Lung cancer is often diagnosed by molecular markers for prediction and treatment. To date, the golden standard for detection of anaplastic lymphoma kinase (ALK) rearrangements is fluorescence in situ hybridization (FISH). We performed a pooled-data analysis on the diagnostic operating characteristics of immunohistochemistry (IHC) assay on non-small cell lung cancer (NSCLC). We searched Embase, Pubmed, and Springer databases. The results of IHC were evaluated using a modified H-score. We used a 2-level bivariate meta-analysis following a random effect model to summarize sensitivity and specificity and fit hierarchical summary receiver-operating characteristic curves. We also performed sensitivity analysis using different antibodies to investigate potential heterogeneity. Twelve studies consisting of a total of 3754 NSCLC specimens were analyzed. When we defined 1+/2+/3+, 2+/3+, and 3+ as ALK positive, we found the sensitivities to be 99% (95% confidence interval [CI], 97%-100%), 86% (95% CI, 73%-93%), and 56% (95% CI, 36%-74%) and the specificities to be 98% (95% CI, 95%-99%), 99% (95% CI, 99%-100%), and 100% (95% CI, 100%-100%), respectively. We demonstrated that when defining 3+ as positive and 0 as negative the sensitivity was 99% and specificity was 100%. In our sensitivity analysis, we found the sensitivity of D5F3 and 5A4 antibodies to be much higher than that of ALK1. We concluded that IHC scores 0 and 3+ were nearly 100% concordant with FISH-negative and FISH-positive status, respectively. However, IHC scores 1+ and 2+ might require further confirmatory testing by FISH assay. IHC assay using D5F3 and 5A4 antibodies reliably detected NSCLC with ALK rearrangement and may be useful as a screening method to identify these tumors.