior to starting any development effort should be considered. This could be considered a “constraint requirement” (you can’t design the assay without this) when thinking about system requirements. As such having sufficient documentation on how this analyte was chosen becomes very important.
One way to cover our assets, would be to compare the CDx assay with another Dx assay that is on the market with acceptable detection of the assay analyte. This could allow us to have some confidence that we are adequately detecting the analyte going into the drug study.
Performance of the drug is done in a pivotal trial. Depending on how this is done there may or may not be sufficient data to fully evaluate the CDx. If those subjects chosen by the CDx do not demonstrate that the drug is effective, we can at least have some notion that it was not due to the fact that the CDx assay did or didn’t detect the biomarker satisfactorily, but maybe the biomarker’s association with the drug mechanism of action was not adequate to determine the population that would or would not respond as required.
In summary, the development of CDx assays is no different than the development of other IVD assays. What is different in how the analyte that has been chosen relates to the clinical condition of the subjects, (response or no response to the drug). Does the developer of the CDx have sufficient information on the association of the analyte such that it will be able to distinguish between these two clinical conditions?
There is no process in the development of the assay that can make it work adequately if the correct analyte has not been chosen to begin with. If you are tasked to create such an assay, then the use of this analyte in the design of the assay becomes a “constraint requirement” at the system level. Documenting how this analyte was chosen by the pharma partner becomes very important and all data or other information should be considered prior to moving forward with any other development work. If there is a way to make sure your apple matches someone else’s apple during development, this could be useful in making sure you are on the right track.
More information on companion diagnostics