Abstract 297P
Background
TNBC, an aggressive subtype of breast carcinoma, has limited treatment options. Improved progression free and overall survival has been observed in patients with unresectable locally advanced and metastatic TNBC (NCT02425891) with atezolizumab and nab-paclitaxel in PD-L1-positive [≥1% on immune cells using VENTANA SP142 PD-L1 IHC]. Although this predictive marker is pivotal for treatment decision making, there are limited data on inter- and intra-pathologist (P) agreement for SP142 PD-L1 scoring. We investigated reproducibility of SP142 IHC in TNBC amongst Ps trained in SP142 scoring and the effect of training on accuracy of assessment by Ps untrained in PD-L1.
Methods
Previously untreated, resected primary invasive TNBCs were collected from 4 centers. SP142 PD-L1 IHC was performed on tissue microarrays. 60 cases, selected for PD-L1 expression close to 1% cut-off, were dichotomized as negative (<1%) or positive (≥1%). Intra- and inter-observer reproducibility assessment involved 10 Ps, conducted over 2 days (D). Five Ps were previously trained in SP142 IHC (subgroup - SG1) and 5 were SP142 assessment naïve (SG2). SG2 received 1 hour of training in SP142 IHC prior to scoring on D2; pre- and post-training scores were compared. Intra-observer reproducibility was evaluated by comparing the scores obtained by SG1 on D1 and D2. Inter-observer reproducibility compared scores between SG1 on D1.
Results
High intra-observer agreement was demonstrated between SG1; overall percent agreement (OPA) 95.0% (95% CI 91.9– 97.0), average positive agreement (APA) 95.2% (95% CI 92.2 – 97.0), average negative agreement (ANA) 94.9% (95% CI 91.7 - 96.9), Cohen’s kappa co-efficient (K) = 0.90. Inter-observer agreement was also high; OPA 93.3% (95% CI 91.1 - 95.1), APA 93.7% (95% CI 91.5 - 95.3), ANA 93.0% (95% CI 90.6 - 94.8), K = 0.87. Following D2 training, SG2 demonstrated significant improvement in SP142 agreement; OPA D1 81.5% (95% CI 76.8 – 85.5); D2 85.7% (95% CI 81.3 - 89.2), p=0.0499. However, OPA was higher in SG1, 96.3% (95% CI 93.6 - 97.9).
Conclusions
High intra- and inter-observer reproducibility was demonstrated for SP142 PD-L1 assay amongst Ps with minimal specific training in TNBC SP142 IHC.
Clinical trial identification
Editorial acknowledgement
Legal entity responsible for the study
Roche Products Pty. Limited.
Funding
Roche Products Pty. Limited.
Disclosure
J-M. Pang: Research grant/Funding (institution): Roche Products Pty. Ltd. B. Castles: Shareholder/Stockholder/Stock options, Full/Part-time employment: Roche Products Pty. Ltd. D.J. Byrne: Research grant/Funding (institution): Roche Products Pty. Ltd. P. Button: Advisory/Consultancy: Roche Products Pty. Ltd. S. Lakhani: Advisory/Consultancy, Travel/Accommodation/Expenses: Roche Products Pty. Ltd; Officer/Board of Directors: Breast Cancer Trial Australia and New Zealand. V. Sivasubramaniam: Advisory/Consultancy, Travel/Accommodation/Expenses: Roche Products Pty. Ltd. W. Cooper: Travel/Accommodation/Expenses: Roche Products Pty. Ltd. J. Armes: Advisory/Consultancy, Travel/Accommodation/Expenses: Roche Products Pty. Ltd. E. Millar: Travel/Accommodation/Expenses: Roche Products Pty. Ltd. W. Raymond: Travel/Accommodation/Expenses: Roche Products Pty. Ltd. J. Beith: Honoraria (self), Advisory/Consultancy, Travel/Accommodation/Expenses: Roche Products Pty. Ltd. S.A. O'Toole: Honoraria (self), Advisory/Consultancy, Travel/Accommodation/Expenses: Roche Products Pty. Ltd; Honoraria (self): BMS; Honoraria (self): Merck. S.B. Fox: Honoraria (institution), Advisory/Consultancy, Research grant/Funding (institution): Roche Products Pty. Ltd; Advisory/Consultancy, Research grant/Funding (institution): BMS; Advisory/Consultancy, Research grant/Funding (institution): AstraZeneca; Advisory/Consultancy: Bayer; Advisory/Consultancy: Pfizer; Advisory/Consultancy: Amgen. All other authors have declared no conflicts of interest.