A Standardized Investigational Ki-67 Immunohistochemistry Assay Used to Assess High-Risk Early Breast Cancer Patients in the monarchE Phase 3 Clinical Study Identifies a Population With Greater Risk of Disease Recurrence When Treated With Endocrine Therapy Alone.

The objectives were to develop a standardized Ki-67 immunohistochemistry (IHC) method for precise, robust, and reproducible assessment of patients with early breast cancer, and utilize this assay to evaluate patients participating in the monarchE study (NCT03155997). The Ki-67 assay was developed and validated for sensitivity, specificity, repeatability, precision, and robustness using a predefined ≥20% cutoff. Reproducibility studies (intersite and intrasite, interobserver and intraobserver) were conducted at 3 external laboratories using detailed scoring instructions designed for monarchE. Using the assay, patient tumors were classified as displaying high (≥20%) or low (<20%) Ki-67 expression; Kaplan-Meier methods evaluated 2-year invasive disease-free survival rates for these 2 groups among patients treated with endocrine therapy (ET) alone. All analytical validation and reproducibility studies achieved point estimates of >90% for negative, positive, and overall percent agreement. Intersite reproducibility produced point estimate values of 94.7%, 100.0%, and 97.3%. External interobserver reproducibility produced point estimate values of 98.9%, 97.8%, and 98.3%. Among 1954 patients receiving ET alone, 986 (50.5%) had high and 968 (49.5%) had low Ki-67 expression. Patients with high Ki-67 had a clinically meaningful increased risk of developing invasive disease within 2 years compared with those with low Ki-67 [2-y invasive disease-free survival rate: 86.1% (95% confidence interval: 83.1%-88.7%) vs. 92.0% (95% confidence interval: 89.7%-93.9%), respectively]. This standardized Ki-67 methodology resulted in high concordance across multiple laboratories, and its use in the monarchE study prospectively demonstrated the prognostic value of Ki-67 IHC in HR+, HER2- early breast cancer with high-risk clinicopathologic features.

Comparative analysis of machine learning approaches to classify tumor mutation burden in lung adenocarcinoma using histopathology images.

Both histologic subtypes and tumor mutation burden (TMB) represent important biomarkers in lung cancer, with implications for patient prognosis and treatment decisions. Typically, TMB is evaluated by comprehensive genomic profiling but this requires use of finite tissue specimens and costly, time-consuming laboratory processes. Histologic subtype classification represents an established component of lung adenocarcinoma histopathology, but can be challenging and is associated with substantial inter-pathologist variability. Here we developed a deep learning system to both classify histologic patterns in lung adenocarcinoma and predict TMB status using de-identified Hematoxylin and Eosin (H&E) stained whole slide images. We first trained a convolutional neural network to map histologic features across whole slide images of lung cancer resection specimens. On evaluation using an external data source, this model achieved patch-level area under the receiver operating characteristic curve (AUC) of 0.78-0.98 across nine histologic features. We then integrated the output of this model with clinico-demographic data to develop an interpretable model for TMB classification. The resulting end-to-end system was evaluated on 172 held out cases from TCGA, achieving an AUC of 0.71 (95% CI 0.63-0.80). The benefit of using histologic features in predicting TMB is highlighted by the significant improvement this approach offers over using the clinical features alone (AUC of 0.63 [95% CI 0.53-0.72], p = 0.002). Furthermore, we found that our histologic subtype-based approach achieved performance similar to that of a weakly supervised approach (AUC of 0.72 [95% CI 0.64-0.80]). Together these results underscore that incorporating histologic patterns in biomarker prediction for lung cancer provides informative signals, and that interpretable approaches utilizing these patterns perform comparably with less interpretable, weakly supervised approaches.

Clinical Utility of the Combined Positive Score for Programmed Death Ligand-1 Expression and the Approval of Pembrolizumab for Treatment of Gastric Cancer.

CONTEXT.­: Regulatory approval of pembrolizumab for treatment of gastric and gastroesophageal junction (G/GEJ) adenocarcinoma required a reproducible scoring method for use of programmed death ligand-1 (PD-L1) protein expression as a companion diagnostic to identify likely responders to therapy. OBJECTIVE.­: To develop an immunohistochemical scoring algorithm that includes PD-L1 expression for tumor and immune cells, that is, the combined positive score. DESIGN.­: Four previously treated tumor types in the KEYNOTE-012 and KEYNOTE-028 studies were analyzed descriptively with a version of the PD-L1 immunohistochemical 22C3 pharmDx assay labeled for investigational use only to determine the relative importance of PD-L1 expression in tumor versus immune cells as a biomarker for pembrolizumab response. A combined positive score was developed as a novel scoring method and was compared with the tumor proportion score in cohort 1 from the KEYNOTE-059 study (G/GEJ cancer). External reproducibility was assessed. RESULTS.­: Per combined positive score cutoff of 1 or more, the prevalence of PD-L1 expression in patients with G/GEJ cancer was 57.6% (148 of 257 patients), with reasonable enrichment of responses (odds ratio, 2.8). Per tumor proportion score cutoff of 1% or more, prevalence was 12.5% (32 of 257 patients), with minimal enrichment (odds ratio, 1.4). External reproducibility assessments demonstrated interpathologist overall agreement of 96.6% (591 of 612; 95% CI, 94.0%-98.7%) and intrapathologist overall agreement of 97.2% (595 of 612; 95% CI, 95.3%-98.9%). CONCLUSIONS.­: Combined positive score is a robust, reproducible PD-L1 scoring method that predicts response to pembrolizumab in patients with G/GEJ cancer. This novel scoring method supported US Food and Drug Administration approval of pembrolizumab as third-line therapy for G/GEJ cancer and has facilitated investigation in other indications.

PD-L1 Immunohistochemistry Assays for Lung Cancer: Results from Phase 1 of the Blueprint PD-L1 IHC Assay Comparison Project.

INTRODUCTION: The Blueprint Programmed Death Ligand 1 (PD-L1) Immunohistochemistry (IHC) Assay Comparison Project is an industrial-academic collaborative partnership to provide information on the analytical and clinical comparability of four PD-L1 IHC assays used in clinical trials. METHODS: A total of 39 NSCLC tumors were stained with four PD-L1 IHC assays (22C3, 28-8, SP142, and SP263), as used in the clinical trials. Three experts in interpreting their respective assays independently evaluated the percentages of tumor and immune cells staining positive at any intensity. Clinical diagnostic performance was assessed through comparisons of patient classification above and below a selected expression cutoff and by agreement using various combinations of assays and cutoffs. RESULTS: Analytical comparison demonstrated that the percentage of PD-L1-stained tumor cells was comparable when the 22C3, 28-8, and SP263 assays were used, whereas the SP142 assay exhibited fewer stained tumor cells overall. The variability of immune cell staining across the four assays appears to be higher than for tumor cell staining. Of the 38 cases, 19 (50.0%) were classified above and five (13%) were classified below the selected cutoffs of all assays. For 14 of the 38 cases (37%), a different PD-L1 classification would be made depending on which assay/scoring system was used. CONCLUSIONS: The Blueprint PD-L1 IHC Assay Comparison Project revealed that three of the four assays were closely aligned on tumor cell staining whereas the fourth showed consistently fewer tumor cells stained. All of the assays demonstrated immune cell staining, but with greater variability than with tumor cell staining. By comparing assays and cutoffs, the study indicated that despite similar analytical performance of PD-L1 expression for three assays, interchanging assays and cutoffs would lead to "misclassification" of PD-L1 status for some patients. More data are required to inform on the use of alternative staining assays upon which to read different specific therapy-related PD-L1 cutoffs.