HER2 Immunohistochemistry: Workflow Experience with Image Analysis Based Interpretation of
CB11 and 4B5 Clones
Jeffrey Fine MD, Rohit Bhargava MD, Urvashi Surthi PhD, and David Dabbs MD
Magee-Womens Hospital of UPMC
Disclaimer
• One of the abstract authors (David Dabbs) is a consultant for Ventana
• None of the other authors have any conflicts of interest to report
Objectives
• Overview of Her2/NEU testing (IHC)
• Recap of validation results
• Discuss implementation of image analysis
Her2/NEU
• Test for responsiveness to Trastuzumab (Herceptin)
• Trastuzumab is cardiotoxic and is very expensive– False positives are highly undesirable
• Trastuzumab can increase survival or reduce risk of recurrence– False negatives are also undesirable
CAP HER2-A Survey (Spring 2007)
• 40 TMA cores (4 slides)
• Stained/interpreted at institutions (350ish)
• Consensus (>80%) in 22 of 40 cases
• Very variable
IHC Variables
• Pre-analytic– Fixation issues, tissue processing
• Analytic– Validation, Calibration, Antibody clone,
Antigen retrieval, Automation, Controls, etc.
• Post-analytic– Interpretation criteria– QA procedures
• Image analysis (17.9% reported using it)
Validation of IA at UPMC
• Formalin fixed (8-48 hours) paraffin embedded tissue
• Automated IHC platform (Ventana)
• CB11 and 4B5 antibodies (Ventana)
• VIAS (Ventana Image Analysis System)
• FISH (Vysis)
VIAS (image from vendor)
Validation Results I
• System differentiated between tumor and stroma (subjective impression)
• Pathologist had to find invasive tumor (and exclude in-situ tumor)
Classification (Tumor vs. Stroma)
Results CB11
• 100% Concordance with FISH (n=52)– 0/1+ IHC with no amp by FISH– 3+ with amp by FISH
• FISH rate (2+ rate):– Expert 22.9% (n=118)– VIAS 21.2%
• (expert was also 100% concordant)
Results 4B5
• 94.6% Concordance with FISH (n=56)– 100% Concordance with new reference range
• FISH rate (2+):– Expert 21.9% (n=114)– VIAS 28.9% (n=117)
• *new reference range
Reference Range
• VIAS assigns a raw number score to each case which is then rounded to the nearest whole number.
• Out of the box—score 2.5 or higher was rounded to “3+”
• New range is conservative—only cases with score 3.5 are called “3+”
Demixed to show brown
Old Workflow
Order Her2/NEU
Retrieve IHCStain
InterpretStain
Dictate Results
Sign out case
MailboxOffice
New Workflow
Order Her2/NEU
Retrieve IHCStain
Perform IA
Dictate Results
Sign out case
Office MailboxIA Workstation
Implementation Details
• Billing– Fee code 88361– Technical charge initiated by order in APLIS
• Documentation– VIAS results printout retained with other case
paperwork (requisition)– IHC results (ER/PR and Her2/NEU) dictated
into canned text that includes VIAS blurb
Documentation Support
• Transcription– New “quick text” with VIAS sentence– Communication with Transcription Team
• Slide/paper management personnel– Communication – do not discard results report
• Back-up of data– Currently performed manually
“Consumables”
• IA charged per “click”
• IHC Laboratory responsible for keeping an adequate supply of the click reagent– Smart Card (100 tests)
Image from nist.gov www site
Training
• Local Domain Expert (me)
• Fellows
• Selected Faculty
• Other Faculty
• Residents
Status Report
• IA is in production for a week
• Bumps being ironed out– Transcription– Training
• End experience varies
Good
• Should increase standardization– Recent switch to 4B5 clone – different
appearance and possible differences in interpretation
• Documented response to pressure for accurate Her2/NEU testing
• Foot in the door for other IA applications• Vendor has been responsive and appears
to want to improve shortcomings
Bad
• IA of new antibody is less accurate than that of discontinued antibody– Reference range work-around
• Workflow involves “travel” and is more labor intensive than traditional method– Quality improvement but does not extend the pathologist
• Operator error is possible– Data entry– Lighting– Focusing
Future IA (More Automation)
• Whole Slide Images (some systems do permit IA)– Slide could be scanned in the IHC lab, and results
(with the electronic slide) delivered straight to the pathologist
• Automated detection of invasive tumor– Transition of IA to non-pathologist staff
• Electronic interfaces to support test ordering and resulting (no more paper print outs or dictation)
Conclusions
• IA is validated and should improve performance of Her2/NEU IHC testing by reducing post-analytic variability
• Current IA set-up not ideal but an important first step:– Successful implementation in a busy
academic setting– Revenue (digital pathology business case)