Characterizing the hallmarks of trials that have led to biomarker label expansions

Molecular biomarkers serve as a sorting factor by which oncology patients can be divided into sub-populations that are more likely to experience favorable clinical benefits from many types of therapies. The first generation of biomarker selection for cancer therapy included HER2 testing for guiding Herceptin prescription, and since then many new biomarkers have been studied and included on drug labels, either through new drug applications (i.e., Rubraca with BRCA mutation testing) or supplemental biologics/drug label expansions (i.e., Pembrolizumab with PD-L1 and MSI). In an effort to understand the data that must be generated in support of adding new biomarkers to drug labels, we conducted an analysis of 17 label-expanding clinical trials in order to identify what characteristics of those studies enabled supplemental FDA submissions. With these data, we have identified 600+ trials that fit the profile of label expanding studies, offering us insights on which companies and therapeutics may be best positioned for near-mid term label expansions that include biomarkers.

See the bottom of this page for an interactive data tool highlighting potentially label-expanding trials

The starting point of our analysis involved reviewing FDA approved companion diagnostic (CDx) tests and identifying which drugs and biomarkers corresponded with each diagnostic. The FDA provides a succinct overview of these approved tests, which we have also included in Table 1 below.

Table 1: FDA Approved Therapeutics & Corresponding CDx Biomarkers (2016 - 2018)
BiomarkerDrugCDx
ABL1NilotinibMolecularMD MRDx BCR-ABL Test
ALK RearrangementsAlectinibF1CDx
CrizotinibF1CDx
CeritinibF1CDx, VENTANA ALK (D5F3) CDx Assay
BRAF V600EDabrafenibF1CDx
VemurafenibF1CDx
Dabrafenib + TrametinibF1CDx, ThermoFisher Oncomine Dx
BRAF V600E + V600KTrametinibF1CDx
Cobimetinib + VemurafenibF1CDx
BRCA1/2OlaparibMyriad BRACAnalysis CDx
RucaparibF1CDx, FoundationFocus CDxBRCA Assay
EGFR MutationAfatinibF1CDx
GefitinibF1CDx, ThermoFisher Oncomine Dx
ErlotinibF1CDx, Roche cobas EGFR Mutation Test v2
OsimertinibF1CDx, Roche cobas EGFR Mutation Test v2
ERBB2TrastuzumabF1CDx
Ado-trastuzumab-emtansineF1CDx
PertuzumabF1CDx
IDH2 MutationsEnasidenibAbbott RealTime IDH2
KRAS WTCetuximabF1CDx
KRAS + NRAS WTPanitumumabF1CDx
PD-L1PembrolizumabDako PD-L1 IHC 22C3 pharmDx

While FISH, PCR, and IHC have been the key testing methodologies for earlier FDA approved CDx tests, NGS is playing an increasingly import role in the in the CDx space, driven in large part by its versatility. For example, while most PCR-based assays are only indicated for one drug, F1CDx currently covers 15+ distinct therapies. And while FDA approved tests provide a snapshot of biomarker activity, it is important to note that laboratory developed tests (LDTs) also play a large role in clinical biomarker testing, and these are not necessarily recognized under FDA filings and are therefore not included in the above table. Foundation Medicine’s FoundationOne test started as an LDT but achieved FDA approval through a PMA application. Guardant Health’s portfolio of liquid biopsy tests still carries the LDT designation, yet they will likely play a central role in generating data for future drug label expansions.

From reviewing the supplemental biologics license application / supplemental new drug application  (sBLA/sNDA) filings for each drug in Table 1, we identified the set of trials that were instrumental in generating the supporting data needed for the inclusion of a biomarker in a pharmaceutical label. This subset of trials only includes label expanding trials; the drugs must have already been brought to market with a prior NDA.

Table 2A: Collection of clinical trials (Phase 2) referenced as for their role in validating biomarker label expansions
BiomarkerDrugSubmission DateApproval DateEnrollmentStart DatePrimary CompletionNCT
BRAFDabrafenib + Trametinib9/22/20166/22/20171746/20/201110/1/201501336634
MSINivolumab2/2/20177/31/20173403/7/201412/3/201802060188
PD-L1Pembrolizumab3/22/20179/22/20173162/3/20155/27/201902335411
Pembrolizumab4/2/201410/2/201554011/20/201211/16/201501704287
Pembrolizumab12/28/20176/12/2018135012/18/20158/28/202302628067
Atezolizumab6/6/20186/19/20181197/3/201410/3/201502951767

Table 2B: Collection of clinical trials (Phase 3) referenced as for their role in validating biomarker label expansions
BiomarkerDrugSubmission DateApproval DateEnrollmentStart DatePrimary CompletionNCT
ALKCeritinib11/28/20165/26/20173757/9/20136/24/201601828099
Ceritinib + Alectinib5/31/201711/6/20173038/19/20142/9/201702075840
BRAFDabrafenib + Trametinib9/22/20166/22/20174225/4/20128/26/201301584648
Dabrafenib + Vemurafenib + Trametinib9/22/20166/22/20177046/4/20124/17/201401597908
BRCA, HER2Olaparib8/18/20171/21/20183023/27/201412/9/201602000622
EGFROsimertinib10/18/20174/18/201867412/3/20146/19/201702296125
Erlotinib11/15/20125/14/20131742/1/200712/1/200900446225
HER2Pertuzumab7/28/201712/20/2017480411/8/201112/19/201601358877
Pembrolizumab6/24/201610/24/20163058/25/20145/9/201602142738
Ph+ CML-CPNilotinib6/13/201712/22/20178527/31/20079/2/200900471497
RASPanitumumab8/29/20166/29/201711838/1/20068/1/200900364013

From the data in Table 2 and analysis of the clinical trial listings, several features we identified as characteristic of registrational trials include:

  • Mid-late stage – All of the analyzed trials were phase 2 or higher, and the majority (11 of the 17 trials included) were phase 3
  • High enrollment volumes – From the above trials, average enrollment was ~750; for comparison, average enrollment across immuno-oncology biomarker trials from our I/O BioMAP tool was ~290 in phase 2+ trials
  • Biomarkers for inclusion – 11 of the 17 trials included the biomarker as an inclusion criterion; 2 were for cohort stratification; 4 did not specify what the biomarker was used for in the trial description
  • Trials are well underway – 4.3 (~1,550 days) was the average time after the trial start date that supplemental drug label expansion requests are filed; the maximum was 3,690 days, and the minimum was 498
  • Primary completion reached – 13 of the 17 trials had achieved their primary completion dates prior to filing for a label expansion
  • Markers not mentioned in outcome measures – The biomarkers included in the label expansions were only mentioned in the outcome measure of 3 trials
  • Pharma primary sponsorship: Every trial had the pharma company as the primary sponsor

Discussion

From the characteristics of the registrational trials presented in table 2, this allows us to highlight what may be considered a ‘prototypical’ registrational biomarker study. While it may not be a surprise that label expanding trials would be phase 2 or higher with large enrollment volumes and sponsored by the relevant pharma company, several of the features we identified above lead to a trial profile that somewhat defies expectations. First, it is surprising that more trials do not list the biomarker as an outcome measure. Given that these studies are exploring outcomes based on prospective biomarker analysis, it would make sense to correlate primary outcomes with the biomarkers of interest. On the other hand, these registrational trials are often established after smaller / earlier stage trials that include biomarkers, which lay the groundwork for prospective inclusion in the enrollment criteria of these larger studies rather than the outcomes fields.

Another intriguing finding that we did not predict was the use of Objective Response Rate (ORR) as one of the key outcome measures on drug labels used to measure overall efficacy. With an average period of 4 years to accumulate study data, we would expect more late-stage outcomes, such as progression-free survival (PFS) or overall survival (OS), to be included on the drug label. However, we found that 15 of the 17 trials included ORR, 9 included PFS, and only 3 included OS in drug label registrational data.

One publication from 2016 addresses exactly this point. The author claims that in recent years ORR has been increasingly seen in accelerated label approvals compared to OS and PFS. A study from JAMA quantified this uptick in single-agent ORR approval rates, finding that ORR’s exceeding 30% received an 89% approval rate, while an ORR of 45%+ were approved 100% of the time. Looking back at the data we collected, ORR is the sole endpoint included for FDA approvals in the two most recent label expansions of 2018, and the other two 2018 approvals included PFS and / or DOR in addition to ORR.

Understanding the characteristics of trials that have led to biomarker-based label expansions allows us to develop a set of criteria that can be used to identify the next generation of registrational studies. The use of ORR as a key endpoint already marks a significant step in enabling more rapid label expansions, and when this is coupled with next generation genomic technologies and markers, it is possible that clear clinical benefits can be predicted and used for registering label expansions after even short times.

Predicting Future Registrational Trials

Based on the characteristics of the registrational clinical trials above, we analyzed the 60,000+ oncology trials on clinicaltrials.gov and identified ~600 trials that have similar profiles as those presented in Table 2. These ~600 trials represent the subset of studies that match the ‘prototypical’ profile of registrational biomarker studies. From this set of trials, we have created an interactive data tool that allows us to further refine search criteria (e.g. keywords, inclusion criteria, outcome measures, indications) to hone in on key trials that may well be pivotal to near-midterm label expansions and shed light on what late-stage clinical programs different pharma companies are prioritizing. It is important to note, however, that we did not pre-filter by biomarkers during this analysis, but in the interactive tool below we offer search functions that allow us to identify mentions of biomarkers in either the inclusion criteria or outcome measures.

To validate the accuracy of the criteria discussed above and the subset of trials it identified, we checked to see if our data could accurately identify several of the top priority trials that have been included in recent sBLA filings. Some of these include:

  • CheckMate-227 – Nivolumab + Ipilumab in TMB-high patients
  • IMpower150 – Atezolizumab explored in EGFR or ALK mutation positive patients, with possible inclusion of Teff gene signature

Even though the trials included in the interactive data tool below may fit the profile in timeline, enrollment, phase etc. to be registrational, their future clinical outcomes (especially ORR) is the feature that will ultimately determine whether FDA filings can commence or if follow-on, more focused clinical studies need to be established. And while we can certainly identify several trials on this list that include established biomarkers in their inclusion criteria (e.g.,  EGFR, PD-L1), we are keen to identify next generation markers in both the I/O and targeted therapy spaces that may still only be included in an exploratory capacity.

For deeper insights on immuno-oncology biomarkers, check out our I/O BioMAP, which has data on over 1,300 I/O clinical trials including trends in biomarker usage, sponsorship data, and a timeline of when to expect catalyzing trial results

Several notes on navigating this data tool:

  • You can click on any of the chart elements to apply filters and narrow down the list of relevant trials at the bottom of the page
  • Clicking on a specific trial will bring up the option to navigate to the corresponding page of clinicaltrials.gov
  • If you are interested in searching for trials containing a particular marker, you can use the free-text search functions on the right side of the screen; as an example, if you search for “V600” in the Inclusion Criteria search box, this will return a subset of ~20 clinical trials that can then be filtered down further; while conducting searches, be aware that the free text searches capture all mentions of a particular query and not just exact (e.g., a search for “ALK” would generate positive hits from “walk” or “talk”)