A Sensitive ALK Immunohistochemistry Companion Diagnostic Test ...

ORIGINAL ARTICLE

A Sensitive ALK Immunohistochemistry Companion Diagnostic Test Identifies Patients Eligible for Treatment with Crizotinib Trish Thorne-Nuzzo, BS,a,* Crystal Williams, MPH,a Alice Catallini, BS,a June Clements, MD,a Shalini Singh, MD,a James Amberson, MD,b Kim Dickinson, MD,b Zoran Gatalica, MD,c Steffan N. Ho, MD, PhD,d Isabell Loftin, PhD,a Abigail McElhinny, PhD,e Penny Towne, MBAa a

Ventana Medical Systems, Tucson, Arizona Laboratory Corporation of America Holdings, Shelton, Connecticut c Caris Life Sciences, Phoenix, Arizona d Pfizer Oncology, La Jolla, California e Personal Genome Diagnostics, Baltimore, Maryland b

Received 28 October 2016; revised 19 December 2016; accepted 6 January 2017 Available online - 29 January 2017

ABSTRACT Introduction: The availability of high-quality, rigorously validated diagnostic tests that can be broadly implemented is necessary to efficiently identify patients with anaplastic lymphoma kinase (ALK)-positive NSCLC who can potentially benefit from treatment with crizotinib. Here we present data on the recently approved Ventana ALK (D5F3) CDx Assay (Ventana Medical Systems, Tucson, AZ), the only immunohistochemistry (IHC)-based assay linked to treatment outcome. Methods: NSCLC specimens prospectively tested for anaplastic lymphoma receptor tyrosine kinase gene (ALK) status by flourescent in situ hybridization (FISH) in the PROFILE 1014 clinical trial of crizotinib versus chemotherapy (N ¼ 1018, including 179 ALK-positive and 754 ALK-negative specimens) were evaluated using the ALK (D5F3) CDx assay. Hazard ratios for progression-free survival comparing crizotinib and chemotherapy for ALK IHC–positive patients and ALK FISH–positive patients, as well as for concordance with the enrollment ALK FISH assay, were determined. Results: Results from both assays were obtained for 933 cases. Percent positive, negative, and overall agreement rates were 86.0% , 96.3%, and 94.3%, respectively. There were 53 discrepant cases, of which 25 were ALK FISH– positive/ALK IHC–negative and 28 were ALK FISH– negative/ALK IHC–positive. The hazard ratios using observed outcomes were 0.401 for ALK FISH–positive/ALK IHC–positive cases and 0.407 for all ALK FISH–positive cases tested with ALK IHC versus 0.454 for all ALK FISH– positive cases enrolled in the trial. Outcome data for ALK FISH–negative/ALK IHC–positive cases were not available for analysis. Between-reader agreement rates for ALK IHC involving three independent laboratories exceeded 98%.

Journal of Thoracic Oncology

Vol. 12 No. 5: 804-813

Conclusions: The ALK (D5F3) CDx assay is a stand-alone companion diagnostic test for identification of patients for treatment with crizotinib. This automated assay provides an effective option to accurately and rapidly identify patients with ALK-positive NSCLC. The simple binary scoring algorithm results in high reader-to-reader precision. 2017 International Association for the Study of Lung Cancer. Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license (http:// creativecommons.org/licenses/by-nc-nd/4.0/). Keywords: Anaplastic lymphoma kinase; NSCLC; Companion diagnostic assay; ALK IHC; Targeted therapy

Introduction NSCLC accounts for approximately 85% of all lung cancer cases in the United States1 and is by far the most common fatal cancer. In more than half of patients with *Corresponding author. Ms. Thorne-Nuzzo and Ms. Williams equally contributed to this work. Disclosure: Ms. Catallini, Dr. Clements, Dr. Dickinson, Dr. Loftin, Ms. Williams, Dr. Singh, Ms. Thorne-Nuzzo, and Ms. Towne are employees of Ventana Medical Systems, Inc. Dr. Ho is an employee of Pfizer, Inc. Dr. McElhinny is an employee of Personal Genome Diagnostics. Address for correspondence: Trish Thorne-Nuzzo, BS, Ventana Medical Systems, Inc., 1910 E. Innovation Drive, Tucson, AZ 85755. E-mail: [email protected] ª 2017 International Association for the Study of Lung Cancer. Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/ 4.0/). ISSN: 1556-0864 http://dx.doi.org/10.1016/j.jtho.2017.01.020

May 2017

newly diagnosed NSCLC the disease has already metastasized, greatly decreasing their likelihood of survival. The overall 5-year survival rate for advanced NSCLC is 1% to 4%, depending on histologic subtype.2 Because of inadequate screening methods, most patients with NSCLC present with inoperable, locally advaned disease or metastatic disease, neither of which currently has any curative treatment options.3 The anaplastic lymphoma kinase (ALK) protein is a member of the insulin receptor superfamily of receptor tyrosine kinases and is not normally expressed in lung. However, anaplastic lymphoma receptor tyrosine kinase gene (ALK) rearrangements have been described in NSCLC, and their expression offers a molecularly defined subgroup of patients with lung cancer who may benefit from ALK-targeted therapy relative to standard first-line chemotherapy.4 In a series of studies utilizing archived NSCLC specimens from patients with a broad range of ethnic backgrounds, the incidence of ALK gene rearrangements ranged from 2% to 7%.4–9 A meta-analysis of these results indicated that the prevalence of ALK gene rearrangements in lung cancer is approximately 6000 patients per year in the United States and up to 40,000 patients per year worldwide.10 The clinical significance of ALK gene rearrangements was demonstrated in clinical studies that examined the efficacy and toxicity of crizotinib (Xalkori [Pfizer, New York, NY]), a receptor tyrosine kinase inhibitor for the treatment of ALK-positive NSCLC.11 The aggregate overall, partial, and complete response rates of these trials were 61.2%, 59.8%, and 1.5%, respectively.12 A recent study in patients with ALK-positive NSCLC demonstrated significantly longer progression-free survival (PFS) (median 10.9 versus 7.0 months) and higher objective response rates (ORRs) (74% versus 45%) with crizotinib treatment compared with standard of care first-line doublet chemotherapy.12 Eligibility for treatment with crizotinib is dependent on the use of an accurate and reproducible companion diagnostic test to reliably identify patients with ALK-positive NSCLC. Fluorescence in situ hybridization (FISH) to detect ALK gene rearrangements was used as the primary clinical trial assay for enrollment into crizotinib clinical trials and therefore represented the first method clinically validated to identify patients eligible for treatment with crizotinib. FISH is technically challenging and requires specialized expertise to execute and evaluate the staining results. For example, the Vysis FISH assay (Vysis, Downers Grove, IL) requires that a minimum of 50 cells be counted and positive tumor cell percentages between 10% and 50% be evaluated by a second reader.13 Immunohistochemical (IHC) methods can detect ALK protein independent of the underlying mechanism mediating expression and therefore provide

ALK IHC Companion Diagnostic

805

an alternative to FISH for determining ALK status in formalin-fixed paraffin embedded (FFPE) tissue specimens. Several IHC methods for the detection of ALK in NSCLC have been published,14–19 but the reported sensitivity and specificity of these assays relative to FISH were variable, depending on the antibody, specific assay conditions, scoring methodology, and/or detection systems used.14,19 Importantly, none of these studies utilized tumor tissue samples from clinical trials that established the clinical benefit of crizotinib and therefore did not associate the results of the ALK IHC testing with patient outcome data after crizotinib treatment. A new, fully automated, U.S. Food and Drug Administration–approved ALK IHC assay, the Ventana ALK (D5F3) CDx assay (Ventana Medical Systems, Inc., Tucson, AZ), was developed to provide high analytical sensitivity and specificity, a high degree of concordance with FISH, and very high interreader reproducibility resulting from the use of a binary scoring algorithm. Clinical validation of the assay was performed by ultilizing samples from a phase 3 clinical trial. The D5F3 antibody detects the kinase (active) region of ALK expression.20 The assay uses a highly sensitive detection system to amplify the ALK-specific signal, thus providing the sensitivity necessary to detect low levels of ALK expression. The absence of endogenous ALK protein expression in NSCLC that has not undergone oncogenic rearrangement of the ALK gene enables verification of very strong signal amplification with assay conditions that maintain specificity. This system further allows evaluation of the staining results by using a simple binary scoring algorithm, according to which the presence of strong granular staining in any percentage of tumor cells is considered positive for ALK. The studies reported here were conducted to determine the clinical utility of the ALK (D5F3) CDx assay as a stand-alone companion diagnostic test for the identification of patients with NSCLC who might benefit from treatment with crizotinib, and they include results from analysis of clinical outcome data (PFS hazard ratios [HRs]), concordance with FISH, and interlaboratory reproducibility data.

Materials and Methods Ventana ALK (D5F3) CDx Assay Development The ALK (D5F3) CDx assay uses a rabbit monoclonal primary antibody that binds to ALK in FFPE tissue sections. The assay was developed as a system with the ALK (D5F3) CDx antibody in combination with the OptiView DAB IHC detection and OptiView Amplification (AMP) kits and rabbit monoclonal negative control immunoglobulin for use on a Ventana BenchMark XT automated staining instrument. The

806 Thorne-Nuzzo et al

amplification kit involves addition of the OptiView Amplifier (3-hydroxy-2-quinoxaline-haptenated tyramide), OptiView Amplification H2O2 (the substrate), and the OptiView Amplification Multimer (an anti3-hydroxy-2-quinoxaline horseradish peroxidase–conjugated monoclonal antibody), which builds on the scaffolding generated from the OptiView DAB IHC Detection Kit.21 During assay development, extensive feasibility testing was conducted on 139 NSCLC specimens (the feasibility case set) representing a range of ALK-positive and ALK-negative samples (79 ALK FISH–positive and 60 ALK FISH–negative) to optimize sensitivity and specificity of the assay, enabling detection of low ALK expressors with high concordance with FISH. All samples were previously characterized by FISH, IHC, and/or reverse-transcriptase polymerase chain reaction. Initial challenges in determining ALK status on focally/weakly ALK-positive NSCLC tumors led to addition of the OptiView Amplification kit in conjunction with the OptiView DAB Detection kit to facilitate assessment of ALK status in focally positive NSCLC specimens. Use of the OptiView Amplification kit also enabled development of a binary scoring algorithm in which any percentage of strong granular cytoplasmic staining in tumor cells could be defined as ALK positive. Clinical Specimens (1014 Sample Set 1). Additionally, 305 ALK FISH–characterized cases sourced from the clinical study PROFILE 101412 (1014 sample set 1) were stained with the ALK (D5F3) CDx assay during its early development to ensure that the associated ALK (D5F3) CDx assay scoring algorithm for NSCLC correlated with the established ALK FISH status, as determined by Vysis ALK Break Apart FISH Probe Kit (Vysis ALK FISH assay [Abbott Molecular Inc., Des Plaines, IL]).13

Interlaboratory Reproducibility Study Multiple tissue sections were cut from 12 FFPE NSCLC specimens (six ALK IHC–positive and six ALK IHC–negative), representing the dynamic range of the intended-use population. The slides were randomized and distributed to three clinical sites (LabCorp, Shelton CT; Caris Life Sciences, Phoenix, AZ; and Integrated Oncology, Irvine, CA) for staining with the ALK (D5F3) CDx assay. The pathologists were blinded to all clinical information about the cases. Each site stained slides from the 12 cases on five nonconsecutive days (over a minimum of 20 days). Thus, 60 slides were stained at each site and two pathologists at each site evaluated the slides independently. All 60 case slides were evaluable by both readers at the three study sites, for a total of 360 observations with evaluable results.


Vol. 12 No. 5

ALK IHC Clinical Utility Study Clinical Specimens (1014 Sample Set 2). FFPE NSCLC tissue specimens for the ALK IHC clinical utility study were prospectively collected under informed consent from patients with locally advanced or metastatic NSCLC who were being considered for enrollment into the Pfizer clinical study PROFILE 1014 (NCT01154140), which evaluated the efficacy and safety of crizotinib compared with standard-of-care first-line doublet chemotherapy.12 As part of the clinical study, tissue specimens were submitted to one of four central laboratories and screened for ALK status by using an investigational use–only form of the Vysis ALK FISH assay.13 Only patients with positive ALK status by FISH (ALK FISH–positive) were eligible for enrollment into the clinical study. Tissue Staining and Evaluation Procedures. Unstained tissue sections from all available specimens screened for the clinical study between approximately July 2012 and the end of enrollment in June 2013 (including ALK FISH–negative and ALK FISH–uninformative specimens) were transferred to an independent central laboratory (LabCorp), which served as the sole investigative site for the clinical utility study. Eligible specimens (those with sufficient tumor for evaluation as determined by hematoxylin and eosin staining) were enrolled and evaluated for ALK status according to the ALK (D5F3) CDx assay. Sufficiency of tumor specimen was left to the discretion of the investigators. All specimens were deidentified and unlinked from patient information before transfer to LabCorp. LabCorp investigators had no access to FISH data from the clinical study and were blinded to all clinical information about the cases. All laboratories (for both the ALK IHC clinical utility and reproducibility studies) performed the staining procedure according to the manufacturer’s instructions and as described elsewhere.20,21 System run controls were included with each staining run. For both the clinical utility and reproducibility studies, case slides were assigned to staining runs on a Ventana BenchMark XT IHC/In Situ Hybridization automated staining instrument according to a randomization scheme. For the reproducibility study, a different random order of cases was used for each staining run at each site. Stained slides were evaluated by trained investigatorpathologists (readers) at each study site according to the assay scoring algorithm.20 ALK IHC–positive status was denoted by the presence of strong granular cytoplasmic staining in any percentage of positive tumor cells. ALK IHC–negative status was denoted by the absence of strong granular cytoplasmic staining in tumor cells.

May 2017

Statistical Methods Reproducibility Study. Frequencies and percentages were calculated for the agreement of ALK status (positive or negative). Average positive agreement, average negative agreement rates, and overall percent agreement (OPA) were calculated. ALK IHC Clinical Utility Study. The primary end points for the ALK IHC clinical utility study (sample set 2) were the positive percent agreement (PPA) and negative percent agreement (NPA) rates for ALK status between the ALK FISH and ALK (D5F3) CDx assays. The PPA, NPA, and OPA rates were calculated by comparing ALK status from the two assays for the same specimens using the ALK FISH status as the reference and standard formulas22 with corresponding two-sided 95% confidence intervals. The secondary end point of the ALK IHC clinical utility study was the therapeutic benefit of the ALK (D5F3) CDx assay for determining suitability for crizotinib treatment. Sample sets 1 and 2 were used for the secondary end point analysis. The HR for patients who were ALK FISH–positive and ALK IHC–positive was calculated between treatment arms by using the Cox proportional hazards model (racial group, brain metastasis, and Eastern Cooperative Oncology Group performance status score were used as strata [n ¼ 141]). This method was also used to calculate the HR for all ALK FISH–positive patients enrolled in the clinical study (n ¼ 343) and the HR for cases of enrolled patients who were ALK FISH–positive and screened restrospectively with the ALK (D5F3) CDx assay (n ¼ 171). Additional analyses that included ALK FISH–negative or ALK FISH–uninformative/ALK IHC– positive patients were conducted by using simulated outcomes (because ALK FISH–negative or ALK FISH– uninformative patients were not enrolled in the crizotinib clinical trial). Outcomes were assigned to ALK FISH–negtive or ALK FISH–uninformative/ALK IHC– positive patients by resampling with replacement from the existing ALK FISH–positive/ALK IHC–positive outcome data. The proportion of outcomes taken from each treatment arm were varied such that best- and worst-case scenarios were assessed. The best-case scenario assumed that ALK FISH–negative or ALK FISH–uninformative/ALK IHC–positive patients responded to crizotinib to the same degree that ALK FISH–positive/ALK IHC–positive patients responded to crizotinib. The worst-case scenario assumed that ALK FISH–negative or ALK FISH–uninformative /ALK IHC– positive patients responded to crizotinib to the same degree that ALK FISH–positive/ALK IHC–positive patients responded to chemotherapy. For each scenario,


807

1000 HRs were calculated (one for each of 1000 bootstrap samples). The ORRs (defined as patients in the clinical study with a complete or partial response to treatment by the Response Evaluation Criteria in Solid Tumors, version, 1.123) for crizotinib and chemotherapy were calculated for patients enrolled into the study as follows: ALK FISH– positive cases that were also tested retrospectively with ALK (D5F3) CDx assay, ALK FISH–positive/ALK IHC–positive cases, and ALK FISH–positive/ALK IHC– negative cases. Exploratory end points for the ALK IHC clinical utility study included the initial and final staining acceptability rates for the ALK (D5F3) CDx and FISH assays and were determined for all cases that were stained with both assays. The first-pass and final evaluable rates (i.e., the proportion of specimens providing a valid ALK status on initial testing and final testing) were calculated for the FISH and IHC assays.

Results Development of the ALK (D5F3) CDx Assay and Initial Concordance with FISH The ALK (D5F3) CDx assay was developed by using NSCLC cases that were characterized for ALK status determined with the Vysis ALK FISH assay. Seventy-nine ALK FISH–positive cases and 60 ALK FISH–negative cases were obtained and stained with the D5F3 antibody and the OptiView detection system under various conditions to optimize assay performance and maximize concordance with the FISH results. The ALK antigen was readily detected in approximately 80% of the ALK FISH– positive cases by standard DAB systems. Most cases were homogeneously positive across the entire tumor area (Supplementary Figure 1B). ALK staining was not present in the stroma adjacent to or within the tumor area, providing an internal negative control in each specimen. Strikingly, in 15% to 20% of the ALK FISH-positive cases, ALK expression was heterogeneous with focally positive staining, often staining very weakly in most of the tumor area (Supplementary Fig. 1A). These “focally positive” cases often exhibited ALK IHC staining intensity that was near the limit of detection by standard DAB detection. To reliably diagnose all ALK-positive NSCLC specimens, an additional amplification step (use of the OptiView AMP kit) was introduced into the IHC staining protocol. This tyramide-based amplification system resulted in a significant increase in staining intensity, allowing easy detection of previously weak or barely detectable staining with minimal increase in background staining that did not affect evaluation of the slides (see Supplementary Fig. 1A).



Multiple pathologists evaluated the FISHcharacterized feasibility case set that was stained by the ALK (D5F3) CDx assay, and a simple binary scoring algorithm that correlated ALK IHC staining with the predefined ALK FISH status was developed. ALK IHC– positive status was denoted by the presence of strong granular cytoplasmic staining in any percentage of positive tumor cells. ALK IHC–negative status was denoted by the absence of strong granular cytoplasmic staining in tumor cells. When this algorithm was used, concordance with FISH was 100% for both PPA and NPA in this initial sample set of 139 cases (Supplementary Fig. 1B). This assay and algorithm, which are described in detail elsewhere,20 were used for the ALK IHC clinical utility and reproducibility studies described here.

Reproducibility Study The ALK (D5F3) CDx assay was highly reproducible across study sites, days, and readers. The between-site average positive agreement, average negative agreement, and OPA rates all exceeded 93% and the betweenday and between-reader rates all exceeded 98% (Table 1).

ALK IHC Clinical Utility Study Patient/Specimen Disposition and Demographics. Of the 2217 patients screened with ALK FISH for the PROFILE 1014 clinical study (1014 sample set 2), 1018 were available for testing with the ALK (D5F3) CDx assay; 933 of these cases had valid ALK FISH and ALK (D5F3) IHC results and were included in the assessment of agreement rates (Supplementary Figure 2). The PROFILE 1014 clinical study was initiated by using the ALK FISH assay for enrollment before initiation of the method comparison study. This led to a subset of the patient cases enrolled into the trial being available for staining in the method comparison study. All available patient cases were prospectively stained with the ALK (D5F3) CDx assay upon formulation lock. Patient characteristics between those with samples stained with the ALK (D5F3) CDx assay and those with samples not Table 1. Interlaboratory Reproducibility of ALK (D5F3) IHC Assay Variable Compared

APA

ANA

OPA

Between sites: % 95% CI Between days: % 95% CI Between readers: % 95% CI

93.8 76.2–100 99.1 96.4–100 98.8 95.2–100

94.9 79.2–100 99.2 96.9–100 99.0 95.8-100

94.4 83.3–100 99.2 97.5–100 98.9 96.7–100

ALK, anaplastic lymphoma kinase; IHC, immunohistochemistry; APA, average positive agreement; ANA, average negative agreement; OPA, overall percent agreement; CI, confidence interval.

Vol. 12 No. 5

stained were not appreciably different, suggesting that the subset of samples stained with the ALK (D5F3) CDx assay is representative of the total trial population. The mean age of patients enrolled in the ALK IHC clinical utility study was 62.1 years, with similar proportions of male and female patients. Approximately 77% of patients were ALK FISH–negative and 18% were ALK FISH–positive (Supplementary Table 1). Concordance between ALK FISH and ALK (D5F3) CDx Assays. The results obtained with the ALK (D5F3) CDx assay were highly concordant with those obtained by ALK FISH assay, with PPA, NPA, and OPA rates of 86.0%, 96.3%, and 94.3%, respectively (Table 2). Diagnostic Clinical Utility. For observed outcomes, the point estimate for HR comparing PFS by treatment arm was lower for ALK FISH–positive/ALK IHC–positive patients (0.401) than for all ALK FISH–positive patients (0.454). The HR for all ALK FISH–positive patients who were tested with the ALK (D5F3) CDx assay (1014 sample sets 1 and 2), including patients who were ALK FISH–positive/ALK IHC–negative, was 0.407 (Table 3). Table 4 provides a summary of the analysis conducted with simulated outcome data. The mean, SD, 2.5th and 97.5th percentiles of the 1000 HRs are presented. The response level scenario column corresponds with the best- to worst-case scenarios, with 0% representing the worst-case scenario. The results demonstrate that even under the worst-case scenario, which is conservative, the HR for the ALK IHC–positive cases in sample sets 1 and 2 (0.492) did not vary significantly from that for the ALK FISH–positive/ALK IHC–positive cases (0.401). The ORR was statistically significantly greater for ALK FISH–positive/ALK IHC–positive cases than for ALK FISH–positive/ALK IHC–negative cases for patients treated with crizotinib (86.7% versus 33.3% [p ¼ 0.0083), whereas for those treated with chemotherapy the difference between the ALK FISH–positive/ ALK IHC–positive cases and the ALK FISH–positive/ALK IHC–negative cases was smaller (44.9% versus 37.5%). The ORR for ALK FISH–positive/ALK IHC–negative cases treated with crizotinib was comparable to the ORR for the chemotherapy arm (33.3% versus 37.5%), yet sample size in this category was small (Supplementary Table 2).

Exploratory End Points Staining acceptability was assessed for both the ALK FISH and ALK (D5F3) CDx assays. If the ALK FISH assessment resulted in an actionable result (positive or negative), the testing attempt was considered acceptable. Not evaluable status was due to a variety of reasons such

May 2017


809

Table 2. Concordance between ALK FISH and ALK IHC Assays for PROFILE 1014 Sample Set 2 ALK FISH Status

Agreement Rates

ALK IHC Status

Positive

Negative

Total

Rate

n of N

%

95% CI

Positive Negative Total

154 25 179

28 726 754

182 751 933

PPA NPA OPA

154 of 179 726 of 754 880 of 933

86.0 96.3 94.3

80.2–90.4 94.7–97.4 92.6–95.6

ALK, anaplastic lymphoma receptor tyrosine kinase gene; FISH, fluorescence in situ hybridization; ALK, anaplastic lymphoma kinase; IHC, immunochemistry; CI, confidence interval; PPA, positive percent agreement; NPA, negative percent agreement; OPA, overall percent agreement.

as no FISH signals, not enough evaluable cells, or tissue that fell off the slide. The ALK (D5F3) CDx assay assessment was considered acceptable if the run controls, negative reagent control, and antibody slide were considered acceptable and an actionable result of positive or negative was recorded. Reasons for not evaluable status include not enough tissue or tumor and inappropriate run controls. Assessment of the acceptability of initial (first-pass) staining showed that the proportion of cases with a valid IHC result (926 of 1018 [91.0%]) was higher than the proportion of cases with a valid FISH result (876 of 1018 [86.1%]). The final staining acceptability rates after repeat testing were 95.8% (975 of 1018) and 94.9% (966 of 1018) for IHC and FISH, respectively. Of the 52 cases that were uninformative on final FISH testing, 75.0% (39 of 52) were ALK IHC– negative and 5.8% (three of 52) were ALK IHC–positive; the remaining 10 cases were not evaluable or had missing ALK IHC data.

Discordant Cases between FISH and IHC in the ALK IHC Clinical Utility Study Fifty-three patients had discordant ALK results between the two assays (25 ALK FISH–positive/ALK IHC–negative and 28 ALK FISH–negative/ALK IHC–positive). Mean and median FISH scores were significantly lower for the ALK FISH–positive/ALK IHC–negative discordant population (31.8 and 22.0%, respectively)

than for the ALK FISH–positive population (57.0% and 58.0%, respectively) (Supplementary Table 3). Fourteen of the 25 ALK FISH–positive/ALK IHC–negative cases had FISH scores of 25% or less. The disagreement between assays appeared to be greatest for cases closest to the FISH diagnostic cutoff (15% of tumor cells exhibiting the translocation). When a subset of cases with FISH scores near the FISH diagnostic cutoff (10%– 50%) were analyzed, the concordance between the assays decreased with lower PPA rates than for the overall population (Supplementary Table 4). Conversely, when cases with FISH scores in this translocation equivocal zone were excluded, the concordance increased with higher PPA rates than for the overall population (Supplementary Table 4). Fifteen of the 25 ALK FISH–positive/ALK IHC–negative patients were enrolled in the PROFILE 1014 study; six of these 15 patients received crizotinib and nine received chemotherapy. Of the six patients who received crizotinib, the two with the best overall response had FISH scores of 67% and 72% and experienced a partial response to treatment, two had FISH scores of 15% and 17% with disease progression, and two had FISH scores of 15% and 18% with stable disease (Fig. 1). No clinical outcome data were available on the 28 cases that were ALK FISH–negative/ALK IHC–positive because the PROFILE 1014 study enrolled only patients who were ALK-positive by FISH. However, to gain additional insight on discordance between ALK IHC and

Table 3. Hazard Ratios of Crizotinib versus Chemotherapy for PROFILE 1014 Sample Sets 1 and 2 Sample Size ALK Status

Hazard Ratioa

SE

95% CI

Chemotherapy Arm

Crizontinib Arm

0.454

0.139

0.346–0.596

171

172

0.407 0.401 1.711

0.214 0.237 0.703

0.267–0.618 0.252–0.639 0.431–6.789

82 63 17

90 78 8

Total enrolled FISH-positive ALK IHC–tested FISH-positiveb FISH-positive/IHC-positive FISH-positive/IHC-negative a

Hazard ratio of crizotinib versus chemotherapy. Results estimated by using a stratified Cox model with the following strata: race, brain metastases, and Eastern Cooperative Oncology Group performance status. b For two ALK FISH-positive patients in the chemotherapy arm and four patients in the crizotinib arm, no positive or negative ALK IHC status was obtained. ALK, anaplastic lymphoma kinase; CI, confidence interval; FISH, fluorescence in situ hybridization; IHC, immunohistochemistry; ALK, anaplastic lymphoma receptor tyrosine kinase gene.



Table 4. HR Estimates Using Simulated Outcomes for PROFILE 1014 Sample Sets 1 and 2 HR Statistics for Simulated ALK IHC– Positive Trialb Response Level Scenarioa

Mean ± SD

100% 75% 50% 25% 0%

0.412 0.433 0.452 0.471 0.492

± ± ± ± ±

2.5th and 97.5th Percentile

0.041 0.043 0.043 0.042 0.041

0.337 0.357 0.373 0.393 0.415

and and and and and

0.503 0.523 0.540 0.563 0.578

a Percentage of ALK FISH–negative or ALK FISH–uninformative/ALK IHC– positive patients randomly assigned to the crizotinib treatment arm of the simulated ALK IHC–positive trial who were assumed to exhibit the same response as the crizotinib-treated ALK FISH–positive/ALK IHC–positive patients. b HR statistics based on 1000 simulated ALK IHC–positive trials per scenario. In each simulated trial, the HR for crizotinib versus chemotherapy was estimated by using a stratified Cox regression model with race, brain metastases, and Eastern Cooperative Oncology Group performance status as strata. HR, hazard ratio; ALK, anaplastic lymphoma kinase; IHC, immunohistochemistry; ALK, anaplastic lymphoma receptor tyrosine kinase gene; FISH, fluorescence in situ hybridization.

FISH, a post hoc review of ALK (D5F3) CDx assay– stained slides from the 28 ALK FISH–negative/ALK IHC–positive cases was conducted by an expert reader to evaluate the positive ALK IHC staining patterns in ALK FISH–negative patient specimens. Representative ALK (D5F3) IHC staining patterns of ALK IHC–positive/ALK FISH–negative cases are shown in Figure 2. Strong homogenous (90% of cells positive) ALK IHC–positive staining was observed in 14 cases (Fig. 2A) and strong but heterogeneous ALK IHC–positive staining was

Vol. 12 No. 5

present in 12 cases (Fig. 2B and C); two cases were unevaluable in post hoc review. Thus, 26 of 28 of these cases (92.9%) were unequivocally ALK IHC–positive with readily detectable strong staining.

Discussion Validated IHC testing is an appropriate and reliable alternative to FISH for the identification of ALK-positive NSCLC.13–15,24 The Ventana ALK (D5F3) CDx Assay was specifically developed to maximize agreement with ALK FISH testing to enable use of ALK IHC as a stand-alone companion diagnostic to identify patients with ALKpositive NSCLC who are eligible for treatment with crizotinib. The assay detection technology is designed to amplify the DAB signal to ensure detection of cases that may have low expression of ALK rearrangements. This amplification technology enables a binary scoring algorithm and is thus less subjective than other IHC assays and highly concordant with FISH. This was demonstrated by the very high interreader precision rates observed in the reproducibility study reported here. The ALK (D5F3) CDx assay demonstrated equivalent performance to FISH for identifying ALK rearrangement in patients with NSCLC. In the ALK IHC clinical utility study, the ALK (D5F3) CDx assay showed excellent positive, negative, and overall agreement rates with the FISH assay (86.0%, 96.3%, and 94.3%, respectively). These high negative and positive agreement rates are in accordance with specificity and sensitivity rates reported elsewhere for the D5F3 antibody when used with the OptiView Detection and Amplification Kit (specificity 95%–100%, sensitivity 86%–100%).14–19 The ALK Study 1014 Discordant Cases: 53

ChemotherapyTreated Cases: 9

FISH score 15-18% Cases: 4

YES

ALK FISH+/ALK IHCCases: 25

ALK FISH-/ALK IHC+ Cases: 28

Crizotinib-Treated Cases: 6

Post-hoc Internal Review of ALK IHC Slide Images

“Equivocal” Fish Score?

Unevaluable: 2 cases

Strong ALK IHC Staining in ≥ 90% of Tumor Cells: 14 cases

Heterogeneous ALK-Positive Staining: 12 cases

NO Stable Disease Cases: 2

Disease Progression Cases: 2

FISH score 66-72%: 2 cases

Partial Response: 2 cases

Figure 1. Outcome data for discordant cases. Abbreviations: ALK, anaplastic lymphoma kinase; FISH, fluorescence in situ hybridization; IHC, immunohistochemistry.

May 2017


A

811

B

ALKIHC+/FISH-

ALKIHC+/FISH-

ALKIHC+/FISHNegave IHC Region

ALKIHC+/FISHPosive IHC Region

C

Figure 2. Patterns of anaplastic lymphoma kinase (ALK) immunohistochemistry (IHC) positivity in ALK IHC–positive/fluorescence in situ hybridization (FISH)–negative discordant cases. (A) ALK IHC–positive/ALK FISH-negative case displaying strong ALK IHC staining in more than 90% of tumor cells. (B) ALK IHC–positive/ALK FISH–negative case displaying heterogeneous ALK IHC staining intensity. Note diffusely distributed ALK-positive and ALK-negative tumor cells. (C) ALK IHC–positive/ALK FISH– negative case displaying heterogeneous ALK IHC staining. Note the distinct tumor areas of ALK-positive staining (right image) and ALK-negative staining (left image); areas of positivity may have been missed by FISH.

(D5F3) CDx assay showed excellent initial and final staining acceptability rates and a higher first-pass rate than the ALK FISH assay. The higher first-pass rate reduces the need for repeat testing, with consequent savings in time and costs and ability to provide the ALK status results to the physician faster, enabling quicker treatment decisions for the patient. The clinical utility of the ALK (D5F3) CDx assay was evaluated by using HRs for PFS and ORRs for patients enrolled in the clinical trial. When observed therapeutic outcomes were used, the HR for PFS between crizotiniband chemotherapy-treated patients was lower for patients who were ALK FISH–positive/ALK IHC–positive

than for all patients who were ALK FISH–positive, demonstrating that the ALK (D5F3) CDx assay is suitable as a stand-alone test to identify patients for crizotinib therapy. These results were supported and confirmed by additional analyses using simulated outcomes for ALK FISH–negative or ALK FISH–uninformative patients who were ALK IHC–positive. With the inclusion of these patients, the HR between the two treatment arms was comparable to the HR reported for all ALK FISH–screened patients who were also screened with the ALK (D5F3) CDx assay. Additionally, the ORR for patients treated with crizotinib was notably higher for ALK FISH–positive/ALK IHC–positive patients than for ALK FISH–positive/ALK


IHC–negative patients; the ORR for the latter group of ALK FISH–positive/ALK IHC–negative patients was similar to that of patients treated with chemotherapy, suggesting possible false-positive FISH results for some cases. Outcome data were not available for ALK FISH–negative/ ALK FISH–positive patients. To address this limitation in the design of the study, an analysis with simulated outcome data was conducted (see Table 4). Fifty-three cases had discordant ALK results for the two assays. Most of the ALK FISH–positive/ALK IHC– negative discordances appeared to be for cases that were close to the FISH diagnostic cutoff, with concordance decreasing when patients in the FISH equivocal zone (10% to 50% positive cells on evaluation of 50 cells) were compared and increasing when patients in the equivocal zone were excluded. On the basis of available outcome data for the ALK FISH–positive/ALK IHC– negative cases, it appears that some false-positive FISH results may have contributed to the ALK FISH–positive/ ALK IHC–negative discordances. Similarly, on the basis of post hoc review of ALK (D5F3) IHC–stained slides, errors in ALK FISH staining or interpretation of slides may have contributed to the ALK FISH–negative/ALK IHC–positive discordances. FISH interpretation can be challenging because of intrachromosomal rearrangements and technical variability in probe signal and separation, which can yield subtle signal splitting, leading to potential false negatives.25 In this study 50% of the ALK FISH–negative/ ALK IHC–positive cases demonstrated heterogeneity. This heterogeneity may have contributed to the discordances, as the binary IHC scoring algorithm and ability to evaluate the whole tissue section may have enabled the reader to identify more positive cases when using ALK (D5F3) IHC versus when using FISH, which often is performed on only a representative area of the neoplasm. There is evidence that some ALK FISH–negative/ALK IHC–positive cases do respond to crizotinib; therefore, the discordant ALK FISH–negative/ALK IHC–positive cases observed in this study may not necessarily represent IHC false positives and could be FISH false negatives.26 The strong, homogenous staining observed in most of these cases, which is typical of ALK FISH– positive/ALK IHC–positive cases, is consistent with true ALK-positive status of the tumor. Although such discordant cases are infrequent, additional clinical data will be needed to determine which method best correlates with patient outcome. The ALK (D5F3) IHC assay is currently being used in additional clinical trials for ALK inhibitors (e.g., NCT01828099 and NCT02075840) to select patients for treatment; data from these studies will further evaluate the utility of ALK (D5F3) assay.27,28 In conclusion, the ALK (D5F3) CDx assay is equivalent to FISH for identifying ALK status in patients with NSCLC. The assay is highly reproducible and suitable for


Vol. 12 No. 5

use as a stand-alone test to identify patients with NSCLC who may benefit from crizotinib treatment. ALK IHC testing using the ALK (D5F3) CDx assay is a rapid and relatively inexpensive alternative method to identify ALK-positive NSCLC; it is also less labor-intensive, does not require specialized training, and unlike FISH, does not require a minimum number of tumor cells present in the sample. The U.S. Food and Drug Administration– approved ALK (D5F3) CDx assay is a highly sensitive and specific automated and standardized ALK IHC assay that has been clinically validated by using patient samples from PROFILE 1014 and can accurately and rapidly identify patients with ALK-positive NSCLC. The detection chemistry used in the ALK (D5F3) CDx assay allows for a simple binary scoring algorithm that results in high reader-to-reader precision.

Acknowledgments This study was supported by Roche Tissue Diagnostic (Ventana Medical Systems) and Pfizer Oncology (La Jolla, CA). Dr. Jacqueline Wu of Castle Peak Medical Writing provided medical writing assistance for this manuscript with funding provided by Ventana Medical Systems. Dr. Hiro Nitta assisted with sample procurement and scientific input during determination of assay feasibility. The contributions of the PROFILE 1014 patients, as well as that of the study team and clinical investigators, are gratefully acknowledged.

Supplementary Data Note: To access the supplementary material accompanying this article, visit the online version of the Journal of Thoracic Oncology at www.jto.org and at http://dx.doi. org/10.1016/j.jtho.2017.01.020.

References 1. Howlander N, Noone A, Krapcho M. SEER cancer statistics review. http://seer.cancer.gov/csr/1975_2010. Accessed September 10, 2013. 2. Cetin K, Ettinger DS, Hei YJ, O’Malley CD. Survival by histologic subtype in stage IV nonsmall cell lung cancer based on data from the Surveillance, Epidemiology and End Results Program. Clin Epidemiol. 2011;3:139–148. 3. Jiang T, Changyun Z, Su C, Ren S, Zhou C. The diagnostic value of circulating cell free DNA quantification in nonsmall cell lung cancer: a systematic review with metaanalysis. Lung Cancer. 2016;100:63–70. 4. Soda M, Choi YL, Enomoto M, et al. Identification of the transforming EML4-ALK fusion gene in non-small-cell lung cancer. Nature. 2007;448:561–566. 5. Choi YL, Takeuchi K, Soda M, et al. Identification of novel isoforms of the EML4-ALK transforming gene in non-small cell lung cancer. Cancer Res. 2008;68:4971–4976. 6. Inamura K, Togashi Y, Ninomiya H, et al. EML4-ALK fusion is linked to histological characteristics in a subset of lung cancers. J Thorac Oncol. 2008;3:13–17.

May 2017 7. Koivunen JP, Mermel C, Zejnullahu K, et al. EML4-ALK fusion gene and efficacy of an ALK kinase inhibitor in lung cancer. Clin Cancer Res. 2008;14:4275–4283. 8. Shinmura K, Kageyama S, Tao H, et al. EML4-ALK fusion transcripts, but no NPM-, TPM3-, CLTC-, ATIC-, or TFGALK fusion transcripts, in non-small cell lung carcinomas. Lung Cancer. 2008;61:163–169. 9. Takeuchi K, Choi YL, Soda M, et al. Multiplex reverse transcription-PCR screening for EML4-ALK fusion transcripts. Clin Cancer Res. 2008;14:6618–6624. 10. Pfizer, Inc. Crizotinib briefing information for the November 30, 2010 meeting of the Pediatric Oncology Subcommittee of the FDA Oncologic Drugs Advisory Committee. http://www.fda.gov/AdvisoryCommittees/ CommitteesMeetingMaterials/Drugs/OncologicDrugsAdvi soryCommittee/ucm234831.htm. Accessed May 18, 2015. 11. Qian H, Gao F, Wang H, Ma F. The efficacy and safety of crizotinib in the treatment of anaplastic lymphoma kinase-positive non-small cell lung cancer: a metaanalysis of clinical trials. BMC Cancer. 2014;14:683. 12. Solomon BJ, Mok T, Kim DW, et al. First-line crizotinib versus chemotherapy in ALK-positive lung cancer. N Engl J Med. 2014;371:2167–2177. 13. Vysis ALK. Break Apart FISH Probe Kit [product information]. Des Plaines, IL: Abbott Molecular, Inc.; 2011. 14. Hutarew G, Hauser-Kronberger C, Strasser F, Llenos IC, Dietze O. Immunohistochemistry as a screening tool for ALK rearrangement in NSCLC: evaluation of five different ALK antibody clones and ALK FISH. Histopathology. 2014;65:398–407. 15. Ying J, Guo L, Qiu T, et al. Diagnostic value of a novel fully automated immunochemistry assay for detection of ALK rearrangement in primary lung adenocarcinoma. Ann Oncol. 2013;24:2589–2593. 16. Wynes MW, Sholl LM, Dietel M, et al. An international interpretation study using the ALK IHC antibody D5F3 and a sensitive detection kit demonstrates high concordance between ALK IHC and ALK FISH and between evaluators. J Thorac Oncol. 2014;9:631–638. 17. Minca EC, Portier BP, Wang Z, et al. ALK status testing in non-small cell lung carcinoma: correlation between ultrasensitive IHC and FISH. J Mol Diagn. 2013;15:341–346. 18. Ali G, Proietti A, Pelliccioni S, et al. ALK rearrangement in a large series of consecutive non-small cell lung cancers: comparison between a new immunohistochemical approach and fluorescence in situ hybridization for the screening of patients eligible for


19.

20. 21.

22.

23.

24.

25.

26.

27.

28.

813

crizotinib treatment. Arch Pathol Lab Med. 2014;138:1449–1458. Le Quesne J, Maurya M, Yancheva SG, et al. A comparison of immunohistochemical assays and FISH in detecting the ALK translocation in diagnostic histological and cytological lung tumor material. J Thorac Oncol. 2014;9:769–774. Ventana ALK. (D5F3) CDx Assay [package insert]. Tucson, AZ: Ventana Medical Systems, Inc.; 2015. Nitta H, Tsuta K, Yoshida A, et al. New methods for ALK status diagnosis in non-small-cell lung cancer: an improved ALK immunohistochemical assay and a new, Brightfield, dual ALK IHC-in situ hybridization assay. J Thorac Oncol. 2013;8:1019–1031. U.S. Food and Drug Administration. Statistical guidance on reporting results from studies evaluating diagnostic tests. http://www.fda.gov/RegulatoryInformation/Guidances/ ucm071148.htm. Accessed September 29, 2016. Eisenhauer EA, Therasse P, Bogaerts J, et al. New response evaluation criteria in solid tumours: revised RECIST guideline (version 1.1). Eur J Cancer. 2009;45:228–247. Leighl NB, Rekhtman N, Biermann WA, et al. Molecular testing for selection of patients with lung cancer for epidermal growth factor receptor and anaplastic lymphoma kinase tyrosine kinase inhibitors: American Society of Clinical Oncology endorsement of the College of American Pathologists/International Association for the study of lung cancer/association for molecular pathology guideline. J Clin Oncol. 2014;32:3673–3679. Galetta D, Rossi A, Pisocnti S, et al. The emerging role of ALK inhibitors in the treatment of advanced non-small cell lung cancer. Expert Opin Ther Targets. 2012;16(2):S45–S54. Sun J, Choi Y, Won J, et al. A dramatic response to crizotinib in a non-small cell lung cancer patient with IHC-positive and FISH-negative ALK. J Thorac Oncol. 2012;7:e36–e38. Novartis Pharmaceuticals. LDK378 versus chemotherapy in previously untreated patients with ALK rearranged non-small cell lung cancer. https://clinicaltrials.gov/ ct2/show/NCT01828099. Accessed September 29, 2016. Hoffmann-LaRoche. ALEX study: a randomized, phase III study comparing alectinib with crizotinib in treatment-naive anaplastic lymphoma kinase-positive advanced non-small cell lung cancer participants. https://clinicaltrials.gov/ ct2/show/NCT02075840. Accessed September 29, 2016.