Advertisement

Clinical Utility of Chromosomal Aneusomy in Individuals at High Risk of Lung Cancer

Open ArchivePublished:June 17, 2017DOI:https://doi.org/10.1016/j.jtho.2017.06.008

      Abstract

      Introduction

      Low-dose computed tomography screening for lung cancer has a high false-positive rate with frequent discovery of indeterminate pulmonary nodules. Noninvasive biomarkers are needed to reduce false positives and improve risk stratification. A retrospective longitudinal evaluation was performed to assess chromosomal aneusomy in sputum by fluorescence in situ hybridization (CA-FISH) in four nested case-control studies.

      Methods

      Receiver operating characteristic analysis resulted in two grouped cohorts: a high-risk cohort (Colorado High-Risk Cohort and Colorado Nodule Cohort [68 case patients and 69 controls]) and a screening cohort (American College of Radiology Imaging Network/National Lung Screening Trial and Pittsburgh Lung Screening Study [97 case patients and 185 controls]). The CA-FISH assay was a four-target DNA panel encompassing the EGFR and v-myc avian myelocytomatosis viral oncogene homolog (MYC) genes, and the 5p15 and centromere 6 regions or the fibroblast growth factor 1 gene (FGFR1) and phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit alpha gene (PIK3CA). A four-category scale (normal, probably normal, probably abnormal, and abnormal) was applied. Sensitivity, specificity, and positive and negative likelihood ratios (LRs) (with 95% confidence intervals [CIs]) were estimated for each cohort.

      Results

      Sensitivity and specificity were, respectively, 0.67 (95% CI: 0.55–0.78) and 0.94 (95% CI: 0.85–0.98) for high-risk participants and 0.20 (95% CI: 0.13–0.30) and 0.84 (95% CI: 0.78–0.89) for screening participants. The positive and negative LRs were, respectively, 11.66 (95% CI: 4.44–30.63) and 0.34 (95% CI: 0.24–0.48) for high-risk participants and 1.36 (95% CI: 0.81–2.28) and 0.93 (95% CI: 0.83–1.05) for screening participants.

      Conclusion

      The high positive LR of sputum CA-FISH indicates that it could be a useful adjunct to low-dose computed tomography for lung cancer in high-risk settings. For screening, however, its low positive LR limits clinical utility. Prospective assessment of CA-FISH in the incidentally identified indeterminate nodule setting is ongoing in the Colorado Pulmonary Nodule Biomarker Trial.

      Keywords

      Introduction

      In the context of low-dose computed tomography (LDCT) screening in current and former smokers, the false-positive rate is high (26% at first National Lung Screening Trial [NLST] screening
      • Aberle D.R.
      • Adams A.M.
      • et al.
      National Lung Screening Trial Research Team
      Reduced lung-cancer mortality with low-dose computed tomographic screening.
      and 13% with Lung-RADS criteria applied to NLST
      • Pinsky P.F.
      • Gierada D.S.
      • Black W.
      • et al.
      Performance of Lung-RADS in the National Lung Screening Trial: a retrospective assessment.
      ) and indeterminate nodules are frequently discovered (on 17% to 51% of chest computed tomography [CT] scans).
      • Deppen S.A.
      • Grogan E.L.
      Using clinical risk models for lung nodule classification.
      Noninvasive biomarkers are urgently needed to reduce false positives with screening LDCT and to improve risk stratification in those identified to have indeterminate nodules. Biofluids (sputum, blood, and urine) have also been considered to help identify participants who should undergo LDCT.
      • Ahmed N.
      • Bezabeh T.
      • Ijare O.B.
      • et al.
      Metabolic signatures of lung cancer in sputum and exhaled breath condensate detected by 1H magnetic resonance spectroscopy: a feasibility study.
      • Su Y.
      • Fang H.
      • Jiang F.
      Integrating DNA methylation and microRNA biomarkers in sputum for lung cancer detection.
      Since 2003, the University of Colorado Lung SPORE has explored a variety of biomarkers in noninvasively collected specimens for early detection of lung cancer. A major effort has been directed toward development of a predictive biomarker applied to sputum epithelial cells on the basis of the common occurrence of chromosomal abnormalities in this disease, mainly copy number variations due to chromosomal aneusomy (CA). A multitarget fluorescence in situ hybridization assay (FISH) has been developed to detect CA in lung cancer and termed CA-FISH.
      • Romeo M.S.
      • Sokolova I.A.
      • Morrison L.E.
      • et al.
      Chromosomal abnormalities in non-small cell lung carcinomas and in bronchial epithelia of high-risk smokers detected by multi-target interphase fluorescence in situ hybridization.
      • Varella-Garcia M.
      • Kittelson J.
      • Schulte A.P.
      • et al.
      Multi-target interphase fluorescence in situ hybridization assay increases sensitivity of sputum cytology as a predictor of lung cancer.
      • Varella-Garcia M.
      • Schulte A.P.
      • Wolf H.J.
      • et al.
      The detection of chromosomal aneusomy by fluorescence in situ hybridization in sputum predicts lung cancer incidence.
      To confirm that CA is an early occurrence, the CA-FISH assay was investigated in premalignant bronchial biopsy specimens. Jonsson et al.
      • Jonsson S.
      • Varella-Garcia M.
      • Miller Y.E.
      • et al.
      Chromosomal aneusomy in bronchial high-grade lesions is associated with invasive lung cancer.
      reported analyses in high-grade dysplasia lesions of 44 case patients with prevalent cancers compared with in 90 cancer-free controls. Approximately 64% of case patients and only 31% of controls exhibited CA (OR = 4.68, 95% confidence interval [CI]: 1.97–11.04). Massion et al.
      • Massion P.P.
      • Zou Y.
      • Uner H.
      • et al.
      Recurrent genomic gains in preinvasive lesions as a biomarker of risk for lung cancer.
      investigated genomic gains in six loci and two chromosomal regions in high-grade dysplasia lesions in case patients with 27 prevalent lung cancers and 43 controls. With a crude OR of 11 (p < 0.05), the sensitivity of a combination of four probes was 82% and the specificity was 58%. Finally, copy number variation was examined by Nakachi et al.
      • Nakachi I.
      • Rice J.L.
      • Coldren C.D.
      • et al.
      Application of SNP microarrays to the genome-wide analysis of chromosomal instability in premalignant airway lesions.
      initially by using a genome-wide approach based on single-nucleotide polymorphism microarrays and subsequently confirming by FISH in paired tumor-normal samples obtained by bronchial biopsy or brushings from six case patients with lung cancer, three with carcinoma in situ, and 11 with high-grade dysplasia. Genomic changes were detected in 100% of cancer cases (six of six) and carcinoma in situ cases (three of three), and in 27% (three of 11) of high-grade dysplasia cases. Both well-described and novel alterations were identified across the spectrum of samples.
      Copy number variations by FISH have also been detected in sputum epithelial cells, confirming the feasibility of using sputum for this purpose.
      • Romeo M.S.
      • Sokolova I.A.
      • Morrison L.E.
      • et al.
      Chromosomal abnormalities in non-small cell lung carcinomas and in bronchial epithelia of high-risk smokers detected by multi-target interphase fluorescence in situ hybridization.
      • Varella-Garcia M.
      • Kittelson J.
      • Schulte A.P.
      • et al.
      Multi-target interphase fluorescence in situ hybridization assay increases sensitivity of sputum cytology as a predictor of lung cancer.
      • Varella-Garcia M.
      • Schulte A.P.
      • Wolf H.J.
      • et al.
      The detection of chromosomal aneusomy by fluorescence in situ hybridization in sputum predicts lung cancer incidence.
      • Kettunen E.
      • Salmenkivi K.
      • Vuopala K.
      • et al.
      Copy number gains on 5p15, 6p11-q11, 7p12, and 8q24 are rare in sputum cells of individuals at high risk of lung cancer.
      • Li R.
      • Todd N.W.
      • Qiu Q.
      • et al.
      Genetic deletions in sputum as diagnostic markers for early detection of stage I non-small cell lung cancer.
      • Kang J.U.
      • Koo S.H.
      • Kwon K.C.
      • Park J.W.
      • Jung S.S.
      Gain of the EGFR gene located on 7p12 is a frequent and early event in squamous cell carcinoma of the lung.
      • Qiu Q.
      • Todd N.W.
      • Li R.
      • et al.
      Magnetic enrichment of bronchial epithelial cells from sputum for lung cancer diagnosis.
      • Katz R.L.
      • Zaidi T.M.
      • Fernandez R.L.
      • et al.
      Automated detection of genetic abnormalities combined with cytology in sputum is a sensitive predictor of lung cancer.
      • Jiang F.
      • Todd N.W.
      • Qiu Q.
      • Liu Z.
      • Katz R.L.
      • Stass S.A.
      Combined genetic analysis of sputum and computed tomography for noninvasive diagnosis of non-small-cell lung cancer.
      • Jiang F.
      • Todd N.W.
      • Li R.
      • Zhang H.
      • Fang H.
      • Stass S.A.
      A panel of sputum-based genomic marker for early detection of lung cancer.
      • Guber A.
      • Greif J.
      • Rona R.
      • et al.
      Computerized analysis of cytology and fluorescence in situ hybridization (FISH) in induced sputum for lung cancer detection.
      Table 1 summarizes 11 of these studies according to design, assay platform, conclusions, and criteria for positivity. Most of these investigations used sputum specimens from patients with diagnosed lung cancer and convenience samples of high-risk individuals (e.g., asbestos workers, heavy smokers, and patients with chronic obstructive pulmonary disease [COPD]) and compared them with specimens from healthy controls or never-smokers. In seven studies, sputum cytology was considered a primary sputum biomarker
      • Romeo M.S.
      • Sokolova I.A.
      • Morrison L.E.
      • et al.
      Chromosomal abnormalities in non-small cell lung carcinomas and in bronchial epithelia of high-risk smokers detected by multi-target interphase fluorescence in situ hybridization.
      • Varella-Garcia M.
      • Kittelson J.
      • Schulte A.P.
      • et al.
      Multi-target interphase fluorescence in situ hybridization assay increases sensitivity of sputum cytology as a predictor of lung cancer.
      • Kettunen E.
      • Salmenkivi K.
      • Vuopala K.
      • et al.
      Copy number gains on 5p15, 6p11-q11, 7p12, and 8q24 are rare in sputum cells of individuals at high risk of lung cancer.
      • Li R.
      • Todd N.W.
      • Qiu Q.
      • et al.
      Genetic deletions in sputum as diagnostic markers for early detection of stage I non-small cell lung cancer.
      • Qiu Q.
      • Todd N.W.
      • Li R.
      • et al.
      Magnetic enrichment of bronchial epithelial cells from sputum for lung cancer diagnosis.
      • Katz R.L.
      • Zaidi T.M.
      • Fernandez R.L.
      • et al.
      Automated detection of genetic abnormalities combined with cytology in sputum is a sensitive predictor of lung cancer.
      • Jiang F.
      • Todd N.W.
      • Qiu Q.
      • Liu Z.
      • Katz R.L.
      • Stass S.A.
      Combined genetic analysis of sputum and computed tomography for noninvasive diagnosis of non-small-cell lung cancer.
      and the presence of chromosomal abnormalities was evaluated with respect to incremental diagnostic value. Among those studies, six (except for that by Kettunen et al.
      • Kettunen E.
      • Salmenkivi K.
      • Vuopala K.
      • et al.
      Copy number gains on 5p15, 6p11-q11, 7p12, and 8q24 are rare in sputum cells of individuals at high risk of lung cancer.
      ) concluded that chromosomal abnormalities improved sensitivity, specificity, or both. Latency of the biomarker was considered in two of the 11 studies
      • Varella-Garcia M.
      • Kittelson J.
      • Schulte A.P.
      • et al.
      Multi-target interphase fluorescence in situ hybridization assay increases sensitivity of sputum cytology as a predictor of lung cancer.
      • Varella-Garcia M.
      • Schulte A.P.
      • Wolf H.J.
      • et al.
      The detection of chromosomal aneusomy by fluorescence in situ hybridization in sputum predicts lung cancer incidence.
      and the results suggested that it could be useful for early detection of lung cancer (12 to 18 months before diagnosis). Finally, criteria for positivity ranged from a percentage of abnormal cells among cells sampled (4%–9%) to counts of cells with abnormal targets (from ≥2 cells up to ≥6 cells with ≥2 targets) to three or more SDs above the mean percent of gain or loss.
      Table 1Studies Showing Feasibility of Using FISH for Detection of Chromosomal Aneusomy in Sputum
      ReferenceStudy ParticipantsGenes/Genomic Regions TestedCriteria for Positive ResultConclusions
      Romeo et al., 2003
      • Romeo M.S.
      • Sokolova I.A.
      • Morrison L.E.
      • et al.
      Chromosomal abnormalities in non-small cell lung carcinomas and in bronchial epithelia of high-risk smokers detected by multi-target interphase fluorescence in situ hybridization.
      (University of Colorado)
      3 case patients with lung cancer, 11 former smokers with normal sputum cytological testing resultsLAVysion (EGFR, MYC, 5p15CEP6)≥2 cells with gain in ≥2 targets in 11-100 DAPI abnormal cellsSputum samples from former smokers with normal cytological testing result had negative FISH result; sputum from patients with lung cancer were abnormal by FISH
      Varella-Garcia et al., 2004
      • Varella-Garcia M.
      • Kittelson J.
      • Schulte A.P.
      • et al.
      Multi-target interphase fluorescence in situ hybridization assay increases sensitivity of sputum cytology as a predictor of lung cancer.
      (University of Colorado)
      33 case patients with incident lung cancer, 33 cancer-free matched controlsLAVysion (EGFR, MYC, 5p15CEP6)≥2 cells with gain in ≥2 targets in 100 DAPI abnormal cellsCA-FISH improved sensitivity of cytological testing within 12 mo of diagnosis
      Kettunen et al., 2006
      • Kettunen E.
      • Salmenkivi K.
      • Vuopala K.
      • et al.
      Copy number gains on 5p15, 6p11-q11, 7p12, and 8q24 are rare in sputum cells of individuals at high risk of lung cancer.
      (Finland)
      20 case patients with lung cancer, 43 asbestos workers, 21 heavy smokers, 15 never-smokersLAVysion (EGFR, MYC, 5p15, CEP6)≥3 cells with gain in ≥2 targets in full slide scannedFISH and cytological testing had similar sensitivity (44%) for finding lung cancer
      Li et al., 2007
      • Li R.
      • Todd N.W.
      • Qiu Q.
      • et al.
      Genetic deletions in sputum as diagnostic markers for early detection of stage I non-small cell lung cancer.
      (MDACC, University of Maryland)
      38 case patients with stage 1 NSCLC, 36 smokers, 28 healthy nonsmokersFHIT - 3p14; HYAL2 - 3p21; SFTPC - 8p21.3Mean ± 3 SD for gain/loss (9%, 10%, and 9%, respectively)3p deletion markers are diagnostic for early stage lung cancer; FISH + cytological testing better than each alone
      Kang et al., 2008
      • Kang J.U.
      • Koo S.H.
      • Kwon K.C.
      • Park J.W.
      • Jung S.S.
      Gain of the EGFR gene located on 7p12 is a frequent and early event in squamous cell carcinoma of the lung.
      (Republic of South Korea)
      52 case patients with lung cancer, 15 former smokersLAVysion (EGFR, MYC, 5p15CEP6)≥6 cells with gain in ≥2 targets in 25 DAPI-abnormal cellsFISH aneusomy has potential utility as diagnostic method for early cancer
      Qiu et al., 2008
      • Qiu Q.
      • Todd N.W.
      • Li R.
      • et al.
      Magnetic enrichment of bronchial epithelial cells from sputum for lung cancer diagnosis.
      (MDACC, University of Maryland)
      29 case patients with stage I NSCLC, 26 heavy smokers, 28 healthy nonsmokersFHIT, 3p14, CEP3Same as Li et al., 2007
      • Li R.
      • Todd N.W.
      • Qiu Q.
      • et al.
      Genetic deletions in sputum as diagnostic markers for early detection of stage I non-small cell lung cancer.
      Magnetic enrichment of bronchial cells from sputum improved diagnostic value of FISH (58% sensitivity vs. 42%) and cytological testing
      Katz et al., 2008
      • Katz R.L.
      • Zaidi T.M.
      • Fernandez R.L.
      • et al.
      Automated detection of genetic abnormalities combined with cytology in sputum is a sensitive predictor of lung cancer.
      (MDACC, University of Maryland)
      35 case patients with lung cancer, 25 high-risk smokers, 6 healthy participants3p22.1/CEP3, SPA 10q22.3/CEP10Deletion: gene < control; mean ± 1 SD for gain/lossAutomated detection of FISH abnormal + cytological testing better than cytological testing alone for diagnosis; sensitivity 74% vs. 37%; specificity 82% vs. 87%
      Jiang et al., 2009
      • Jiang F.
      • Todd N.W.
      • Qiu Q.
      • Liu Z.
      • Katz R.L.
      • Stass S.A.
      Combined genetic analysis of sputum and computed tomography for noninvasive diagnosis of non-small-cell lung cancer.
      (MDACC, University of Maryland)
      33 case patients with stage I NSCLC, 49 cancer-free participants with other diseasesMini-chip FISH (HYAL2 - 3p21; FHIT - 3p14; p16 - 9p21; SPA - 10q22.3)Mean ± 3 SD for gain/loss (10%, 9%, and 9%, respectively) in 2000 scored nucleiFISH + CT was better than each alone for lung cancer diagnosis (CT sensitivity 85%, specificity 89%; FISH sensitivity 70%, specificity 92%; and CT + FISH sensitivity 91%, specificity 92%)
      Varella-Garcia et al., 2010
      • Varella-Garcia M.
      • Schulte A.P.
      • Wolf H.J.
      • et al.
      The detection of chromosomal aneusomy by fluorescence in situ hybridization in sputum predicts lung cancer incidence.
      (University of Colorado)
      100 case patients with incident lung cancer, 96 cancer-free matched controlsLAVysion (EGFR, MYC, 5p15, CEP6)≥2 cells with gain in ≥2 targets in 100 DAPI abnormal cellsCA-FISH higher sensitive within 18 mo of diagnosis; better sensitivity for squamous cell carcinoma than for other tumor types
      Jiang et al., 2010
      • Jiang F.
      • Todd N.W.
      • Li R.
      • Zhang H.
      • Fang H.
      • Stass S.A.
      A panel of sputum-based genomic marker for early detection of lung cancer.
      (University of Maryland)
      49 case patients with lung cancer, 49 case patients with COPD, 6 healthy smokersMini-chip FISH (15 genes and controls)Same as Jiang et al., 2009
      • Jiang F.
      • Todd N.W.
      • Qiu Q.
      • Liu Z.
      • Katz R.L.
      • Stass S.A.
      Combined genetic analysis of sputum and computed tomography for noninvasive diagnosis of non-small-cell lung cancer.
      6 genes were better (for gain ENO1, SKP2, 14-3-3zeta; for loss FHIT, HYAL2, p16); better for squamous than for adenocarcinoma histological subtype
      Guber et al., 2010
      • Guber A.
      • Greif J.
      • Rona R.
      • et al.
      Computerized analysis of cytology and fluorescence in situ hybridization (FISH) in induced sputum for lung cancer detection.
      (MDACC, BioView [Billerica, MA])
      12 case patients with lung cancer, 15 healthy nonsmokers3p22.1/CEP3, SPA 10q22.3/CEP10≥4% abnormal cells in 300 cellsComputer-assisted approach for sputum analyses combining cell morphology and FISH reached higher sensitivity (91.7%) and specificity (80%) for lung cancer detection
      FISH, fluorescence in situ hybridization; MYC, v-myc avian myelocytomatosis viral oncogene homolog; CEP6, centromere 6; CA, chromosomal aneusomy; DAPI, 4,6-diamino-2-phenylindole; MDACC, M. D. Anderson Cancer Center; FHIT, fragile histidine triad gene; HYAL2, hyaluronoglucosaminidase 2; SFTPC, surfactant protein C gene; CEP3, centromere 3; SPA, surfactant protein A gene; CEP10, centromere 10, CT, computed tomography; ENO1, enolase 1; SKP2, S-phase kinase associated protein 2.
      Validation of biomarkers is challenging. Study design standards for evaluating the predictive accuracy of diagnostic, screening, and prognostic biomarkers have been developed by Pepe et al.
      • Pepe M.S.
      • Feng Z.
      • Janes H.
      • Bossuyt P.M.
      • Potter J.D.
      Pivotal evaluation of the accuracy of a biomarker used for classification or prediction: standards for study design.
      Known as the Prospective Specimen Collection, Retrospective Blinded Evaluation (PRoBE) practice standards, their hallmark features are prospective ascertainment of the biomarker before occurrence of the relevant outcome (e.g., disease, recurrence, and progression), randomly selected case patients and controls from a cohort, and blinding when performing the biomarker assay. Testing of the biomarker, either alone or in combination with other biomarkers and clinical data, through the nested case-control design must take place within the intended clinical context, and the performance criteria for predictive accuracy must be prespecified.
      This article reports on a retrospective longitudinal evaluation incorporating many of the features of a PRoBE design that was performed to assess the potential of a CA-FISH biomarker for early detection of lung cancer.

      Materials and Methods

      Participant Population

      The study population for the testing and validation of a CA-FISH biomarker for lung cancer in nested case-control study designs comprised the populations of four trials investigating biomarkers to predict lung cancer or use of LDCT to diagnose lung cancer: the Colorado High-Risk Cohort, the NLST, the Pittsburgh Lung Screening Study (PLuSS), and the Colorado Pulmonary Nodule Biomarker Trial (PuNBaT). All participants signed a consent form and the studies were institutional review board–approved.

      Colorado High-Risk Cohort

      The Colorado High-Risk Cohort comprises participants recruited from 1993 to 2003 in community and academic pulmonary clinics primarily in the metropolitan area of Denver, Colorado, to determine the relationship between sputum cytology and incident lung cancer. The methods of accrual and processing of specimens have been previously described.
      • Kennedy T.C.
      • Proudfoot S.P.
      • Franklin W.A.
      • et al.
      Cytopathological analysis of sputum in patients with airflow obstruction and significant smoking histories.
      Briefly, at the time of enrollment, all participants had a cigarette smoking history of at least 30 pack-years and significant COPD but no history of cancer or acute respiratory infection. These patients were asked to collect two consecutive 3-day, early-morning, spontaneous cough sputum specimens in jars containing Saccomanno’s fixative (2% carbowax and 50% alcohol) annually. Sputum specimens were stored at room temperature in the Tissue Procurement Core of the University of Colorado SPORE in Lung Cancer. Cohort members were followed by active methods, including telephone and mail contact, and by passive methods, including matching to the Colorado Department of Public Health and Environment Vital Statistics and Central Cancer Registry records. In 2005, 114 case patients with incident lung cancer were identified and matched (according to sex, age, and date of enrollment) with 114 controls who had remained clinically cancer-free for at least 5 years. CA-FISH assay results were obtained from 88% of the case patients and 84% of the controls. From this study, 55 case patients with sputum samples collected within 18 months of diagnosis and 59 controls were included in the current evaluation.

      ACRIN/NLST

      The NLST was a randomized trial of screening carried out between 2002 and 2010 that compared LDCT with chest radiography (CXR) with regard to reductions in lung cancer mortality.
      • Aberle D.R.
      • Adams A.M.
      • et al.
      National Lung Screening Trial Research Team
      Reduced lung-cancer mortality with low-dose computed tomographic screening.
      Participants were recruited from the general community between 2002 and 2004. Eligible participants were between 55 and 74 years of age at the time of randomization, had a history of cigarette smoking of at least 30 pack-years, and if former smokers, had quit within the previous 15 years. Persons who had previously received a diagnosis of lung cancer, had undergone chest CT within 18 months before enrollment, had hemoptysis, or had an unexplained weight loss of more than 6.8 kg (15 lb) in the preceding year were excluded. A subgroup of trial participants were asked to provide sputum and blood for future studies, using protocols originally developed by the University of Colorado Cancer Center (UCCC) Lung SPORE Tissue Core. UCCC investigators were approved to use sputum samples by an NLST–American College of Radiology Imaging Network (ACRIN) ad hoc committee and the ACRIN Data Access Committee in 2011. Samples from 91 case patients and 182 controls (1:2) matched on age, sex, and enrollment site were received for evaluation. A cytological preassessment showed that 151 spontaneously collected sputum samples met the criteria to be used for the CA-FISH assay. Successful and quantifiable CA-FISH results were obtained for 49 case samples (42 were collected within 18 months of diagnosis), and 96 control samples were included in the current evaluation. The case patients were more likely than the controls to be current smokers and more likely to have COPD at enrollment. The ACRIN samples originated from both the CT and the CXR study arms. Among the 145 ACRIN samples investigated in this study, 86 belonged to the CT arm and 59 to the CXR arm (59% and 41% of the total, respectively). Of the ACRIN case samples, 30 came from the CT arm and 19 from the CXR arm (61% and 39% of the total number of ACRIN cases, respectively).

      PLuSS

      The Pittsburgh Lung Screening Study (PLuSS) was a community-based research cohort of current and ex-smokers who were being followed for lung cancer.
      • Wilson D.O.
      • Weissfeld J.L.
      • Fuhrman C.R.
      • et al.
      The Pittsburgh Lung Screening Study (PLuSS): outcomes within 3 years of a first computed tomography scan.
      Briefly, between January 2002 and April 2005, PLuSS used mass mailings and advertisements to enroll 3754 volunteers, mainly from southwestern Pennsylvania, with the following characteristics: (1) age 50 to 79 years; (2) no personal history of lung cancer; (3) current or ex-smoker of at least 10 cigarettes per day for at least 25 years and, if an ex-smoker, stopped smoking not more than 10 years before enrollment; (4) no chest computed tomography within 12 months; and (5) body weight less than 400 pounds. Of the 3754 volunteers enrolled, 92.9% were white and more than 95% had a cigarette smoking history at enrollment of at least 20 pack-years. Sputum samples collected from 60 case patients and 120 controls matched on age, sex, and enrollment were preevaluated; 139 met the criteria to be used for the CA-FISH assay. Successful CA-FISH results were obtained from 139 samples. Of these, samples from 48 case patients and 89 controls were included in this analysis. Two samples were excluded because of inability to verify lung cancer status.

      Colorado PuNBaT

      The Pulmonary Nodule Biomarker Trial (PuNBaT) is a prospective study that was started in 2010 to establish a cohort of 500 individuals with lung nodules of indeterminate etiology incidentally identified by CT scans; it is currently in the accrual stage. Patients are adults from 18 to 85 years old who were referred to pulmonologists, oncologists, or thoracic surgeons for evaluation of peripheral lung nodules found on CT scan. Repeat CT scans, biopsy, or surgical excision are clinically indicated to determine the etiology of the nodule, and one or more lung nodules must be between 8 mm and 30 mm in the greatest diameter. The main exclusion criteria include (1) lung nodules or masses larger than 30 mm in the greatest dimension; (2) CT evidence of partial or complete obstruction of a lobar bronchus, mainstem bronchus, or trachea; (3) lung nodules with benign calcification; (4) a diagnosis of cancer, with the exception of nonmelanoma skin cancer, within 2 years before study entry; and (5) life expectancy less than 6 months. Participants allowed investigators access to CT scan images and spirometry data, and they provided blood, sputum, urine, and exhaled breath samples over the course of the study. A malignant nodule developed in approximately 33%. Cytological preassessment identified 25 of 36 sputum samples as acceptable for FISH assay. Of these, 13 case patients with lung cancer and 10 controls were included in the current evaluation. Two case patients' cancers that were not lung primaries were excluded from analysis.

      Measures

      Lung cancer incidence was determined in all studies by following cohort members using active methods (including telephone and mail contact) and passive methods (including matching to the state and national vital statistics records and to the state and hospital cancer registry records). Review of medical, pathology, and tumor-staging records was also performed.
      Clinical characteristics such as demographic data, cigarette smoking history, medical history (including pulmonary lung function, COPD status, and personal history of cancer), and current medical problems were ascertained by questionnaire at enrollment. A set of common clinical measures was established to compare risk across the cohorts by using the Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial Modified 2012
      • Tammemagi M.C.
      • Katki H.A.
      • Hocking W.G.
      • et al.
      Selection criteria for lung-cancer screening.
      lung cancer risk model. For risk fields missing or unavailable, zero added risk was assumed during calculation of the PLCOM2012 risk score, with the exception of education, which was assumed to be post–high school training for all participants.
      This evaluation followed most of the PRoBE design principles within each of the clinical contexts represented by the four respective cohort studies, with the exception that the classification criterion was not predefined but rather established within the current evaluation.
      On the basis of receiver operating characteristic (ROC) analysis of the CA-FISH score and examination of the PLCOM2012 risk score, participants from these four studies were combined into two groups: a high-risk group (Colorado High-Risk Cohort [55 case patients and 59 controls] and Colorado Nodule Cohort [13 case patients and 10 controls]) and a screening group (ACRIN/NLST [49 case patients and 96 controls] and PLuSS [48 case patients and 89 controls]).

      Analytical Methods

      Sputum suspensions were collected and frozen, or pelleted and frozen, and stored at –86oC for 1 to 3 years in the cytological fixative Saccomanno’s fluid (Fisher HealthCare, Houston, TX). After thawing and incubation in Sputolysin reagent (Calbiochem, San Diego, CA), the cells were pelleted and resuspended in 1× phosphate-buffered saline and cytospin preps were generated. One slide per case patient was stained with Papanicolaou stain for assessment of adequacy, and specimens with at least 100 epithelial cells and 50 alveolar macrophages per quadrant were subjected to FISH. A detailed description is presented in the Supplementary Materials and Methods.
      The CA-FISH assay was a four-color, four-target panel including genomic sequences encompassing the EGFR and v-myc avian myelocytomatosis viral oncogene homolog (MYC) genes and the 5p15 and centromere 6 regions (FISH panel 1) or the fibroblast growth factor 1 gene (FGFR1) and phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit alpha gene (PIK3CA) genes (FISH panel 2). Specific probe information is presented in the Supplementary Materials and Methods. The CA-FISH assays were performed as previously described
      • Varella-Garcia M.
      • Schulte A.P.
      • Wolf H.J.
      • et al.
      The detection of chromosomal aneusomy by fluorescence in situ hybridization in sputum predicts lung cancer incidence.
      ; the details are provided in the Supplementary Materials and Methods. CA-FISH analyses were performed with the Papanicolaou-stained slide (Sigma-Aldrich, St. Louis, MO) to guide the identification of cancer cells in the preparations, and two subpopulations of cells were selected and analyzed (subpopulation 1, which included normal-appearing epithelial cells, and subpopulation 2, which included cells with a tumor-like appearance and other cells, as described in Supplementary Materials and Methods). The interpretation criteria were based on copy number gain, and at the participant level, the assay was scored on a four-category scale representing (1) abnormal, (2) probably abnormal, (3) probably normal, and (4) normal, as described and illustrated in Figure 1.
      Figure thumbnail gr1
      Figure 1Scoring criteria and illustration of the four-category scale used for the assay determining chromosomal aneusomy in sputum by fluorescence in situ hybridization. CNG, copy number gain.

      Statistical Analysis

      Sex, COPD status, smoking status and history of tobacco exposure, PLCOM2012 lung cancer risk score, lung function, age, time from sputum collection to lung cancer diagnosis, and lung cancer histologic subtype for cases were compared by using analysis of variance and chi-square tests, as appropriate. For each of the nested case-control studies, a ROC analysis was performed by using logistic regression models with lung cancer status as the outcome and the four-category CA-FISH score as the only predictor. The results of that analysis showed that the discriminatory ability of CA-FISH differed between the two screening cohorts and the two high-risk cohorts but was similar within those two groupings. The best cutoff for CA-FISH was identified within each grouping by using Youden’s index.
      • Youden W.J.
      Index for rating diagnostic tests.
      For that cutoff, the sensitivity, specificity, positive likelihood ratio (LR), and negative LR were estimated with their respective 95% CIs.
      • Simel D.L.
      • Samsa G.P.
      • Matchar D.B.
      Likelihood ratios with confidence: sample size estimation for diagnostic test studies.
      These analyses were also performed separately in the subgroup of participants with COPD.

      Results

      Characteristics of the participants in each of the nested case-control studies are shown in Table 2. The study groups are similar in variables such as mean age (62–68 years), mean duration of smoking (43.4–46.5 years), and mean number of cigarettes smoked per day (26.4–31.5). Several differences were observed among the study groups, most notably with regard to pulmonary function (forced expiratory volume in 1 second percent of predicted 42.3–77.8), COPD status (48%–97%), former smoker status (26%–62%), and mean years from quitting smoking (4.7–8.1). Tumor histological subtypes also varied across the cohorts, including nearly equal numbers of squamous cell carcinomas and adenocarcinomas in the Colorado High-Risk Cohort and PLuSS cohort and greater percentages of adenocarcinoma in the Colorado Nodule Cohort and ACRIN samples. The median time elapsed between the collection of sputum samples and clinical diagnosis was shortest in the Colorado Nodule Cohort (−0.7 months) and longest in the PLuSS cohort (8.5 months) with the ACRIN sample and Colorado High-Risk Cohort falling in between (4 and 8 months, respectively).
      Table 2Risk Characteristics, by Cohort, of Participants with CA-FISH Results
      Source CohortCO High Risk (n = 114; 55 LC Cases)ACRIN (n = 145; 49 LC Cases)PLuSS (n = 137; 48 LC Cases)CO Nodule (n = 23; 13 LC Cases)ACRIN and PLuSS (n = 282; 97 LC Cases)CO High Risk and CO Nodule (n = 137; 68 LC Cases)
      N%N%N%N%N%N%
      Sex
      Statistically significant difference between the four cohorts.
       Male8675%9263%10879%2087%20071%10677%
       Female2825%5337%2921%313%8229%3123%
      COPD or emphysema
      Statistically significant difference between the four cohorts.
      ,
      Percentages adding up to less than 100% indicate missing data. Emphysema as determined on CT scan for participants with CT scan available.
      ,
      Statistically significant difference between the two combined cohorts.
       Yes11197%8357%9771%1748%18064%12893%
       No00%6142%4029%643%10136%65%
      Smoking
      Percentages adding up to less than 100% indicate missing data. Emphysema as determined on CT scan for participants with CT scan available.
       Former smoker
      Statistically significant difference between the four cohorts.
      ,
      Statistically significant difference between the two combined cohorts.
      7162%5337%3526%730%8831%7857%
       Current smoker4337%9263%10274%1670%19469%5943%
       Mean No. of cigarettes per day (SD)
      Statistically significant difference between the four cohorts.
      ,
      Statistically significant difference between the two combined cohorts.
      31.5 (14.2)26.4 (9.1)26.4 (9.0)26.5 (11.8)26.4 (9.0)30.6 (13.9)
       Mean duration of smoking (SD), y
      Statistically significant difference between the four cohorts.
      43.8 (9.3)43.4 (7.6)46.5 (6.9)43.8 (8.9)44.9 (7.5)43.8 (9.2)
       Mean time since quitting smoking (SD), y
      Statistically significant difference between the four cohorts.
      ,
      Statistically significant difference between the two combined cohorts.
      9.1 (8.1)8.4 (4.1)4.7 (3.9)5.9 (8.4)6.7 (4.4)8.8 (8.1)
      Mean PLCOM2012 risk (SD)
      Statistically significant difference between the four cohorts.
      ,
      Statistically significant difference between the two combined cohorts.
      7.41 (5.10)4.49 (3.17)5.90 (4.62)5.16 (4.89)5.17 (3.99)7.04 (5.12)
      Mean FEV1 (SD), % predicted
      Statistically significant difference between the four cohorts.
      ,
      Statistically significant difference between the two combined cohorts.
      42.3 (16.8)77.8 (24.8)73.6 (20.5)67.0 (23.9)75.8 (22.8)46.2 (20.1)
      Mean age (SD), y
      Statistically significant difference between the four cohorts.
      ,
      Statistically significant difference between the two combined cohorts.
      68 (7.5)63 (4.9)65 (6.4)62 (9.0)64 (5.7)67 (8.1)
      Lung cancer histological subtype for cases
      Statistically significant difference between the four cohorts.
      ,
      Statistically significant difference between the two combined cohorts.
       Adenocarcinoma1731%2653%1225%
       Squamous cell carcinoma1629%510%1429%
       SCLC59%1020%817%
       Other1324%816%1429%
       Unknown47%00%00%
      Median mo (IQR 25th and 75th percentile) from sputum collection to lung cancer diagnosis
      Statistically significant difference between the four cohorts.
      ,
      Unavailable for seven ACRIN LC cases.
      8 (1, 11)4 (1, 7)8.5 (3.5, 21)–0.7 (–1.4, 0.5)
      CA, chromosomal aneusomy; FISH, fluorescence in situ hybridization; CO, Colorado; LC, lung cancer; ACRIN, American College of Radiology Imaging Network; PLuSS, Pittsburgh Lung Screening Study; PLCOM2012, Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial Modified 2012; FEV1, forced expiratory volume in 1 second; IQR, interquartile range; CT, computed tomography.
      a Statistically significant difference between the four cohorts.
      b Percentages adding up to less than 100% indicate missing data. Emphysema as determined on CT scan for participants with CT scan available.
      c Statistically significant difference between the two combined cohorts.
      d Unavailable for seven ACRIN LC cases.
      Table 3 shows the distributional properties of the sputum CA-FISH score in each of the four studies. A high percentage of controls in all of the nested case-control studies (ranging from 80% to 95%) exhibited normal FISH scores; only 5% to 20% of controls displayed scores considered abnormal. In the ACRIN/NLST and PLuSS cohorts a low percentage of case patients (16% and 25%, respectively) had abnormal FISH scores, whereas in the other two studies a much higher percentage of case patients (46% and 73%) had abnormal FISH scores.
      Table 3Four-Category CA-FISH Score within Cohort by Case-Control Status
      CA-FISH ScoreCO High Risk (n = 114; 55 LC Cases)ACRIN (n = 145; 49 LC Cases)PLuSS (n = 137; 48 LC Cases)CO Nodule (n = 23; 13 LC Cases)ACRIN and PLuSS (n = 282; 97 LC Cases)CO High Risk and CO Nodule (n = 137; 68 LC Cases)
      CasesControlsCasesControlsCasesControlsCasesControlsCasesControlsCasesControls
      Sensitivity72.7%16.3%25.0%46.2%20.6%67.6%
      Specificity94.9%80.2%89.9%90.0%84.8%94.2%
      Positive LR14.300.822.474.6211.661.36
      Negative LR0.291.040.830.600.340.93
       Abnormal40 (73%)3 (5%)8 (16%)19 (20%)12 (25%)9 (10%)6 (46%)1 (10%)20 (21%)28 (15%)46 (68%)4 (6%)
       Not abnormal15 (27%)56 (95%)41 (84%)77 (80%)36 (75%)80 (90%)7 (54%)9 (90%)77 (79%)157 (85%)22 (32%)65 (95%)
      Probably abnormal1 (2%)2 (3%)4 (8%)19 (19%)2 (4%)6 (7%)3 (23%)1 (10%)
      Probably normal7 (13%)22 (38%)6 (12%)8 (8%)8 (16%)18 (20%)2 (15%)0 (0%)
      Normal7 (13%)32 (54%)31 (63%)50 (52%)26 (54%)56 (63%)2 (15%)8 (80%)
      CA, chromosomal aneusomy; FISH, fluorescence in situ hybridization; CO, Colorado; LC, lung cancer; ACRIN, American College of Radiology Imaging Network; PLuSS, Pittsburgh Lung Screening Study; LR, likelihood ratio.
      The ability of the four-category sputum CA-FISH score to discriminate between those participants in whom lung cancer did and did not develop is summarized with the use of ROC curves in Figure 2. These curves demonstrate that the score appeared to operate similarly in the two high-risk cohorts (Colorado High-Risk Cohort and Colorado Nodule Cohort) and similarly in the two screening cohorts (PLuSS and ACRIN/NLST), and thus, the remainder of the analyses was based on these two groupings.
      Figure thumbnail gr2
      Figure 2Receiver operating characteristic curves and areas under the curve (AUC) for assay determining chromosomal aneusomy in sputum by fluorescence in situ hybridization, by nested case-control study cohort. CO High-Risk, Colorado High-Risk Cohort; CO Nodule; Colorado Nodule Cohort; PLuSS, Pittsburgh Lung Screening Study; ACRIN/NLST, American College of Radiology Imaging Network/National Lung Screening Trial.
      By applying Youden’s index to each of the potential cutoffs for the CA-FISH score, we identified the cutoff of abnormal versus not abnormal as optimal with regard to maximizing the unweighted sum of sensitivity and specificity. When this cutoff was used for the high-risk group, the sensitivity and specificity were 0.67 (95% CI: 0.55–0.78) and 0.94 (95% CI: 0.85–0.98), respectively; for the screening group, the sensitivity and specificity were 0.20 (95% CI: 0.13–0.30) and 0.84 (95% CI: 0.78–0.89), respectively. The respective positive and negative LRs for high-risk participants were 11.66 (95% CI: 4.44–30.63) and 0.34 (95% CI: 0.24–0.48), and the respective positive and negative LRs for screening participants were 1.36 (95% CI: 0.81–2.28) and 0.93 (95% CI: 0.83–1.05). Similar results were observed when only COPD participants were analyzed (data not shown).

      Discussion

      Lung carcinomas have been known for a long time to frequently exhibit marked chromosome aneusomy reflecting disruptions in the cellular mechanisms ensuring genetic stability.
      • Tonon G.
      • Wong K.K.
      • Maulik G.
      • et al.
      High-resolution genomic profiles of human lung cancer.
      • Weir B.A.
      • Woo M.S.
      • Getz G.
      • et al.
      Characterizing the cancer genome in lung adenocarcinoma.
      • Hammerman P.S.
      • Lawrence M.S.
      Cancer Genome Atlas Research Network
      Comprehensive genomic characterization of squamous cell lung cancers.
      Most common are the defects in the mitotic spindle checkpoints, the abnormal centrosome formation, and failures in the cytokinesis process, which are frequently associated in complex arrangements.
      • Masuda A.
      • Takahashi T.
      Chromosome instability in human lung cancers: possible underlying mechanisms and potential consequences in the pathogenesis.
      • Silk A.D.
      • Zasadil L.M.
      • Holland A.J.
      • Vitre B.
      • Cleveland D.W.
      • Weaver B.A.
      Chromosome missegregation rate predicts whether aneuploidy will promote or suppress tumors.
      We meant to take advantage of this general feature to screen for aneusomic cells in sputum specimen as a biomarker for detection of lung cancer. A similar principle was examined for its potential role as a prognostic marker in lung cancer studies. For instance, increased copy number in EGFR, MYC, 5p15.2, and centromere 6 and deletion in cyclin-dependent kinase inhibitor 2A detected by FISH was associated with poor overall survival rate in lung squamous cell carcinoma.
      • Yoo J.W.
      • Seo K.W.
      • Jang S.J.
      • et al.
      The relationship between the presence of chromosomal instability and prognosis of squamous cell carcinoma of the lung: fluorescence in situ hybridization analysis of paraffin-embedded tissue from 47 Korean patients.
      Deletions in 3p detected by comparative genomic hybridization were associated with decreased overall survival in lung squamous cell carcinomas independent of clinicopathological parameters.
      • Danner B.C.
      • Hellms T.
      • Jung K.
      • et al.
      Prognostic value of chromosomal imbalances in squamous cell carcinoma and adenocarcinoma of the lung.

      Challenges of the CA-FISH Analyses

      The CA-FISH assay offers significant challenges for its performance in sputum. First, sputum is a dirty and viscous suspension that includes particles and cellular debris with fluorescent characteristics, which contribute to a high background noise and impair the clear visualization of cells. Second, Saccomanno’s fluid is not an optimal long-term fixative and the cell fragility conflicts with the stringency of the protocol necessary to reduce the background noise. It is possible that tumor cells are preferentially destroyed during the procedure because they are known to be less resistant than normal cells to temperature and chemical treatments. Selection of nuclei to score is subjective. We choose to select cells on the basis of features that have documented association with lung carcinoma, such as nucleus of large size, irregular borders, and grainy chromatin texture. We also considered that tumor cells typically stained dimmer with 4,6-diamino-2-phenylindole than did nontumor cells, on account of a looser degree of chromatin condensation. These features seemed to effectively aid in the cell selection, but the manual analytical process is tedious and time-consuming; ideally, an automated or computer-assisted platform should be used. An interesting approach was described by Guber et al.
      • Guber A.
      • Greif J.
      • Rona R.
      • et al.
      Computerized analysis of cytology and fluorescence in situ hybridization (FISH) in induced sputum for lung cancer detection.
      using the automated scanning and analyses by the Duet (BioView, Billerica, MA). Combining FISH variables with parameters from cell morphology, they reached high sensitivity (∼92%) and specificity (∼80%) in the detection of lung cancer. Moreover, tumor cells are scarce in sputum fluids (<1%),
      • Thunnissen F.B.
      Sputum examination for early detection of lung cancer.
      and strategies to enrich the biofluid samples for tumor cells such as those described by Qiu et al.
      • Qiu Q.
      • Todd N.W.
      • Li R.
      • et al.
      Magnetic enrichment of bronchial epithelial cells from sputum for lung cancer diagnosis.
      may also be necessary for clinical implementation of the assay.
      The limited number of targets that can be simultaneously tested in a FISH assay is also a caveat. Because we are looking for chromosomal numerical changes significantly common in lung cancer, in our efforts we have examined four DNA targets. Nevertheless, screening for a larger panel of probes would very likely increase both the sensitivity and specificity of the CA-FISH assay. An interesting in situ mini-chip approach for FISH probes has been successfully developed and tested in sputum, including sets of seven
      • Jiang F.
      • Todd N.W.
      • Qiu Q.
      • Liu Z.
      • Katz R.L.
      • Stass S.A.
      Combined genetic analysis of sputum and computed tomography for noninvasive diagnosis of non-small-cell lung cancer.
      and of 15 DNA sequences.
      • Jiang F.
      • Todd N.W.
      • Li R.
      • Zhang H.
      • Fang H.
      • Stass S.A.
      A panel of sputum-based genomic marker for early detection of lung cancer.
      Although the targets used in those studies do not overlap with ours and mostly covered genomic areas known for being deleted in lung cancer, in both studies the authors demonstrated an improvement of sensitivity and specificity of their test when more targets were simultaneously tested.

      Challenges of Investigating CA-FISH as a Biomarker to Predict/Detect Lung Cancer

      After promising performance of CA-FISH as a biomarker for lung cancer in sputum from patients with COPD was observed with the use of a nested case-control design, further evaluation based on PRoBE design principles has highlighted the considerable variability that occurs in lung cancer. Our attempt at a pooled analysis of CA-FISH in sputum in 165 case patients and 254 matched controls illustrates the difficulty in carrying out a standard biomarker validation in the lung cancer environment. Further statistical analysis has illustrated, however, the potential of this marker in patients whose clinical presentation merited specialized care by pulmonologists. This is in contrast to the biomarker’s lack of performance in high-risk patients from the general population identified by LDCT screening. In combination, these results support the well-developed idea that lung cancer, like breast, prostate, and colon cancer, represents a family of disease, with LDCT screening yielding patients that followed several different paths of carcinogenesis. This heterogeneity is supported by the preliminary positive performance of CA-FISH in sputum from a small group of patients who were identified to have nodules 8 to 30 mm in diameter and were referred to pulmonologists for further evaluation (as described in Materials and Methods).

      Interval between Testing and Clinical Diagnosis

      In the screening setting it is critical to consider the optimal interval for applying the biomarker testing. Tests performed too often are unnecessarily expensive, and tests performed too rarely miss the window of opportunity for preclinical diagnosis. Our previous studies have observed abnormal cells in sputum in specimens collected 12 and 18 months before clinical diagnosis, but not in specimens collected long before.
      • Varella-Garcia M.
      • Kittelson J.
      • Schulte A.P.
      • et al.
      Multi-target interphase fluorescence in situ hybridization assay increases sensitivity of sputum cytology as a predictor of lung cancer.
      • Varella-Garcia M.
      • Schulte A.P.
      • Wolf H.J.
      • et al.
      The detection of chromosomal aneusomy by fluorescence in situ hybridization in sputum predicts lung cancer incidence.
      Recent data also support a similar conclusion. An investigation of DNA hypermethylation in sputum of asymptomatic participants in the NELSON trial who were at risk for lung cancer
      • Hubers A.J.
      • Heideman D.A.
      • Duin S.
      • et al.
      DNA hypermethylation analysis in sputum of asymptomatic subjects at risk for lung cancer participating in the NELSON trial: argument for maximum screening interval of 2 years.
      analyzed data on the basis of numerous screening interval durations ranging between 1 and 5 years. The results indicated that 2 years was the longest screening interval maintaining the discriminatory capacity of the biomarker. At 3- to 5-year screening intervals, the detection of preclinical disease was significantly reduced. The median time elapsed between sputum sample collection and diagnosis in the current study was on average shorter than 18 months, as shown in Table 2.

      Challenges of Biomarker Validation Analysis

      We carried out exploratory analyses to better understand the operating characteristics of CA-FISH as a biomarker for early detection of lung cancer in a variety of cohort settings. In a true PRoBE evaluation, all cutoffs would have been predefined to establish validation. Further work will be needed in the future with one or more external cohorts to accomplish a full PRoBE-style validation study.
      As part of PRoBE, the clinical context in which the biomarker will be applied is of critical importance. Our work has clarified that CA-FISH in sputum has potential for reducing diagnostic errors in a clinical setting in which false-positive rates are high (e.g., CT imaging for nodules). More specifically, a positive CA-FISH test result could improve medical decision making in patient management, as illustrated briefly in the next section.

      Potential Clinical Utility of CA-FISH in High-Risk Settings

      The high positive LR (>10.0) of sputum CA-FISH seen in the high-risk participants in our study indicates that this noninvasive biomarker could be a clinically useful adjunct to LDCT among high-risk patients referred for evaluation for lung cancer, such as the small subset of patients with nodules included in the high-risk group (described in Materials and Methods). As an example, if a hypothetical patient with indeterminate nodules has a pretest (CA-FISH) lung cancer risk of 20%, a positive CA-FISH test result with a positive LR of 11.66 would raise their posttest probability of lung cancer to 78%. This would likely have an impact on clinical management to support biopsy rather than a strategy involving serial imaging to determine nodule growth. In the screening setting, however, the low positive LR (<2.0) of CA-FISH limits its clinical utility both as a prescreening tool and for postscreen risk stratification.

      Conclusion

      Our evaluation of CA-FISH in sputum suggests the potential of this noninvasive biomarker to complement CT imaging and other diagnostic tests in very high-risk settings, such as the management of individuals with indeterminate nodules. However, the results do not suggest that the biomarker would be useful in low-risk settings, such as screening with LDCT. Whether a similarly high positive LR will be reproducible in a larger high-risk cohort of persons with indeterminate nodules remains to be seen. Prospective assessment of sputum CA-FISH is ongoing in the Colorado Pulmonary Nodule Biomarker Trial supported by the Colorado SPORE in Lung Cancer.

      Acknowledgments

      This work was supported by the National Institutes of Health/National Cancer Institute Lung Cancer Specialized Programs of Research Excellence grants P50CA058187 (University of Colorado) and P50 090440 (University of Pittsburgh), the LUNGevity Foundation, and the Gift of Life and Breath Foundation. Technical support was provided through National Cancer Institute Cancer Center Support Grant NCI CCSG P30CA046934 (UCCC/Molecular Pathology Shared Resource). Specimen support was obtained from the National Lung Screening Trial/American College of Radiology Imaging Network (Drs. Franklin and Garg were principal investigators on our study.)

      Supplementary Data

      References

        • Aberle D.R.
        • Adams A.M.
        • et al.
        • National Lung Screening Trial Research Team
        Reduced lung-cancer mortality with low-dose computed tomographic screening.
        N Engl J Med. 2011; 365: 395-409
        • Pinsky P.F.
        • Gierada D.S.
        • Black W.
        • et al.
        Performance of Lung-RADS in the National Lung Screening Trial: a retrospective assessment.
        Ann Intern Med. 2015; 162: 485-491
        • Deppen S.A.
        • Grogan E.L.
        Using clinical risk models for lung nodule classification.
        Semin Thorac Cardiovasc Surg. 2015; 27: 30-35
        • Ahmed N.
        • Bezabeh T.
        • Ijare O.B.
        • et al.
        Metabolic signatures of lung cancer in sputum and exhaled breath condensate detected by 1H magnetic resonance spectroscopy: a feasibility study.
        Magn Reson Insights. 2016; 9: 29-35
        • Su Y.
        • Fang H.
        • Jiang F.
        Integrating DNA methylation and microRNA biomarkers in sputum for lung cancer detection.
        Clin Epigenetics. 2016; 8: 109
        • Romeo M.S.
        • Sokolova I.A.
        • Morrison L.E.
        • et al.
        Chromosomal abnormalities in non-small cell lung carcinomas and in bronchial epithelia of high-risk smokers detected by multi-target interphase fluorescence in situ hybridization.
        J Mol Diagn. 2003; 5: 103-112
        • Varella-Garcia M.
        • Kittelson J.
        • Schulte A.P.
        • et al.
        Multi-target interphase fluorescence in situ hybridization assay increases sensitivity of sputum cytology as a predictor of lung cancer.
        Cancer Detect Prev. 2004; 28: 244-251
        • Varella-Garcia M.
        • Schulte A.P.
        • Wolf H.J.
        • et al.
        The detection of chromosomal aneusomy by fluorescence in situ hybridization in sputum predicts lung cancer incidence.
        Cancer Prev Res (Phila). 2010; 3: 447-453
        • Jonsson S.
        • Varella-Garcia M.
        • Miller Y.E.
        • et al.
        Chromosomal aneusomy in bronchial high-grade lesions is associated with invasive lung cancer.
        Am J Respir Crit Care Med. 2008; 177: 342-347
        • Massion P.P.
        • Zou Y.
        • Uner H.
        • et al.
        Recurrent genomic gains in preinvasive lesions as a biomarker of risk for lung cancer.
        PLoS One. 2009; 4: e5611
        • Nakachi I.
        • Rice J.L.
        • Coldren C.D.
        • et al.
        Application of SNP microarrays to the genome-wide analysis of chromosomal instability in premalignant airway lesions.
        Cancer Prev Res (Phila). 2014; 7: 255-265
        • Kettunen E.
        • Salmenkivi K.
        • Vuopala K.
        • et al.
        Copy number gains on 5p15, 6p11-q11, 7p12, and 8q24 are rare in sputum cells of individuals at high risk of lung cancer.
        Lung Cancer. 2006; 54: 169-176
        • Li R.
        • Todd N.W.
        • Qiu Q.
        • et al.
        Genetic deletions in sputum as diagnostic markers for early detection of stage I non-small cell lung cancer.
        Clin Cancer Res. 2007; 13: 482-487
        • Kang J.U.
        • Koo S.H.
        • Kwon K.C.
        • Park J.W.
        • Jung S.S.
        Gain of the EGFR gene located on 7p12 is a frequent and early event in squamous cell carcinoma of the lung.
        Cancer Genet Cytogenet. 2008; 184: 31-37
        • Qiu Q.
        • Todd N.W.
        • Li R.
        • et al.
        Magnetic enrichment of bronchial epithelial cells from sputum for lung cancer diagnosis.
        Cancer. 2008; 114: 275-283
        • Katz R.L.
        • Zaidi T.M.
        • Fernandez R.L.
        • et al.
        Automated detection of genetic abnormalities combined with cytology in sputum is a sensitive predictor of lung cancer.
        Mod Pathol. 2008; 21: 950-960
        • Jiang F.
        • Todd N.W.
        • Qiu Q.
        • Liu Z.
        • Katz R.L.
        • Stass S.A.
        Combined genetic analysis of sputum and computed tomography for noninvasive diagnosis of non-small-cell lung cancer.
        Lung Cancer. 2009; 66: 58-63
        • Jiang F.
        • Todd N.W.
        • Li R.
        • Zhang H.
        • Fang H.
        • Stass S.A.
        A panel of sputum-based genomic marker for early detection of lung cancer.
        Cancer Prev Res (Phila). 2010; 3: 1571-1578
        • Guber A.
        • Greif J.
        • Rona R.
        • et al.
        Computerized analysis of cytology and fluorescence in situ hybridization (FISH) in induced sputum for lung cancer detection.
        Cancer Cytopathol. 2010; 118: 269-277
        • Pepe M.S.
        • Feng Z.
        • Janes H.
        • Bossuyt P.M.
        • Potter J.D.
        Pivotal evaluation of the accuracy of a biomarker used for classification or prediction: standards for study design.
        J Natl Cancer Inst. 2008; 100: 1432-1438
        • Kennedy T.C.
        • Proudfoot S.P.
        • Franklin W.A.
        • et al.
        Cytopathological analysis of sputum in patients with airflow obstruction and significant smoking histories.
        Cancer Res. 1996; 56: 4673-4678
        • Wilson D.O.
        • Weissfeld J.L.
        • Fuhrman C.R.
        • et al.
        The Pittsburgh Lung Screening Study (PLuSS): outcomes within 3 years of a first computed tomography scan.
        Am J Respir Crit Care Med. 2008; 178: 956-961
        • Tammemagi M.C.
        • Katki H.A.
        • Hocking W.G.
        • et al.
        Selection criteria for lung-cancer screening.
        N Engl J Med. 2013; 368: 728-736
        • Youden W.J.
        Index for rating diagnostic tests.
        Cancer. 1950; 3: 32-35
        • Simel D.L.
        • Samsa G.P.
        • Matchar D.B.
        Likelihood ratios with confidence: sample size estimation for diagnostic test studies.
        J Clin Epidemiol. 1991; 44: 763-770
        • Tonon G.
        • Wong K.K.
        • Maulik G.
        • et al.
        High-resolution genomic profiles of human lung cancer.
        Proc Natl Acad Sci U S A. 2005; 102: 9625-9630
        • Weir B.A.
        • Woo M.S.
        • Getz G.
        • et al.
        Characterizing the cancer genome in lung adenocarcinoma.
        Nature. 2007; 450: 893-898
        • Hammerman P.S.
        • Lawrence M.S.
        • Cancer Genome Atlas Research Network
        Comprehensive genomic characterization of squamous cell lung cancers.
        Nature. 2012; 489: 519-525
        • Masuda A.
        • Takahashi T.
        Chromosome instability in human lung cancers: possible underlying mechanisms and potential consequences in the pathogenesis.
        Oncogene. 2002; 21: 6884-6897
        • Silk A.D.
        • Zasadil L.M.
        • Holland A.J.
        • Vitre B.
        • Cleveland D.W.
        • Weaver B.A.
        Chromosome missegregation rate predicts whether aneuploidy will promote or suppress tumors.
        Proc Natl Acad Sci U S A. 2013; 110: E4134-E4141
        • Yoo J.W.
        • Seo K.W.
        • Jang S.J.
        • et al.
        The relationship between the presence of chromosomal instability and prognosis of squamous cell carcinoma of the lung: fluorescence in situ hybridization analysis of paraffin-embedded tissue from 47 Korean patients.
        J Korean Med Sci. 2010; 25: 863-867
        • Danner B.C.
        • Hellms T.
        • Jung K.
        • et al.
        Prognostic value of chromosomal imbalances in squamous cell carcinoma and adenocarcinoma of the lung.
        Ann Thorac Surg. 2011; 92: 1038-1043
        • Thunnissen F.B.
        Sputum examination for early detection of lung cancer.
        J Clin Pathol. 2003; 56: 805-810
        • Hubers A.J.
        • Heideman D.A.
        • Duin S.
        • et al.
        DNA hypermethylation analysis in sputum of asymptomatic subjects at risk for lung cancer participating in the NELSON trial: argument for maximum screening interval of 2 years.
        J Clin Pathol. 2017; 70: 250-254