The goal of personalized medicine is to treat patients with a therapy predicted to be efficacious based on the molecular characteristics of the tumor, thereby sparing the patient futile or toxic therapy. Anaplastic lymphoma kinase (ALK) inhibitors are effective against ALK-positive non–small-cell lung cancer (NSCLC) tumors, but to date the only approved companion diagnostic is a break-apart fluorescence in situ hybridization (FISH) assay. Immunohistochemistry (IHC) is a clinically applicable cost-effective test that is sensitive and specific for ALK protein expression. The purpose of this study was to assemble an international team of expert pathologists to evaluate a new automated standardized ALK IHC assay.
Archival NSCLC tumor specimens (n =103) previously tested for ALK rearrangement by FISH were provided by the international collaborators. These specimens were stained by IHC with the anti-ALK (D5F3) primary antibody combined with OptiView DAB IHC detection and OptiView amplification (Ventana Medical Systems, Inc., Tucson, AZ). Specimens were scored binarily as positive if strong granular cytoplasmic brown staining was present in tumor cells. IHC results were compared with the FISH results and interevaluator comparisons made.
Overall for the 100 evaluable cases the ALK IHC assay was highly sensitive (90%), specific (95%), and accurate relative (93%) to the ALK FISH results. Similar results were observed using a majority score. IHC negativity was scored by seven of seven and six of seven evaluators on three and two FISH-positive cases, respectively. IHC positivity was scored on two FISH-negative cases by seven of seven readers. There was agreement among seven of seven and six of seven readers on 88% and 96% of the cases before review, respectively, and after review there was agreement among seven of seven and six of seven on 95% and 97% of the cases, respectively.
On the basis of expert evaluation the ALK IHC test is sensitive, specific, and accurate, and a majority score of multiple readers does not improve these results over an individual reader’s score. Excellent inter-reader agreement was observed. These data support the algorithmic use of ALK IHC in the evaluation of NSCLC.