Background: Her-2/neu overexpression, by immunohistochemistry (IHC) or fluorescence in-situ hybridization (FISH), is highly correlated with response to trastuzumab and these are currently the gold-standard, FDA-cleared testing methods for assigning treatment to Her-2-directed therapies. However, substantial variability has been documented between community and central laboratory IHC and FISH testing. Biologically, Her-2 overexpression may reflect increased gene copy number, gene expression and/or protein production, and these can be measured by other platforms, including comparative genomic hybridization (CGH), expression arrays and quantitative protein assays, respectively. We sought to determine the degree to which community IHC/FISH results differed from centrally-assessed IHC, FISH, and other assessment platforms within the I-SPY Trial and whether response to neoadjuvant chemotherapy (NAC) differed by platform.

Methods: The I-SPY Trial enrolled 237 women 2002–06 with invasive breast tumors at least 3 cm in clinical/radiographic size who subsequently underwent anthracycline/taxane NAC, serial core biopsies and imaging. Pathologic complete response (pCR) was determined at time of surgery and 3-year follow up has been reached. Trastuzumab was given to Her2+ patients at physician discretion, based upon community IHC/FISH results, and became more widespread after 2005. Central I-SPY laboratories determined Her2 copy number by MIP array, gene expression by Affymetrix and Agilent arrays, and Her2 protein by reverse-phase protein array (RPMA). Unsupervised clustering algorithms were used to evaluate expression patterns. Composite variables were constructed for DNA, RNA and protein positivity as well as for community and central IHC/FISH. Platforms were compared and Kaplan-Meier curves were constructed to compare outcomes by platform.

Results: 222 women were evaluable, though not all patients had results for all platforms. Community composite IHC/FISH was positive in 64/214 (30%) but only 41 of these (64%) were confirmed by central IHC/FISH and 4 additional cases were centrally positive despite negative community testing. Concordance was high among centrally-assessed Her2 platforms, but was lower between community IHC/protein and central RNA (90%), DNA (91%) and protein (91%). Among patients receiving trastuzumab (n=36), the pCR rate was ∼50% regardless of Her2-assessment platform; in contrast, those not receiving trastuzumab had pCR rates below 30%. Among the 64 patients deemed Her2+ by community IHC/FISH, 30 (48%) had pCR and 15 (25%) have had distant relapse. Five distant relapses have occurred despite pCR; all received trastuzumab, all were Her2 positive by multiple central platforms and 3/5 were ER-positive. Sites of distant relapse included brain, bone and viscera; only 1 of 5 had isolated brain relapse.

Conclusions: Community IHC/FISH testing for Her2 expression in the I-SPY Trial overcalled Her2 positivity compared to central testing while central results were highly concordant among DNA, RNA and protein platforms. Despite the high rate of community “false positives”, relapse after pCR occurred only in central Her2 “true positives,” exclusively among those receiving trastuzumab, and was rarely isolated to CNS sanctuary sites.