Virtual Library
Start Your Search
Michael P A Davies
Author of
-
+
P1.04 - Immunooncology (Not CME Accredited Session) (ID 936)
- Event: WCLC 2018
- Type: Poster Viewing in the Exhibit Hall
- Track:
- Presentations: 1
- Moderators:
- Coordinates: 9/24/2018, 16:45 - 18:00, Exhibit Hall
-
+
P1.04-24 - Digital Core Needle-Biopsy to Assess PD-L1 Expression in Non-Small Cell Lung Cancer: Optimal Sampling and Need for Re-Biopsy (ID 12059)
16:45 - 18:00 | Author(s): Michael P A Davies
- Abstract
Background
Assessing expression of PD-L1 on tumour cell membranes by immunochemistry is an important complementary or crucial companion diagnostic test to guide the use of immune modulating checkpoint inhibitors in the treatment of non-small cell lung cancer (NSCLC). Unfortunately, the known temporal and spatial heterogeneity of PD-L1 expression raises the important question of how to ensure that the small biopsy specimens with which this assessment is usually made are adequately representative of PD-L1 expression by the whole tumour.
a9ded1e5ce5d75814730bb4caaf49419 Method
Expression of PD-L1 was assessed in sections of 94 tissue blocks from 50 primary pulmonary adenocarcinomas using the Ventana SP263 antibody and a validated protocol. Scoring was performed by two appropriately-trained pathologists with extensive experience in its interpretation. After conventional assessment, slides were digitally scanned and divided into squares of 1mm² area to form a digital database (mean of 150 data-points per tumour), which were assigned co-ordinates and re-scored. By these means, multiple, “digital core biopsies” (DCBx) approximating a 17 gauge needle were simulated in sequential fashion, and expression in these was compared to that in the whole tumour and categorised by current UK prescribing guidelines*
4c3880bb027f159e801041b1021e88e8 Result
8eea62084ca7e541d918e823422bd82e ConclusionPD-L1 score (%)
Total number of cases
Cases where PD-L1% from single DCBx changed scoring category vs whole tumour*
Cases where PD-L1% from two DCBx changed scoring category vs whole tumour*
Cases where PD-L1% from three DCBx changed scoring category vs whole tumour*
Cases where PD-L1% from four DCBx changed scoring category vs whole tumour*
Cases where PD-L1% from five DCBx changed scoring category vs whole tumour*
Focal expression primary pattern in non-correlative cases
<1
14
2
2
2
2
2
Y
1-10
13
6
3
1
1
0
Y
11-49
10
1
1
0
0
0
Y
50-100
13
0
0
0
0
0
n/a
All
50
9 (18%)
6 (12%)
3 (6%)
3 (6%)
2 (4%)
PD-L1, programmed death ligand 1; DCBx, Digital Core Biopsy
*Based on pembrolizumab categories as: 1st line ≥50%, 2nd line 1-49%; nivolumab categories as: ≥1% (for adenocarcinoma)
In the majority of cases, three digital core biopsies achieved closest correlation with the whole tumour, with little greater accuracy achieved by assessing four cores or more. Correlation was weakest when expression was low and very focal, an important consideration in view of the importance of the ‘1% cut-off’ used commonly to guide immune checkpoint therapy. Using this model as a guide, a single good quality biopsy (2x10mm² area) is sufficient for most tumours scoring 11% or greater PD-L1 expression. However, in the lower range of expression, re-biopsy might be routinely considered if there is doubt about specimen adequacy.
6f8b794f3246b0c1e1780bb4d4d5dc53
-
+
P2.09 - Pathology (Not CME Accredited Session) (ID 958)
- Event: WCLC 2018
- Type: Poster Viewing in the Exhibit Hall
- Track:
- Presentations: 1
- Moderators:
- Coordinates: 9/25/2018, 16:45 - 18:00, Exhibit Hall
-
+
P2.09-19 - Utilising Heterogeneity: Using a Digital Database of Lung Cancers and Immune Profile to Complement Subjective Assessment (ID 12594)
16:45 - 18:00 | Author(s): Michael P A Davies
- Abstract
Background
Traditional pathological assessment of tissue sections involves subjective analysis of complex and heterogeneous features, typified by the challenge of ‘measuring’ PD-L1 expression in non-small cell lung cancer (NSCLC) as a guide to its treatment with immune checkpoint inhibitors. Such heterogeneity is generally perceived as a problem but might, in fact, reflect not only biologically important epitope variation, but also important features of the tumour microenvironment and, by extension, be a tool for predicting behaviour. In-depth analysis of a single slide of a tumour by digital pathology, image analysis and machine learning makes more accurate and meaningful analysis a possibility.
a9ded1e5ce5d75814730bb4caaf49419 Method
Expression of PD-L1 was assessed by immunochemistry in 250 sections from 137 resected NSCLCs using the Ventana SP263 antibody and a validated protocol and its distribution compared with morphology as revealed by corresponding H&E-stained sections. Slides were scanned to create a digital image using Aperio Scanscope with division of images into 1mm² squares using QuPath opensource software, each of which was assigned x and y co-ordinates. Squares were assessed subjectively by two pathologists for morphological features and PD-L1 expression and also subject to automatic image analysis including cell counting and membrane detection. Co-ordinates and values were stored in Microsoft Excel and a digital database was generated for every slide. In-depth analysis of digital data points was achieved using “R” software custom algorithms that included simulating biopsy sampling and applying spatial analysis packages.
4c3880bb027f159e801041b1021e88e8 Result
The resulting database, comprising approximately 30,000 data points from the 137 tumours, is being used to simulate needle-core biopsies, assess heterogeneity of PD-L1 expression and relate this to the tumour micro-environment including immune cell populations, immune signature and tumour mutational burden.
8eea62084ca7e541d918e823422bd82e Conclusion
The vast amount of information in every NSCLC cannot be extracted by conventional histopathological analysis. By utilising new technologies and considering alternative paradigms for data acquisition, powerful new approaches may be developed that give information pertaining to not just diagnostic and prognostic features of a tumour, but behavioural traits including likely responses and resistances to novel drugs such as immune checkpoint inhibitors. The methodology described here is an attempt to extract these data in a more objective way and complement the still crucial subjective analysis that is traditionally the prerogative of the histopathologist.
6f8b794f3246b0c1e1780bb4d4d5dc53
