Data Mining Used To Characterize Discordance In Gastric Cancers HER2 Status Determination To Help For A Better Treatment Decision

Full text

Data mining used to characterize discordance in gastric cancers
HER2 status determination to help for a better treatment decision.
Authors: 1David Pau, 2Geneviève Monges, 3Benoit Terris, 4Marie-Pierre Chenard, 5Frédéric Bibeau, 6Frédérique Penault-Llorca, 1Jessica Martin, 7Justine Rabut, , 8David Leroux, 9Laurent Doucet
1Roche, Boulogne-Billancourt, France; 2Hôpital Pasteur, Nice, France; 3Hôpital Cochin, Paris, France; 4Hôpital Hautepierre, Strasbourg, France; 5Institut du Cancer de Montpellier, France; 6Centre Jean Perrin, Clermont Ferrand, France; 7Lincoln system, Boulogne-Billancourt, France; 8Hays Pharma, Paris, France; 9Hôpital Cavale Blanche CHU de Brest, France
BACKGROUND
Following the results of the ToGA trial (1), Herceptin® (trastuzumab) was approved for the
treatment of HER2 positive (+) metastatic gastric or gastroesophageal junction (GEJ) tumors
in January 2010 in France. From then, HER2 testing should be part of routine assessment to
individualize therapeutic strategies and to decide target treatment initiation.
OBJECTIVE
Considering the specifi c heterogeneity of these tumors, an optimal HER2 scoring system was
developed. As only limited data were available on assays for HER2 scoring on gastric and GEJ
tumors, this study mainly aimed to evaluate the agreement between HER2 status assessed by
local (any technique) and centralized [immunohistochemistry (IHC) 4B5 and silver In Situ
Hybridization (SISH) for all specimens] laboratories.
We used data from this study to perform exploratory analysis to evaluate variables
infl uencing discordance between a local and central evaluation of HER2 overexpression
determination.
METHODS
Study description
• HERable was a French non-interventional, histological study conducted between July 2012
and february 2014. 19 centers (including 13 CHU) and 4 centralized laboratories have
participated to the study.
• Samples collected: Gastric or GEJ adenocarcinoma specimens of all stages, histologically documented, fi xed in paraffi n,
and originated from patients with no objection to participate in the study were included.
• HER2 positive status (HER2+) was defi ned as immunohistochemistry (IHC) 3+ or IHC2+/In Situ Hybridization
(ISH)+(2)
• The agreement between HER2 status assessed by local and centralized laboratories was evaluated using the kappa
coeffi cient. (3)
• Methodology and primary results of HERable study were presented as a poster at ASCO GI 2015. (4)
Analysis method
• False negative (HER2+ according to central laboratory and HER2– according to local laboratory ) and false positive
(HER2- according to central laboratory and HER2+ according to local laboratory ) were studied in this exploratory
analysis.
The following results focus only on false negative discordance, because of the low percentage of false positive.
• Analysis method: decision trees (5) are supervised datamining analysis methods which allowed to detect false negative
cases by:
– Obtaining subgroups with high percentage of false negative from simple decision rules.
– Identifying at each step the most infl uential factor. For continuous factors, the method can determine the most
discriminant threshold.
• Software used for this analysis was R version 3 (‘rpart’ package). For all others analysis SAS 9.4 software was used.
RESULTS
394 specimens from 367 patients were analyzed. These specimens included 162 biopsies,
206 surgical specimens, 15 adenopathies and 11 distant metastases.
Table 1: Patients and disease characteristics
Patients population
N= 367
Age (years) – Mean (SD) 66.4 (13.2)
Gender: Female – n (%) 113 (30.8)
Initial location of the adenocarcinoma: Stomach – n (%) 218 (59.6)
Histological type according to Lauren classifi cation (6) – n (%)
Intestinal 200 (54.8)
Diffuse 108 (29.6)
Mixed 35 (9.6)
Table 2: Agreement between HER2 status assessed by local
and central laboratories
Specimens population (N=394)
Centralized laboratory: HER2 status
+ N=73 - N=321 All N=394
Local laboratory: HER2 status N 73 320 393
+53 (72.6%) 16 (5.0%) 69 (17.6%)
-20 (27.4%) 304 (95.0%) 324 (82.4%)
In table 2, the overall discordance rate was 9 % (95% CI [6-12]), with 27% (20/73) of false
negative and 5 % (16/320) of false positive.
The true concordance between centralized and local laboratory HER2 testing was
acceptable, with a kappa coeffi cient of 0.69 (95% CI [0.60; 0.78]).
This non-interventional HERable study showed that the true concordance between
central and local laboratory HER2 testing in gastric and GEJ adenocarcinoma
specimens was acceptable with a kappa value of 0.69 (95% CI [0.60-0.78]).
The cumulative rate of false negatives and positives in gastric and GEJ tumors
was 9% (95% CI [6-12]).
REFERENCES
1. Bang YJ, et al. Lancet 2010; 376: 687-97.
2. Ruschoff J, et al. Mod Pathol 2012; 5: 637-50
3. Landis & Koch, Biometrics, 1977
4. Monges G, and al. Presented at the 2015 ASCO GI annual meeting, USA
5. Tufféry S, Data mining et statistique décisionnelle, 2012, Technip
6. Lauren P. Ann Pathol Microbiol Scand 1965; 64: 31-49
The HERable study was founded by Roche France
CONFERENCE NAME
ISPOR EUROPE 2016, VIENNA, AUSTRIA
73 samples (Central HER2 +)
27 % of false negative
(20 cases)
28 samples
54 % of false negative
(15 cases)
35 samples
0 % of false negative
(no cases)
10 samples
50 % of false negative
(5 cases)
45 samples
11 % of false negative
(5 cases)
Central lab: IHC percentage
of stained cells <60%
• 100% of heterogeneous (central)
• 53% of ISH clusters (central)
• 79% of local IHC 2+
• 14% of heterogeneous (central)
• 85% of ISH clusters (central)
• 0% of local IHC 2+
• 40% of heterogeneous (central)
• 40% of ISH clusters (central)
• 100% of local IHC 2+
Central lab: IHC percentage
of stained cells ≥60%
Local lab IHC 0,1+, 3+ Local lab IHC 2+
Figure 1: Decision tree for modeling false negative
Using the 73 samples HER2+ within the central laboratories, the 20 false negative were characterized as follow:
• 15 cases had < 60% IHC stained cells centrally and all tumor were heterogeneous
• 5 cases had >= 60% IHC stained cells centrally and were all locally IHC2+
#PMD6
Data mining analysis showed that the main reason for HER2 status discordance between local and
centralized analyses was the tumor heterogeneity in gastric cancer. Given the short history of the HER2
scoring system developed for gastric cancer, a retesting of specimens by a centralized ISH laboratory
could be advisable to ensure the best therapeutic decision.
Data mining are useful techniques to characterize a sub-population of interest in a clinical study.
CONCLUSION