Welcome to Our Kitchen
Turning pathologists into chefs
David L. Rimm | | Opinion
In this food-crazy world, every chef has a special recipe. Unfortunately, that approach has now been forced onto pathologists, who would understandably rather consider themselves scientists than chefs. The difference? A chef’s secret recipes are highly successful, but hard to reproduce. Pathologists aspire to high-level assay reproducibility. We often create our own tests (known as lab-derived tests, or LDTs), rather than use a kit, so that we know exactly what is in each component of the assay. It’s the protocol – not the recipe – that leads to a high level of reproducibility.
With the FDA’s approval of atezolizumab with the SP142 companion diagnostic test, Roche Diagnostics has relegated pathologists to the role of short-order cook. Although it is true that the IMpassion 130 trial shows that atezolizumab improves survival in patients with high PD-L1 (1), their recipe for detection of high PD-L1 is no less secret than a master chef’s prize dish. The SP142 assay has been shown by two independent prospective multi-institutional studies not only to have lower sensitivity for detection of PD-L1 protein by immunohistochemistry, but also – when using the prescribed measurement criteria – to be non-reproducible (2,3).
Why would the FDA approve such an assay? Clearly, the drug is the dog and the assay is the tail. No doubt there is great political pressure to make available a drug for triple-negative breast cancer that increases median survival by 10 months in the PD-L1-positive group. So pathologists just need to buy the kit and do the assay – right? Unfortunately, data in the literature suggests that it will be hard to accurately reproduce the assay approved by the FDA.
First, pathologists are asked to score “immune cells” – a task shown to be reproducible between the two or three company pathologists in the SP142 summary of safety and effectiveness data from the FDA, but not between the 13 or 25 pathologists participating in statistically powered, prospective studies done in the real world. What will happen when thousands of pathologists around the world are expected to read this assay?
Second, the assay is less sensitive than others that detect PD-L1. Because the recipe is secret, we can’t be sure why, but both pathologist-read and objective cell line measurement studies have shown the assay to be negative in cases or cell lines that are positive by other assays (4). Whether the pathologist uses the FDA-approved test or an LDT, there is no clear way to standardize and quality control the assay. With the other assays, validation can be done by comparing them to one another. This is especially important for companion diagnostic tests. We know that the SP142 assay is less sensitive, so there is no standard for comparison or validation.
In Roche’s defense, they did not know when they began the IMpassion 130 trial that their assay would be less sensitive or that their reading system would show poor reproducibility. Furthermore, with over 900 patients, a 10-month improvement in overall survival is hard to ignore. Nevertheless, the whole trial hinges on the reads of one pathologist in a central lab placing 369 patients into a negative or positive PD-L1 group – hardly a foolproof process.
What is the way forward? One possibility would be for Roche to test an RT-PCR assay that they have, in their Poplar trial, shown to be predictive in lung cancer (5). RT-PCR of three specific mRNAs could be much more objective and probably more reproducible. Another approach would be to use cell lines with known PD-L1 protein concentrations to standardize their IHC assay against other assays, or to test the IMpassion 130 tissues with multiple IHC assays, including their own SP263 assay, to allow harmonization with existing assays.
Because other PD-L1 assays are more sensitive than the SP142 test, we will never know how many patients will receive treatment based on those assays and show no benefit. Nor will we know how many patients who won’t receive treatment (due to the challenge of accurately reading the assay) who might have benefited. 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. It is my hope that Roche/Genentech/Ventana will work with pathologists to find a solution by bringing some science to the table.
- P Schmid et al., “Atezolizumab and Nab-Paclitaxel in Advanced Triple-Negative Breast Cancer”, N Engl J Med, 379, 2108-2121 (2018). PMID: 30345906.
- DL Rimm et al., “A Prospective, Multi-institutional, Pathologist-Based Assessment of 4 Immunohistochemistry Assays for PD-L1 Expression in Non-Small Cell Lung Cancer”, JAMA Oncol, 3, 1051-1058 (2017). PMID: 28278348.
- MS Tsao et al., “PD-L1 Immunohistochemistry Comparability Study in Real-Life Clinical Samples: Results of Blueprint Phase 2 Project”, J Thorac Oncol, 13, 1302-1311 (2018). PMID: 29800747.
- S Martinez-Morilla et al., “Quantitative Assessment of PD-L1 as an Analyte in Immunohistochemistry Diagnostic Assays using a Standardized Cell Line Tissue Microarray” - under review.
- L Fehrenbacher et al., “Atezolizumab versus docetaxel for patients with previously treated non-small-cell lung cancer (POPLAR): a multicentre, open-label, phase 2 randomised controlled trial”, Lancet, 387, 1837–1846 (2016). PMID: 26970723.