So you believe in personalize medicine. You've spent years identifying a kinase that has two distinct gain of function mutations that drive malignant transformation in a subset of several common tumors. You've developed an antibody that detects the mutant proteins, and a small molecule that inhibits the altered enzyme. Your nonclinical program is successfully completed and you're ready for the clinic.

And the regulators write back "Where's your Companion Diagnostic?"
My what?

The rapid advances in molecular medicine have provided powerful tools for identifying the biologic processes behind clinical diseases and exploiting the insights to develop new therapeutics. But, "with great power comes great responsibility." Highly targeted therapies require highly accurate means to identify those patients who will, and will not, benefit.

Hippocrates had no lab tests at all. His only diagnostic tool was to talk with and examine his patients. By the 19th century there were microscopes, tissue fixation, histologic stains, and the emergence of histopathologic diagnosis. By the mid-20th century, there were "routine" laboratory tests for hematology, chemistry, urine analysis, and microbiology, which were performed manually using variations of published "protocols".

From 1960's the emphasis was on automation – chemistries were performed by multichannel, continuous flow analyzers; hematology results came from automated cell counters; antibiotic susceptibilities from microculture plate systems. Initially, oversight of these instruments was the domain of the organizations that accredited the laboratories, the hospitals, and their staff.

A diagnosis was the product of integrating the clinical data acquired from the patient by physicians and nurses, imaging studies performed by a range of radiologic technics, and laboratory data obtained by analysis of tissue, blood, urine, etc. Treatment depended on the diagnosis, but not on the result of any one test.

Molecular tests have fundamentally changed the 'diagnosis – treatment' paradigm. Treatment is no longer directed at a broad clinical diagnosis. Treatments are developed for specific molecular targets and are only effective against the specific target (or variant). The archetype is treatment directed at the HER2 oncogene found first in breast cancer and later in gastric cancer. Indeed, until very recently, the majority of approved Companion Diagnostics were directed at HER2.

Conceptually, a "diagnostic" test can be applied at several different points in a clinical trial of an investigational therapy:

  1. Post-hoc analysis: Patients are enrolled and treated based on traditional clinical diagnosis, independent of the test. Samples from before, during, and/or after treatment are collected, tested, and the results analyzed with regard to outcomes – prognosis, clinical efficacy, adverse events, etc.
  2. Stratification: Patients are enrolled based on clinical diagnosis, and the results of specific laboratory tests are used to assign individuals to different strata prior to randomization. Note that no one is included or excluded based solely on the test result. Rather, the goal is simply used to assure that the distribution of positive and negative results at baseline are balanced across the treatments being studied.
  3. Selection: Patients with a particular clinical diagnosis are tested as part of screening and only patients who are test "positive" are enrolled. Test "negative" patients are excluded from the investigational trial.

Only in scenario #3 is the test considered a Companion Diagnostic.

Why do agencies that regulate drugs care about a diagnostic test?
If a treatment is only effective for and applied to patients with a specific target, then the accuracy of detecting that target will directly impact the efficacy achieved in clinical practice. Considered from a different perspective, the risk-benefit of the treatment depends not only on the clinician making the correct clinical diagnosis, but also – potentially more – on the performance of a specific laboratory test.

Therapeutic efficacy is now linked to laboratory accuracy and assessing efficacy is central to regulating treatments.
The issue becomes particularly acute if the target of the investigational therapy has a relatively low prevalence among patients with the same clinical diagnosis.

Regulatory agencies have long been involved in assessing therapeutic 'devices'. The archetype is therapeutic radiation. (In the U.S., the division responsible for companion diagnostics is called the Center for Devices and Radiological Health or CHDR ). Laboratory tests are classified as "devices" because the result is generated by a piece of equipment. The "device" includes both the "platform" (the particular technology) and the specific test. As medical care decisions are increasingly impacted by the results of laboratory tests, the agency has become more involved in their regulation. Specific examples include microbial susceptibility testing to determine the choice of antibiotics and consumer tests for pregnancy and blood sugar.
In the current context, all new diagnostic tests start as research tools – by definition, these are exploratory. In academic centers, clinical labs may quickly adopt these as "Lab Developed Tests" (LDTs). In a CLIA-certified lab, the Lab Director is expected to document at least a modest level of reproducibility, and the tests are then often referred to as "CLIA-certified Tests". In the FDA's nomenclature, however, such tests are "Research Use Only" (RUO), and the results must be reported as "For Research Use Only. Not for use in diagnostic procedures."

Tests that are being formally evaluated per regulatory procedures may advance to the level of "Investigational Use Only" and results labeled "For Investigational Use Only. The performance characteristics of this product have not been established." (Some observers have quietly wondered if it might not be more helpful to adopt labels analogous to Phase 1, Phase 2, Phase 3.) Per FDA Guidance (25 Nov 2013), until a test has received regulatory approval it may neither be represented as "effective in vitro diagnostic products" nor be "used in clinical diagnosis, including use in making medical treatment decisions."

Why has this all become so important in the context of personalized medicine?
The typical hypothesis driving development is that the treatment will be effective if and only if the specific target is present. The immediate corollary is that the appropriate population for evaluating therapeutic efficacy is patients who test positive for that target. In the US, there are regulatory requirements for using the results of an unapproved laboratory test as an inclusion / exclusion criterion in an clinical trial of an investigational drug. Early in the development program the Sponsor may apply for a "No Risk Assessment."

Note this is not an assessment of the safety of the treatment. This is an assessment of the diagnostic test, being performed by CDHR, a Center for devices. The risk of concern is that that test result may be wrong. If it's false positive, someone will get treated who shouldn't. If it's false negative, someone won't get treatment who should. Those are reasonable concerns if the hypothesis is that only test-positive patients will benefit from the treatment.

Key factors that favor a successful "No Risk Assessment"

  • The patients to be enrolled have no proven treatment options(e.g. relapsed / refractory cancer patients).
  • The test is plausibly reliable(e.g., detects a well-defined mutation or chromosomal alteration). Assaying for a novel antigen or antibody, or for an altered level of gene expression may be considered problematic.
  • The lab(s) that will be performing the test are prespecified and are a credible component of a clear plan for the development of a validated IUO assay that will be available for use in the Phase 2/3 clinical trials.

Even with this approval in hand numerous challenges remain. Enrollment is restricted to patients who test positive in the designated lab. Even if the Pi discovered the target, you still can't use the result from her lab. If the test requires a tissue sample, then sample collection, handling, transport, and processing must be standardized. That process may take time, may delay starting treatment, and thus may frustrate investigators and patients. In addition, the enrollment rate will be limited by the frequency of patients with positive tests among those with the relevant clinical diagnosis. And impacted further if test-positive patients have poorer prognosis or increased co-morbidities.
Relatively few companies have the expertise or capacity to meet the regulatory requirements for developing a Companion Diagnostic. The effort and the cost of contracting the program is often an unexpected high. Even the process of choosing an appropriate "partner" can be daunting, since if things go well (as hoped), the two companies will be working together through and beyond approval of both the treatment and diagnostic.

In summary, if your treatment is truly "personalized" and target-restricted, then the development plan likely needs to incorporate a Companion Diagnostic. And the process is best initiated as early as is feasible.