26
IASLC ATLAS OF ALK TESTING IN LUNG CANCER
rare instances, most of the tumor cells in ALKpositive NSCLCs may exhibit only isolated 3' signals without normal B copies of ALK (Figure 13D). Cells with such a pattern are regarded as nonevaluable by conventional scoring rules A C D because of the possi- Figure 13. Atypical patterns seen in lung tumors with break-apart FISH. 13A: 3'-5-3' (redbility of a hybridization doublet pattern); 13B and 13C: 3'-3'-5'-3' (red-triplet pattern); 13D: Isolated 3’ signals, mostly without normal ALK signals). Arrows indicate ALK rearrangement. failure of the 5' probe, and ALK-positive cancers with such a pattern may be overlooked. Although it is not yet completely clear how consistently such atypical signal patterns predict fusion status, these patterns should at least raise a high index of suspicion and prompt testing with other diagnostic modalities (e.g., RT-PCR, IHC). Future studies may identify other atypical signal profiles that are associated with false-negative FISH results. Borderline Rates of Rearrangement-Positive Cells In approximately 8% of NSCLC cases, the rate of rearrangement-positive cells falls within the range of 10% to 20% (Camidge 2013). Although the currently accepted cutoff of 15% technically classifies such cases as either positive or negative for ALK rearrangement, our experience is limited as to whether such a borderline prevalence of rearrangement signals by FISH accurately represents presence of the fusion gene. We recommend that cells be carefully counted again on such specimens, with particular attention paid to the morphologic differentiation between tumor and nontumor cells. Including nontumor cells in the count dilutes the rate of rearrangement-positive cells. The recommended two-step scoring algorithm should minimize these technical errors. Similar attention should be paid to the vertically split signals along the z-axis of the tissue plane, which could be mistaken as a fused signal. This latter pitfall is particularly relevant in a laboratory in which the evaluation is performed on the captured digital images that consolidate multiple focus levels to produce one image (z-stacking). These borderline cases may also harbor atypical signal profiles, as described earlier. Specifically, a red-doublet pattern may initially stand out as a borderline rate of rearrangement-positive cells. Analysis using a single-color filter may facilitate the identification of closely apposed signals that may be overlooked by a dual-color (red and green) or a triple-color (blue, red, and green) filter. If the rate of rearrangement-positive cells is still borderline on careful reassessment, the final interpretation may be either positive or negative using the 15% cutoff, with the addition of a cautionary note to consider other diagnostic modalities, such as IHC or RT-PCR, for further workup. Need for Repeat Testing The findings of a few reports have suggested that the ALK status, including ALK copy number gain over the clinical course, can change, particularly after treatment with an ALK inhibitor (Doebele