Faculty Bibliography

Immunohistochemical (IHC) stain for PD-L1 as a biomarker for immunotherapy is recommended in non-small cell lung cancer (NSCLC). Under the FDA, the selection of patients for pembrolizumab requires companion diagnostic testing using the Dako Agilent PD-L1 IHC 22C3 pharmDx kit performed on the Dako Autostainer Link 48 platform. However, because it is not widely available, there is need for cross-platform validation. Existing studies provide incomplete protocol detail. In our study, 73 lung tumors were stained using the FDA-approved test ('gold standard'). The same blocks were stained using two different models of the Ventana DISCOVERY platform (ULTRA, n = 73 and XT, n = 70) using different parameters, and interpreted by three pathologists. The ULTRA group met College of American Pathologists (CAP) validation criteria (concordance 91.8%) while the XT group did not (concordance 67.1%). Using tumor proportion score (TPS) ≥1% and TPS ≥50% as cut-offs, the ULTRA protocol had higher sensitivity (97.8% and 91.7%) than XT (73.3% and 60.9%) and similar specificity (ULTRA 88.9% and 100%, XT 88% and 100%). Discordance between ULTRA and XT was 27%, and in all these cases ULTRA was concordant with gold standard. Interobserver reliability was substantial for ULTRA and almost perfect for XT, providing evidence that staining rather than observer variability accounts for XT's inferior performance. Cross-validation of the clinically used 22C3 anti PD-L1 antibody test with substantial interobserver agreement is possible on the commonly used the Ventana DISCOVERY ULTRA automated instrument, while the validation failed on the XT. Cautious attention to detail must be paid when choosing cross-validation parameters.

BACKGROUND:Increased utilization of hepatitis C virus (HCV)-positive donors has increased transplantation rates. However, high levels of viremia have been documented in recipients of viremic donors. There is a knowledge gap in how transient viremia may impact acute cellular rejections (ACRs). METHODS:In this study, 50 subjects received hearts from either viremic or non-viremic donors. The recipients of viremic donors were classified as nucleic acid amplification testing (NAT)+ group, and the remaining were classified as NAT-. All patients were monitored for viremia levels. Endomyocardial biopsies were performed through 180 days, evaluating the incidence of ACRs. RESULTS:A total of 50 HCV-naive recipients received hearts between 2018 and 2019. A total of 22 patients (44%) who received transplants from viremic donors developed viremia at a mean period of 7.2 ± 0.2 days. At that time, glecaprevir/pibrentasvir was initiated. In the viremia period (<56 days), 14 of 22 NAT+ recipients (64%) had ACR vs 5 of 28 NAT- group (18%) (p = 0.001). Through 180 days, 17 of 22 NAT+ recipients (77%) had a repeat rejection biopsy vs 12 of 28 NAT- recipients (43%) (p = 0.02). NAT+ biopsies demonstrated disparity of ACR distribution: negative, low-grade, and high-grade ACR in 84%, 12%, and 4%, respectively, vs 96%, 3%, and 1%, respectively, in the NAT- group (p = 0.03). The median time to first event was 26 (interquartile range [IQR]: 8-45) in the NAT+ group vs 65 (IQR: 44-84) days in the NAT-. Time to first event risk model revealed that NAT+ recipients had a significantly higher rate of ACR occurrences, adjusting for demographics (p = 0.004). CONCLUSIONS:Transient levels of viremia contributed to higher rates and severity of ACRs. Further investigation into the mechanisms of early immune activation in NAT+ recipients is required.

INTRODUCTION/BACKGROUND:A grading system for pulmonary adenocarcinoma has not been established. The IASLC pathology panel evaluated a set of histological criteria associated with prognosis aimed at establishing a grading system for invasive pulmonary adenocarcinoma. DESIGN/METHODS:A multi-institutional study involving multiple cohorts of invasive pulmonary adenocarcinomas was conducted. A cohort of 284 stage I pulmonary adenocarcinomas was used as a training set to identify histological features associated with patient outcomes (recurrence-free survival [RFS] and overall survival [OS]). ROC curve analysis was used to select the best model, which was validated (n=212) and tested (n=300, including Stage I-III) in independent cohorts. Reproducibility of the model was assessed using kappa statistics. RESULTS:The best model (AUC= 0.749 for RFS and 0.787 for OS) was composed of a combination of predominant plus high-grade histological pattern with a cut off of 20% for the latter. The model consists of: Grade 1: lepidic predominant tumor; Grade 2: acinar or papillary predominant tumor, both with no or less than 20% of high-grade patterns; and Grade 3: any tumor with 20% or more of high-grade patterns (solid, micropapillary and or complex gland). Similar results were seen in the validation (AUC = 0.732 for RFS and 0.787 for OS) and test cohorts (AUC of 0.690 for RFS and 0.743 for OS), confirming the predictive value of the model. Inter-observer reproducibility showed good agreement (k=0.617). CONCLUSION/CONCLUSIONS:A grading system based on the predominant and high-grade patterns is practical and prognostic for invasive pulmonary adenocarcinoma. ACKNOWLEDGEMENT/UNASSIGNED:This project was supported by grants from the International Association for the study of Lung Cancer (IASLC), and NIH/NCI Early Detection Research Network 1U01CA214195-01 to HP used for biospecimen collection and salary support (DFL). We would like to thank the staff of the Center of Biospecimen Research and Development (CBRD) from the New York University for their help in digital pathology and histology for this project. CBRD is partially supported by the Cancer Center Support Grant P30CA016087 at the Laura and Isaac Perlmutter Cancer Center.

In the 21st century, there has been a dramatic shift in the management of advanced-stage lung carcinoma, and this has coincided with an increasing use of minimally invasive tissue acquisition methods. Both have had significant downstream effects on cytology and small biopsy specimens. Current treatments require morphologic, immunohistochemical, and/or genotypical subtyping of non-small cell lung carcinoma. To meet these objectives, standardized classification of cytology and small specimen diagnoses, immunohistochemical algorithms, and predictive biomarker testing guidelines have been developed. This review provides an overview of current classification, biomarker testing, methods of small specimen acquisition and triage, clinical management strategies, and emerging technologies.

OBJECTIVES/OBJECTIVE:Recent investigations have shown strong correlations between cytology and surgical non-small cell lung carcinoma (NSCLC) specimens in programmed death-ligand 1 (PD-L1) immunohistochemical (IHC) evaluations. Our study aims to evaluate the reproducibility of PD-L1 IHC scoring in NSCLC cytology cell blocks (CBs) and to assess the impact of CB cellularity, method of sample collection, and observer subspecialty on scoring agreement. METHODS:PD-L1 IHC was performed on 54 NSCLC cytology CBs and was scored independently by seven cytopathologists (three of seven with expertise in pulmonary pathology). Three-tier scoring of negative (<1%), low positive (1%-49%), and high positive (≥50%) and interrater agreement were assessed. RESULTS:Total and majority agreement among cytopathologists was achieved in 48% and 98% of cases, respectively, with κ = 0.608 (substantial agreement; 95% confidence interval, 0.50-0.72). Cytopathologists with pulmonary pathology expertise agreed in 67% of cases (κ = 0.633, substantial agreement), whereas the remaining cytopathologists agreed in 56% of cases (κ = 0.62, substantial agreement). CB cellularity (P = .36) and sample collection type (P = .59) had no statistically significant difference between raters. CONCLUSIONS:There is substantial agreement in PD-L1 IHC scoring in cytology CB specimens among cytopathologists. Additional expertise in pulmonary pathology, sample collection type, and CB cellularity have no statistically significant impact on interobserver agreement.

Immune checkpoint inhibitor therapies have revolutionized the management of patients with non-small cell lung carcinoma (NSCLC) and have led to unprecedented improvements in response rates and survival in a subset of a patients with this fatal disease. However, the available therapies work only for a minority of patients, are associated with substantial societal cost, and may lead to significant immune-related adverse events. Therefore, patient selection must be optimized through use of relevant biomarkers. PD-L1 protein expression by immunohistochemistry is widely used today for selection of PD-1 inhibitor therapy in NSCLC patients, however this approach lacks both robust sensitivity and specificity for predicting response. Tumor mutation burden (TMB), or the number of somatic mutations derived from next generation sequencing techniques, has been widely explored as an alternative or complementary biomarker for response to immune checkpoint inhibitors. In theory, a higher TMB increases the probability of tumor neoantigen production and, therefore, likelihood of immune recognition and tumor cell killing. While TMB alone is a simplistic surrogate of this complex interplay, it is a quantitative variable that can be relatively readily measured using currently available sequencing techniques. A large number of clinical trials and retrospective analyses employing both tumor and blood-based sequencing tools have examined the performance of TMB as a predictive biomarker and in many cases show a correlation between high TMB and immune checkpoint inhibitor response rates and progression-free survival. Many challenges remain prior to implementation of TMB as a biomarker in clinical practice. These include: identification of therapies whose response is best informed by TMB status; robust definition of a predictive TMB cutpoint; acceptable sequencing panel size and design; need for robust technical and informatic rigor to generate precise and accurate TMB measurements across different laboratories. Finally, effective prediction of response to immune checkpoint inhibitor therapy will likely require integration of TMB with a host of other potential biomarkers, including tumor genomic driver alterations, tumor-immune milieu, and other features of the host immune system. This perspective piece will review the current clinical evidence for TMB as a biomarker and address the technical sequencing considerations and ongoing challenges to use of TMB in routine practice.

Background: Molecular screening for therapeutically targetable alterations is considered standard of care in the management of non-small cell lung cancer. However, most molecular assays utilize tumor tissue, which may not always be available. This has led to the development of "liquid biopsies": Plasma-based Next Generation Sequencing (NGS) tests that use circulating tumor DNA as a substrate to identify relevant targets. In this study, we sought to determine the level of agreement between the two tests as they are used in clinical practice and to investigate the utility of concurrent plasma/tissue testing.
Design(s): We identified 47 cases of lung adenocarcinoma diagnosed over the past 2 years, who received concurrent testing (within 24 weeks) with both our institution's tissue (DNA and RNA based) NGS assay and a commercial plasma-based NGS assay. The results were reviewed to establish concordance in the identification of mutations or fusions deemed clinically relevant or for which a targeted therapy was available.
Result(s): Patients in our cohort represented both new diagnoses (31 cases, 66%) and disease progression on treatment (16 cases, 34%). The majority (83%) had stage 4 disease. Tissue NGS identified clinically relevant mutations in 39 cases (83%), including in 14 (88%) of the previously treated cases. By comparison, plasma NGS identified clinically relevant mutations in 20 cases (43%, p<0.001), including 6 treated cases (38%, p=0.01). Tissue NGS identified therapeutic targets in 55% of cases and 75% of previously treated cases; while plasma NGS identified targets in 28% and 25% respectively (p<0.001 and p=0.01 respectively). All clinically relevant mutations identified by plasma NGS were also detected by tissue NGS, while plasma NGS detected only 51% those identified by tissue NGS. Discrepant cases involved hotspot mutations and actionable fusions including those in EGFR, KRAS, and ROS1 (Table 1).(Table presented)
Conclusion(s): Tissue NGS detects more clinically relevant alterations and therapeutic targets compared to plasma NGS, especially in the post-treatment setting, suggesting that tissue NGS should be the preferred method for molecular testing of lung adenocarcinoma. Additionally, all clinically relevant mutations identified by plasma NGS were also detected by tissue NGS, suggesting that tissue/plasma cotesting provides little additional benefit over tissue NGS alone. Plasma NGS can detect clinically relevant targets, and still plays an important role when tissue testing is impractical or not possible

Background: There is considerable interobserver variability in the pathological diagnosis of UIP and other ILD. The variability in finding "ground truth" presents a difficult problem during the development of deep learning platforms designed to diagnose ILD using AI. We describe the use of a smartphone application that allows for the collection of multiple opinions from sixteen pulmonary pathologists from nine countries as a novel method for finding "ground truth" for UIP diagnoses.
Design(s): Whole slide images from 14 consecutive interstitial lung disease cases were scanned by Aperio CS2 at 20x objective and sliced into multiple individual 7x7mm images, resulting in 355 JPEG image patches. These individual images were randomized and shown over the internet to 16 expert pathologists from 9 countries using the novel smartphone application, BonBon system. Expert lung pathologists were asked to classify each image into one of 8 categories: UIP/IPF; CTD/UIP; CHP/UIP; UIP/other cause; Non-UIP; "not sure"; "normal"; and "exclude". Images classified as "exclude" by any of the experts were deleted from analysis. All individual diagnostic classes selected for individual images were grouped by case, and the predominant class for each case, was used as "diagnosis". All diagnoses were analyzed using clustering analysis and Kaplan Meier statistics using JMP. Interobserver agreement was calculated by Fleiss kappa coefficient using R.
Result(s): The diagnosis of each of the 14 cases by each of the 16 pathologists showed that interobserver agreement using the 7 categories was poor (k=0.19). None of the cases were classified as CHP-UIP or UIP/others by a majority of pathologists. In order to simplify the diagnoses into more clinically relevant classes, the categories UIP/IPF, CHP-UIP, and UIP/others were grouped into UIP and Non-UIP and CTD-UIP into Non-UIP as shown in Figure 1. Clustering analysis stratified the diagnoses of each case into 3 clusters as shown in Figure 2A (k= 0.77, 0.63, and 0.03, respectively), validating cluster A and B were meaningful. Log rank test showed significant survival difference between UIP and Non-UIP groups only for cases in cluster A (P=0.017) (Fig 2B) but not for cases in cluster B (Fig 2 C). Eventually, 269 H&E images with high agreement of UIP and non-UIP group were selected as "ground truth" for AI training. (Figure presented)
Conclusion(s): Classification of image patches from WSI over APP offers a useful method to standardize UIP diagnosis and to develop diagnostic AI

Background: Spread through air spaces (STAS) has been reported to be associated with a worse prognosis in adenocarcinoma of lung. Recently it has been proposed that STAS be reported on frozen sections (FS) as an indication for more aggressive surgery (lobectomy vs sublobar resection). We undertook this study to evaluate the reliability of STAS assessment on FS compared to FS controls (FSC) and non-frozen remaining tumor (RT).
Design(s): Cases of adenocarcinoma that had FS of the tumor were identified retrospectively from two institutions. For each case, the following was recorded: Presence(+)/absence(-) of STAS on FS, FSC, and RT; and % of tumor patterns: Lepidic(L), acinar(A), papillary(P), micropapillary(M), solid(S). Grades were defined as follows: G1=L predominant with A/P G2=A/P predominant with <20% M/S G3=M/S predominant or >=20% M/S Cross-tabulations and Spearman's correlations (rs) were performed in SPSS (see Table). rs 0.3-<0.5 = low correlation.
Result(s): 165 cases were found. In 2 cases the tumor was only present on FS/FSC slides. STAS+ was present on FS in 47/165(28%), of which 28/47(60%) had STAS+ on FSC (rs=0.39) and 22/46(48%) had STAS+ on RT (rs=0.34) (1 tumor was only present on FS/FSC). Of the 40 STAS+ cases on RT, 18/40(45%) did not have STAS (STAS-) on FS (rs=0.34). 118/165 of cases were STAS- on FS; of these, 18/117(15%) were STAS+ on RT (rs=0.34) (1 tumor was only present on FS/FSC). Fig 1. Of the 47 cases with STAS+ on FS, 4(15%) were G1, 15(32%) were G2, and 28(60%) were G3 (rs =0.30). Of the 15 G2 cases with STAS+ on FS, 7 had 10% to <20% high grade pattern (M/S). Of the 28 G3 cases with STAS+ on FS, 13 had >= 30% high grade pattern. Of the 40 cases with STAS+ on RT, 1(2.5%) case was G1, 7(18%) cases were G2, and 32(80%) cases were G3 (rs=0.46). Fig 2. (Table presented)
Conclusion(s): The correlations among FS, FSC, and RT are low. FS STAS+ cases remain STAS+ only in 60% of FSC and 48% of RT. STAS shows higher correlation with grade on RT than it does on FS. These results show a lack of reliability in the assessment of STAS on FS, and do not support the proposal of reporting STAS in FS to make intraoperative clinical decisions, as doing so may subject patients to unnecessarily aggressive surgery

Background: When the first diagnosis of a well-differentiated neuroendocrine neoplasm (WDNEN) is made on a biopsy, crush artifact may impede mitotic counting, and the Ki-67 proliferative index (KPI) plays a more important role in grading. Few studies have examined the reliability of KPI in the grading of metastatic (met) WDNEN.
Design(s): Cases were retrieved from 2008-19. For each primary (1degree) and met, a 40x hotspot image of a Ki-67 stained slide was taken. KPI was analyzed by the ImmunoRatio Plugin using ImageJ software. All were validated by independent manual counting. In 1 case, crush artifact precluded use of ImageJ. If multiple mets were present, the largest was selected. Cytology and cases with fewer than 100 cells were omitted. Tumors were graded as G1(KPI<3%), G2(KPI 3-20%), or G3(KPI>20%). Grade and KPI were compared between each 1degree and met.
Result(s): 67 cases, including 38 mets from 29 1degree WDNEN, were evaluated. There were 12, 2, 2, 1, 17, and 4 mets from lung, thymus, middle ear, pancreas, small bowel, & colon respectively. The greatest variations in KPI from 1degree to met were seen in lung compared to all other sites. In 24/38(63%) of all mets, the grade was the same as 1degree. In 14/38(37%) of all mets, the grade was different: 4/14(29%) were lower while 10/14(71%) were higher than 1degree. In mets from lung, 2 had lower, 6 had the same, & 4 had higher grade. In mets from middle ear, the grades increased. In mets from small bowel 2 had lower, 12 had the same, and 3 had higher grade. In mets from colon 3 had the same and 1 had higher grade. And in mets from thymus & pancreas, the grades remained the same. See Fig 1. 3 mets were diagnosed before 1degree (2 days to 2 months). Mets with higher grade than 1degree tended to have longer time interval between 1degree and met, but this was not significant at the 95% confidence level (Kruskal-Wallis test, p=0.08). 28 mets were regional and 10 were distant. There was no correlation between met site and grade change (Person Chi-square, p=0.33). Also see Table 1. (Table presented)
Conclusion(s): Our study showed that in most met WDNEN (63%), the grade remained the same as the 1degree. The grade was higher in 10/38(26%) and lower in 4/38(11%). There was a trend toward higher grade in mets that occurred at a longer time interval after the 1degree. Fig 2 shows a potential diagnostic pitfall when the met shows much higher KPI than 1degree. In conclusion, if a met WDNEN is suspected as a firsttime diagnosis, KPI must be used cautiously for classification of WDNEN since over/under-grading may occur