'Like a Coin Flip': Assay Denies Some Cancer Patients New Drug

Nick Mulcahy

March 17, 2020

In December, at a major breast cancer conference, some attendees couldn't find a seat and were told to leave an overcrowded session on immunotherapy for metastatic triple-negative breast cancer (TNBC).

They refused, and pushed in to hear what was being said.

The crowd might have been surprised to learn that the main draw of the event, a successful new drug, was not all it might have been for women with the disease, being handicapped by a test that determines who is eligible for it.

"That room was overpacked ― there were five people deep against the wall.... It was amazing," said Janice Cowden of Bradenton, Florida. She attended the meeting, the San Antonio Breast Cancer Symposium, as a patient advocate.

Cowden lives with metastatic TNBC, which is known for poor prognoses, aggressiveness, and a lack of targeted treatment options.

"Stage IV is a state of desperation. We just want something to work," she said.

That's why the conference room was packed ― the session was focused on something that had been found to work ― the immunotherapy atezolizumab (Tecentriq, Genentech/Roche).

Atezolizumab had recently been conditionally approved for first-line use in advanced TNBC, having been shown to significantly slow disease progression and, in some patients, to possibly improve survival. A pair of medical oncologists reviewed the clinical trial data during the session.

One important point from the trial data was that the benefit was greater in patients whose tumors had the biomarker PD-L1, and so the FDA approval of the drug specified that it should be used only in those patients.

The drug approval was accompanied by approval of a companion diagnostic test used to identify this PD-L1-positive subgroup of patients, the Ventana SP142 Assay (Roche Diagnostics).

Dr David Rimm

At the meeting, pathologist David Rimm, MD, of Yale University, New Haven, Connecticut, discussed the biomarker PD-L1 and the test.

Rimm had a subtle but unsettling message about the crucial test: that the SP142 diagnostic assay, when used by increasing numbers of pathologists, resulted in increasing rates of PD-L1 scores that were not concordant.

A related meeting poster, presented the next day with Rimm as senior author, was more explicit and concluded that "more than half of the pathologists in real-world situations may mis-assign" patient scores with SP142 (and another Roche assay) because of wide variability in readouts.

"They've made a test that is inadequate ― it just doesn't work. It's like flipping a coin," he told Medscape Medical News about Roche's SP142 assay in everyday practice.

They've made a test that is inadequate ― it just doesn't work. Dr David Rimm

The general problem is not a new one ― for some years there have been problems with the use of PD-L1 as a biomarker for immunotherapy and with assays for that biomarker, with many groups questioning both accuracy and reproducibility.

But the problems with SP142 are "the most egregious," said Rimm, who has served as a paid consultant to Roche Diagnostics in the past.

In clinical practice, Rimm's overall message is that because of the difficulty of reading SP142 assay results, some TNBC patients who were PD-L1-positive would not get the drug, and some who were not positive would get the drug.

Patient advocate Cowden was not worried about overtreatment. She was concerned about patients who "might die without receiving a potentially life-extending treatment."

In an essay in the Pathologist, Rimm echoed that sentiment about undertreatment (as well as overtreatment) with atezolizumab for breast cancer: "In all cases, the patients are the potential victims, but this appears to be completely under the radar of the hype surrounding this new drug."

Roche Disputes Problems With Assay

Roche, manufacturer of both atezolizumab and the companion diagnostic test, disputes that there is a problem.

The FDA and multiple health authorities worldwide have approved atezolizumab and the companion diagnostic SP142 assay for use in TNBC, points out Eslie Dennis, MD, vice president of medical affairs at Roche Tissue Diagnostics.

"The role of a companion diagnostic assay is to discriminate between responders and non-responders for a specific therapeutic product in a specific indication, with a cut-off based on clinical outcomes," she wrote in an email to Medscape Medical News.

Data from the pivotal IMpassion130 trial show that the assay was effective at that task.

Among the 369 patients in the 902-patient trial whose tumors were ≥1% positive for PD-L1, those treated with atezolizumab (and nab-paclitaxel; n = 185) had a median progression-free survival (PFS) of 7.4 months, vs 4.8 months among those treated with placebo (and nab-paclitaxel; n = 184) (P < .0001).

"Exploratory analysis showed no [PFS] benefit in PD-L1-negative patients as tested by the SP142 assay [in IMpassion130]," Dennis and three other physicians write in a reply to Rimm in a letter published in July 2019 in the Pathologist.

The same held true for overall survival in exploratory analysis ― there was no benefit with atezolizumab among the PD-L1-negative patients, they write.

Notably, overall survival benefit for patients who were PD-L1 positive was about 10 months (at the first interim analysis; at the second analysis, the benefit dropped to 7 months and was not statistically significant).

But Rimm points out that the pivotal trial used only one pathologist in a central lab to determine PD-L1 status, who was undoubtedly an expert with the SP142 assay.

Further, Rimm observes that additional data submitted to the FDA to show that SP142 test results are reproducible outside of the pivotal trial setting were performed with only three pathologists and thus unsurprisingly yielded high rates of agreement ― all above 90%.

The data from both of these circumstances are problematic, Rimm said, because in the real world, hundreds of pathologists will score the SP142 assay ― all in the context of a busy day reading a variety of other tests for other diseases.

It's one thing to get an FDA approval for an assay, and it's another thing to be a reliable, well-functioning assay in the real world, he summarized.

Last year, Roberto Salgado, MD, PhD, a pathologist at the Université Libre de Bruxelles, Belgium, commented that "a positive phase III trial should not be taken as a guarantee that the assay used in the trial can be implemented in daily practice" in an opinion piece in the Pathologist.

SP142 Identifies the "Fewest Possible Patients"

The SP142 assay has been shown in multiple studies to have lower sensitivity for PD-L1 than other competing PD-L1 assays, said Rimm, citing examples such as a 2017 study and a 2018 study.

Dr Angela DeMichele

Angela DeMichele, MD, a medical oncologist at the University of Pennsylvania in Philadelphia, agreed and explained what that meant in practical terms for women whose tumors are tested with SP142. "It means that the test is going to identify the fewest possible PD-L1-positive patients [relative to the other available assays]," she said.

"It [the SP142 assay] is far from a perfect test for this situation," added DeMichele, an expert on biomarkers in breast cancer clinical trials.

She said that biomarker tests, like many products of science, tend to become dated with the passage of time, as more is learned about the target and new assays are developed. "Unfortunately, you can't change assays midstream," said DeMichele. She has received a grant from Roche and Stand Up to Cancer to study atezolizumab and another drug in a clinical trial among patients with metastatic TNBC who have minimal residual disease.

DeMichele also said that "David Rimm is one of the most knowledgeable people in the world about this issue."

But DeMichelle also points out the practical: "We're stuck as clinicians" because regulatory bodies and insurance companies only pay for atezolizumab when the SP142 assay indicates PD-L1 positivity.

That's not the case in Europe, where health authorities do not specify which PD-L1 assay is to be used with atezolizumab for breast cancer, pointed out Belgium's Salgado last year.

Another Level of Complexity

At the immunotherapy session in San Antonio, Rimm discussed the results of a study of 68 TNBC archived cases in which specimens were stained with the SP142 assay at Yale and were distributed via electronic images to 19 pathologists at 14 institutions across the United States for PD-L1 scoring.

The study, coauthored by academics from Iowa, Texas A&M, UC San Diego, Mayo Clinic, Memorial Sloan Kettering, and others, used a novel method to determine the minimum number of evaluators needed to estimate "concordance" or agreement about a test result among large numbers of readers.

The consensus/agreement was as high as 80% when eight or fewer pathologists' scores were compared, but was as low as 40% when results from more than eight pathologists were included, said Rimm.

These are some of the data that led him to declare that using the assay is no better than flipping a coin.

Yes, PD-L1 testing is a challenge, and it has "introduced another level of complexity" for pathologists in reading assays, write experts Emina Torlakovic, MD, University of Saskatchewan, Canada, and Allen Gown, MD, PhenoPath Laboratories, Seattle, in response to Rimm last year.

But there is "poor" consensus among pathologists, they point out, "for many scoring systems that are still clinically applied (such as Gleason grading)." Consensus "improves with education and training," the pair add.

To that end, Roche has initiated a global training program for pathologists using the SP142 assay for TNBC. At San Antonio, Roche's Dennis reported that among 432 pathologists from 58 countries, there was overall agreement of 98.2% in scoring assays.

Rimm commented that such high agreement would not be a surprise if testing took place soon after any such training program.

In an email to Medscape Medical News, Torlakovic encouraged pathologists who wish to practice their skill in interpreting assays, including SP142, to visit CBQAReadout.ca, a testing site. The site, which was founded by Torlakovic and may be one of a kind, offers CME credits and is sponsored by independent pathology organizations, such as CAP-ACP and the Saskatchewan Health Authority, as well as pharmaceutical companies, including Roche.

No Clue

Patient advocate Cowden believes the controversy about PD-L1 testing for atezolizumab is largely unknown among breast cancer patients.

She learned about SP142 assay ambiguities in San Antonio, when the Florida Breast Cancer Foundation funded her trip to the meeting and the Alamo Breast Cancer Foundation asked her to write a report on Rimm's presentation.

Cowden is a member of a Facebook group for stage IV TNBC, which has about 1500 members. She estimates that 75% to 80% would be willing to try atezolizumab "no matter what," meaning they don't care about PD-L1 positivity being associated with efficacy.

The Facebook group members "know there is a test and if you are positive, there is an immunotherapy for their breast cancer," said Cowden.

None know that women may be excluded from treatment because of shortcomings with the SP142 test. "They have no clue," she said.

Rimm and DeMichele have financial ties to Roche and other companies. Dennis is an employee of Roche. Torlakovic has ties to multiple companies, including Roche, for whom she has acted as a paid consultant, grant recipient, and paid lecturer. Gown did not respond to a request for financial disclosures. Cowden reports no relevant financial relationships.

Follow Medscape's Nick Mulcahy on Twitter. For more from Medscape Oncology, follow us on Twitter.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.
Post as: