Advertisement

Interobserver Variation among Pathologists and Refinement of Criteria in Distinguishing Separate Primary Tumors from Intrapulmonary Metastases in Lung

Open ArchivePublished:November 07, 2017DOI:https://doi.org/10.1016/j.jtho.2017.10.019

      Abstract

      Multiple tumor nodules are seen with increasing frequency in clinical practice. On the basis of the 2015 WHO classification of lung tumors, we assessed the reproducibility of the comprehensive histologic assessment to distinguish second primary lung cancers (SPLCs) from intrapulmonary metastases (IPMs), looking for the most distinctive histologic features. An international panel of lung pathologists reviewed a scanned sequential cohort of 126 tumors from 48 patients and recorded an agreed set of histologic features, including tumor typing and predominant pattern of adenocarcinoma, thereby opining whether the case was SPLC, IPM, or a combination thereof. Cohen κ statistics of 0.60 on overall assessment of SPLC or IPM indicated a good agreement. Likewise, there was good agreement (κ score 0.64, p < 0.0001) between WHO histologic pattern in individual cases and SPLC or IPM status, but the proportions diversified for histologic pattern and SPLC or IPM status (McNemar test, p < 0.0001). The strongest associations for distinguishing between SPLC and IPM were observed for nuclear pleomorphism, cell size, acinus formation, nucleolar size, mitotic rate, nuclear inclusions, intraalveolar clusters, and necrosis. Conversely, the associations for lymphocytosis, mucin content, lepidic growth, vascular invasion, macrophage response, clear cell change, acute inflammation keratinization, and emperipolesis did not reach significance with tumor extent. Comprehensive histologic assessment is recommended for distinguishing SPLC from IPM with good reproducibility among lung pathologists. In addition to main histologic type and predominant patterns of histologic subtypes, nuclear pleomorphism, cell size, acinus formation, nucleolar size, and mitotic rate strongly correlate with pathologic staging status.

      Keywords

      Introduction

      Multiple tumor nodules may result from the formation of separate primary lung cancers (SPLCs) or intrapulmonary metastases (IPMs) (also termed clinically as separate tumor nodules). From 1975 until recently, distinctions between these two types of nodules were defined by the criteria of Martini and Melamed. According to these criteria, SPLCs were defined as tumors occurring in different lobes, featuring different major histologic types, or being separated by a time interval of more than 2 years.
      • Martini N.
      • Melamed M.R.
      Multiple primary lung cancers.
      Recently however, these criteria have been supplanted by comprehensive histologic assessment (CHA).
      • Girard N.
      • Deshpande C.
      • Lau C.
      • et al.
      Comprehensive histologic assessment helps to differentiate multiple lung primary nonsmall cell carcinomas from metastases.
      CHA involves determination of major histologic type, assessment of predominant and minor histologic patterns according to histologic subtyping, and evaluation of cytologic features. CHA has been shown to significantly improve the pathologic distinction between SPLC and IPM to a level comparable to that by molecular analysis.
      • Girard N.
      • Deshpande C.
      • Lau C.
      • et al.
      Comprehensive histologic assessment helps to differentiate multiple lung primary nonsmall cell carcinomas from metastases.
      Pathologic staging of lung cancer remains integral to patient management. The recently revised eighth editions of the American Joint Committee on Cancer and Union for International Cancer Control staging manuals
      include updates in the T, N and M components for both NSCC and SCLC.
      • Nicholson A.G.
      • Chansky K.
      • Crowley J.
      • et al.
      The International Association for the Study of Lung Cancer Lung Cancer Staging Project: proposals for the revision of the clinical and pathologic staging of small cell lung cancer in the forthcoming eighth edition of the TNM classification for lung cancer.
      • Eberhardt W.E.
      • Mitchell A.
      • Crowley J.
      • et al.
      The IASLC Lung Cancer Staging Project: proposals for the revision of the M descriptors in the forthcoming eighth edition of the TNM classification of lung cancer.
      • Rami-Porta R.
      • Bolejack V.
      • Crowley J.
      • et al.
      The IASLC Lung Cancer Staging Project: proposals for the revisions of the T descriptors in the forthcoming eighth edition of the TNM classification for lung cancer.
      These updates reflect increased interest in staging of patients with multiple tumor nodules
      • Detterbeck F.C.
      • Bolejack V.
      • Arenberg D.A.
      • et al.
      The IASLC Lung Cancer Staging Project: background data and proposals for the classification of lung cancer with separate tumor nodules in the forthcoming eighth edition of the TNM classification for lung cancer.
      • Detterbeck F.C.
      • Franklin W.A.
      • Nicholson A.G.
      • et al.
      The IASLC Lung Cancer Staging Project: background data and proposed criteria to distinguish separate primary lung cancers from metastatic foci in patients with two lung tumors in the forthcoming eighth edition of the TNM classification for lung cancer.
      • Detterbeck F.C.
      • Marom E.M.
      • Arenberg D.A.
      • et al.
      The IASLC Lung Cancer Staging Project: background data and proposals for the application of TNM staging rules to lung cancer presenting as multiple nodules with ground glass or lepidic features or a pneumonic type of involvement in the forthcoming eighth edition of the TNM classification.
      • Travis W.D.
      • Asamura H.
      • Bankier A.A.
      • et al.
      The IASLC Lung Cancer Staging Project: proposals for coding T categories for subsolid nodules and assessment of tumor size in part-solid tumors in the forthcoming eighth edition of the TNM classification of lung cancer.
      because of the increased frequency of presentation
      • Detterbeck F.C.
      • Bolejack V.
      • Arenberg D.A.
      • et al.
      The IASLC Lung Cancer Staging Project: background data and proposals for the classification of lung cancer with separate tumor nodules in the forthcoming eighth edition of the TNM classification for lung cancer.
      and advances in classification of tumor subtypes.
      • Travis W.D.
      • Asamura H.
      • Bankier A.A.
      • et al.
      The IASLC Lung Cancer Staging Project: proposals for coding T categories for subsolid nodules and assessment of tumor size in part-solid tumors in the forthcoming eighth edition of the TNM classification of lung cancer.
      Although TNM categories for multiple tumor nodules remain the same in the eighth edition as in the seventh, categories have been affected by changes in the histologic classification of lung cancer,
      in particular, adenocarcinomas.
      • Travis W.D.
      • Brambilla E.
      • Noguchi M.
      • et al.
      International Association for the Study of Lung Cancer/American Thoracic Society/European Respiratory Society international multidisciplinary classification of lung adenocarcinoma.
      SPLCs continue to be staged individually, with the recommendation that multiple lesions, including ground glass lesions, be grouped with the number of lesions in brackets (e.g., [2] or [m] for multiple). Patients with IPM are staged as T3 (same lobe), T4 (different lobe in ipsilateral lung), and M1a (contralateral lung). IPMs also include those presenting with multiple areas of pneumonic consolidation that frequently correspond to invasive mucinous adenocarcinomas.
      • Rami-Porta R.
      • Bolejack V.
      • Crowley J.
      • et al.
      The IASLC Lung Cancer Staging Project: proposals for the revisions of the T descriptors in the forthcoming eighth edition of the TNM classification for lung cancer.
      • Travis W.D.
      • Brambilla E.
      • Noguchi M.
      • et al.
      International Association for the Study of Lung Cancer/American Thoracic Society/European Respiratory Society international multidisciplinary classification of lung adenocarcinoma.
      This distinction is key to patient management because classification as SPLCs may determine that the patient be treated with surgical therapy instead of with other therapeutic approaches such as chemotherapy and radiotherapy.
      The ability of pathologists to apply this approach in a reproducible fashion has not been tested, and revisions of histologic classification in the 2015 classification
      potentially allow further refinement of criteria within CHA. We therefore sought to assess the reproducibility of CHA and to identify the most useful histologic features.

      Methods

      To test consistency of opinion among a panel of pathologists on the question of primary versus metastatic lung adenocarcinoma in multiple tumor nodules, pathology members of the multiple nodules subgroup of the International Association for the Study of Lung Cancer Staging and Prognostic Factors Committee and invited participants from the International Association for the Study of Lung Cancer Pathology Committee, a group comprising a global representation of practicing pathologists with specialist interest in lung cancer, assessed whether two or more nodules in each case were SPLCs, IPM, or a combination of both by using an image-linked on-line survey. After review of a training cohort of five cases,
      • Franklin W.A.
      • Nicholson A.G.
      • Torkko K.
      • Travis W.D.
      • Detterbeck F.
      Pilot Internet survey of interobserver variability in pathology diagnoses of multiple tumor nodules.
      one representative slide from each nodule resected from a test cohort of 48 patients was scanned into an image database using an Aperio AT2 slide scanner (Leica Biosystems) with an ×40 objective. The cases comprised a sequential series of cases from 2013 and 2014 with multiple resected nodules identified within the diagnostic archive of Royal Brompton Hospital. Digitized images were linked to specific questions via a URL, and participants filled in their answers online until the survey was completed. A sample form is shown in Figure 1. Participants were aware of the site of each nodule but not the nodal status of the case, and no immunohistochemical or molecular data were provided.
      Figure thumbnail gr1
      Figure 1A representative image from the sample survey form.
      A Red Cap database with online survey capabilities was created for data entry and tracking at the University of Colorado. The database incorporated a registration (demographic) survey and a case-by-case password-protected review survey. Participants provided a histologic diagnosis for each nodule, with additional subtyping of adenocarcinomas according to the 2015 WHO classification of lung tumors
      and scoring of a detailed set of histologic parameters. The criteria for CHA were based primarily on those of Girard et al.,
      • Girard N.
      • Deshpande C.
      • Lau C.
      • et al.
      Comprehensive histologic assessment helps to differentiate multiple lung primary nonsmall cell carcinomas from metastases.
      with additional parameters proposed at meetings reviewing the initial training study. These included the presence or absence of features such as lepidic growth or similarities of morphologic features in comparing tumors side by side, such as cell size, mitotic rate, and nuclear pleomorphism (Table 1).
      Table 1Histologic Features Assessed by Survey Participants, Including Responses to Relevant Question
      Histological FeatureResponses
      Histologic diagnosis 1WHO code and descriptor
      Histologic diagnosis 2WHO code and descriptor
      Lepidic growthPresent/absent in each tumor
      Gland/acinus formationPresent/absent in each tumor
      Intraalveolar clustersPresent/absent in each tumor
      Compare cell size difference between tumorsSimilar/significantly different
      Compare mitotic rates between tumorsSimilar/significantly different
      Compare degree of nuclear pleomorphism between tumorsSimilar/significantly different
      Nucleolar prominence and appearanceSimilar/significantly different
      Nuclear inclusionsSimilar/significantly different
      Necrosis patternPresent/absent in each tumor
      Vascular invasionPresent/absent in each tumor
      Mucin contentSimilar/significantly different
      KeratinizationSimilar/significantly different
      Clear cell changeSimilar/significantly different
      Cytoplasmic granulesSimilar/significantly different
      LymphocytosisSimilar/significantly different
      Macrophage responseSimilar/significantly different
      Acute inflammationPresent/absent in each tumor
      EmperipolesisPresent/absent in each tumor
      Other distinguishing microscopic featuresFree text
      Comparing tumor 2 with tumor 1, is tumor 2 metastatic or primary?Primary/metastatic
      The study was approved within the Royal Brompton and Harefield National Health Service Foundation Trust as a service evaluation study in the context of assessing the reproducibility of pathologists in applying current staging and histologic criteria.

      Statistical Analysis

      The submitting diagnosis was regarded as the criterion standard for determination of metastatic status of each nodule because the submitting institution had access to clinical history, imaging data, and immunohistochemical results, which contributed final determination of metastatic status. WHO classifications and numbers of each tumor type as provided by the submitting institution are shown in Table 2. Level of agreement regarding WHO classification criteria and histologic patterns among pathologists and with the submitting diagnosis was assessed by κ statistics. Agreement between histologic pattern and SPLC/IPM status was evaluated by the Cohen κ score and McNemar test. Cases with more than two tumors in which pathologists considered that IPMs and SPLCs were both present (3 of 17 cases) were categorized as IPM for the purposes of analysis, as IPM would indicate higher-stage disease.
      Table 2Histologic Diagnoses of Submitting Institution
      Tumor Types (Reference Diagnosis)WHO CodeNo.
      Acinar adenocarcinoma8551/331
      Adenocarcinoma, NOS8140/32
      Adenocarcinoma in situ, nonmucinous8410/26
      Atypical adenomatous hyperplasia8250/02
      Colloid adenocarcinoma8480/31
      Invasive mucinous adenocarcinoma8253/315
      Keratinizing squamous cell carcinoma8071/34
      Large cell carcinoma8012/37
      Lepidic adenocarcinoma8250/35
      Micropapillary adenocarcinoma8265/37
      Minimally invasive adenocarcinoma, nonmucinous8250/27
      Mixed invasive mucinous and nonmucinous adenocarcinoma8254/36
      Nodule 2 too small to classify1
      Nodule 3 too small to see many features1
      Nonkeratinizing squamous cell carcinoma8072/39
      Other (hamartoma, carcinoid, etc.)3
      Papillary adenocarcinoma8260/34
      Small cell carcinoma8041/35
      Solid adenocarcinoma8230/315
      Total131
      NOS, not otherwise specified.
      To determine which histologic features weighed most heavily in arriving at a final determination of SPLC/IPM status, we studied the independent effects of each question posed in the survey. The results for each pathologist were summed across all tumor pairs for each question. We included data from all pathologists (N = 17 with an effective sample size of n = 16, as one scorer did not complete the survey for all 48 tumor pairs). We included only conclusions of SPLC or IPM and excluded instances in which the pathologist entered “inconclusive” (Table 3). Data were reduced to 2 × 2 tables in which findings were categorized as similar versus dissimilar, none/inconsistent versus present in both or all, etc., and the conclusions were primary or metastasis. Differences in proportions in these 2 × 2 tables were determined by using Fisher’s exact tests. The number of pathologists with significant differences in findings by histologic feature (Fisher’s exact test p < 0.05) was tabulated for each category (Table 4). The findings for each pathologic feature were also tested by using a binomial test to determine whether the number was different from 50-50 (i.e, from the result of a coin toss or, in other words, from chance alone). For example, 14 of 16 or 10 of 16 was tested against eight of 16. A p value less than 0.05 indicated that the number of pathologists with a significant difference is different from chance (half the pathologists found a difference and half did not).
      Table 3Level of Agreement of Each Participant with Reference Diagnosis in Regard to the Overall Question of Primary versus Metastatic Tumor
      ReviewerUncertain
      Percentage of 47.
      Missing
      Percentage of 47.
      Reviewer and Ref AgreeReviewer and Ref DisagreeAgreeDisagreeκMcNemar's
      Both PrimaryBoth MetRev:Prim/Ref:MetRev:Met/Ref:Prim
      n%n%n%n%n%n%%%Valuep Value
      Tests that Kappa is not = 0; p<0.05 means that there is more than zero agreement.
      p Value
      McNemar's p<0.05 means that the reviewer did not agree with the reference to a level that reaches statistical significance.
      1510.6%00%1228.6%2457.1%37.1%37.1%85.7%14.3%0.689<0.00011.00
      212.1%00%1123.9%2656.5%36.5%613.0%80.4%19.6%0.5640.00010.317
      300.0%00%1531.9%2553.2%48.5%36.4%85.1%14.9%0.688<0.00010.705
      400.0%00%919.1%2859.6%12.1%919.1%78.7%21.3%0.5080.00020.011
      512.1%00%1430.4%2758.7%24.3%36.5%89.1%10.9%0.764<0.00010.655
      648.7%12%1023.8%2559.5%12.4%614.3%83.3%16.7%0.624<0.00010.059
      736.4%00%1431.8%2556.8%24.5%36.8%88.6%11.4%0.758<0.00010.655
      812.2%12%920.0%2453.3%36.7%920.0%73.3%26.7%0.4120.0040.083
      900.0%00%1634.0%2655.3%36.4%24.3%89.4%10.6%0.777<0.00010.655
      1024.3%00%1431.1%2657.8%12.2%48.9%88.9%11.1%0.762<0.00010.180
      11817.0%00%717.9%2153.8%37.7%820.5%71.8%28.2%0.3640.0170.132
      1224.3%00%1328.9%2146.7%613.3%511.1%75.6%24.4%0.4950.0010.763
      131225.5%00%1440.0%1542.9%514.3%12.9%82.9%17.1%0.661<0.00010.102
      14510.6%00%1126.2%2559.5%37.1%37.1%85.7%14.3%0.679<0.00011.00
      1524.3%00%817.8%2760.0%24.4%817.8%77.8%22.2%0.4710.0010.058
      1648.5%00%1637.2%1227.9%1432.6%12.3%65.1%34.9%0.3560.0050.001
      1700.0%3370%428.6%750.0%17.1%214.3%78.6%21.4%0.5530.0360.56
      Mean81.2%18.8%0.596
      Ref, reference; Met, metastatic; Rev, reviewer; Prim, primary.
      a Percentage of 47.
      b Tests that Kappa is not = 0; p<0.05 means that there is more than zero agreement.
      c McNemar's p<0.05 means that the reviewer did not agree with the reference to a level that reaches statistical significance.
      Table 4Scoring of Histologic Features
      Histologic FeatureFindingsScore for FeatureSPLC/IPM Status
      n
      Based on a theoretical total of 816 total findings for each of 48 tumor pairs for 17 pathologists. Not all pathologists answered every question. Does not include inconclusive findings.
      %SPLCIPM
      Strong correlation with SPLC/IPM status
       Gland/acinus formationDissimilar32646.7%68.4%31.6%
      Similar33553.3%9.9%90.1%
       Cell sizeVariable size29241.4%75.3%24.7%
      Similar size41458.6%8.7%91.3%
       MitosesDifferent mitoses22031.0%71.8%28.2%
      Similar mitoses49069.0%19.8%80.2%
       Nuclear pleomorphismDifferent level24638.3%74.8%25.2%
      Similar level39761.7%12.1%87.9%
       Nucleolar appearancesDifferent29141.2%75.6%24.4%
      Similar41558.8%8.2%91.8%
      Noninformative
       Lepidic growthInconsistent20627.9%52.4%47.6%
      Consistent37772.1%29.1%70.9%
       Alveolar clustersNo or inconsistent clusters51471.7%41.1%58.9%
      Consistent clusters20328.3%24.6%75.4%
       Necrosis patternDissimilar22032.0%63.6%36.4%
      Similar46768.0%23.3%76.2%
       Mucin contentDifferent16523.8%67.9%32.1%
      Similar52876.2%27.5%72.5%
       LymphocytosisDissimilar20229.4%65.3%34.7%
      Similar48570.6%23.9%76.1%
       Macrophage responseDissimilar17124.9%64.9%35.1%
      Similar51675.1%26.7%73.3%
       NeutrophilsDissimilar in both/all tumors13920.1%48.9%51.1%
      Similar in both/all tumors55279.9%33.2%66.8%
       Vascular invasionNone/inconsistent59287.6%39.9%60.1%
      Present in both/all tumors8412.4%15.5%84.5%
      Noninformative with one finding uncommon (<10%)
       Nuclear inclusionsNone/inconsistent66594.7%35.9%64.1%
      Present in both/all tumors375.3%37.8%62.2%
       KeratinizationNone/inconsistent65993.9%37.8%62.2%
      Present in both/all tumors436.1%2.3%97.7%
       Clear cell changeNone/inconsistent66495.8%37.3%62.7%
      Present in both/all tumors294.2%10.3%89.7%
       GranulesNone/inconsistent58791.6%39.2%60.8%
      Present in both/all tumors458.4%16.7%83.3%
       EmperipolesisNone/inconsistent69299.6%36.4%63.6%
      Present in both/all tumors30.4%33.3%66.7%
      SPLC, separate primary lung cancer; IPM, intrapulmonary metastasis.
      a Based on a theoretical total of 816 total findings for each of 48 tumor pairs for 17 pathologists. Not all pathologists answered every question. Does not include inconclusive findings.
      All analyses were performed with SAS software (version 9.4, SAS Institute, Cary, NC) and GraphPad Prism (version 7.01, GraphPad Software, La Jolla, CA). All statistical tests were two sided, with a p value less than 0.05 determining a statistically significant finding.

      Results

      The overall results are summarized in a consensus diagram (Fig. 2) with cases sorted horizontally according to the number of nodules evaluated in each case. Sixteen pathologists completed the survey (21 questions), evaluating 126 tumors from 48 patients and generating more than 42,000 data points. One pathologist partially completed the survey. Excluding uncertain diagnoses, there were four cases that were unanimously scored as SPLC and 10 as IPM. In a further eight cases, a large majority of pathologists (at least 12 of 16) classified multiple nodules as SPLCs. Of these, six showed different major histologic types and six showed multiple histologic subtypes of ADC, two with more than two nodules.
      Figure thumbnail gr2
      Figure 2In this consensus map, determination for second primary lung cancer (SPLC)/intrapulmonary metastasis (IPM) status is shown for each tumor sorted horizontally by number of tumors and all slide readers. As indicated in the first row, 31 cases with two tumors and 17 cases with more than two tumors were assessed. In four cases the unanimous opinion was that all nodules were SPLC, and 10 cases were unanimously determined to be metastases. In all but six cases most panelists agreed with the submitting opinion. In one case, the majority opinion was that a diagnosis could not be reached. Ref, reference; Dx, diagnosis.
      There was strong agreement between SPLC/IPM status among panelists and reference diagnosis, with overall agreement of 81% (κ 0.596) (Table 3). There was also good agreement as measured by the Cohen κ statistic (0.64, p < 0.0001) between WHO histologic pattern and SPLC/IPM status, but the relationship between histologic pattern and SPLC or IPM status was not identical (McNemar test, p < 0.0001). Figure 3, Figure 4, Figure 5, Figure 6 show examples of cases for which there were varying levels of agreement among those polled, as indicated in the figure legends.
      Figure thumbnail gr3
      Figure 3(Case 2 from .) Unanimous opinion of second primary lung cancer. Differences in architecture and nuclear appearances are evident. (A) The tumor is a solid adenocarcinoma. Nuclei show homogenous chromatin without nucleoli. (B) The tumor is an acinar adenocarcinoma. Tumor cells show vesicular chromatin and prominent nucleoli.
      Figure thumbnail gr4
      Figure 4(Case 33 from .) Unanimous opinion of intrapulmonary metastasis. Similar nuclear features (large nuclear size and prominent nucleoli) and acinar predominant patterns were present in multiple nodules (A and B), although a solid region was also encountered in one nodule (C).
      Figure thumbnail gr5
      Figure 5(Case 13 from .) Only 47% agreement regarding second primary lung cancer/intrapulmonary metastasis status despite similarities in (acinar) pattern and nuclear configuration being observed. Reasons for the mixed opinion in this case are not known.
      Figure thumbnail gr6
      Figure 6(Case 32 from .) Multiple (three) tumors are included. (A) Tumor shows a solid pattern. (B) Tumor shows a papillary and acinar pattern. (C) Tumor shows an acinar pattern. Nuclear features also differ. There was 100% agreement on a second and third primary lung cancers in this case.
      To analyze components of CHA, the summed data from the tumor pairs for each question for each pathologist were tested for differences in proportions using the Fisher’s exact test (Table 4). Those with a p value less than 0.05 were considered to have found a difference in histologic feature that contributed to an SPLC or IPM conclusion. In addition, the number of pathologists with significant differences was summed and tested against the hypothesis that this number is different from the result of a coin toss (Table 5). For example, acinus formation significantly correlated with SPLC/IPM status for 15 of 16 pathologists (94%). A binomial test was used to determine whether 15 of 16 was different from 8 of 16, where 8 of 16 represents the expected proportion that was not different from a the result of a coin toss. The p value for this test was 0.0005, indicating that the number of pathologists who found a difference in proportions is better than random chance.
      Table 5Fisher's Exact Test Results for Histologic Features Determined by Each Pathologist versus SPT/IMP Status
      Histologic feature% of Pathologists Finding Differences in Feature by SPLC/IPM StatusBinomial p Value
      Informative
       Cell size100% (16 of 16)<0.0001
       Nucleolar appearances100% (16 of 16)<0.001
       Mitoses87% (14 of 16)0.004
       Nuclear pleomorphism93% (14 of 15)0.001
       Gland/acinus formation94% (15 of 16)0.0005
      Noninformative
       Alveolar clusters19% (3 of 16)0.02
       Vascular invasion25% (4 of 16)0.08
       Lepidic growth29% (5 of 17)0.14
       Neutrophils37% (6 of 16)0.45
       Mucin content44% (7 of 16)0.80
       Necrosis pattern56% (9 of 16)0.80
       Macrophage response62% (10 of 16)0.79
       Lymphocytosis69% (11 of 16)0.21
      Noninformative with one finding uncommon (<10%)
       Cytoplasmic granules27% (4 of 15)0.12
       Nuclear inclusions37% (6 of 16)0.45
       Keratinization44% (7 of 16)0.80
       Clear cell change44% (7 of 16)0.80
       Emperipolesis56% (9 of 16)0.80
      SPLC, separate primary lung cancer; IPM, intrapulmonary metastasis.
      NOTE: Items in boldface are p<0.05. Based on number of Fisher's exact p values less than 0.05 calculated from the proportion differences of primary/metastases by finding for each pathologist.
      Table 6 shows the numbers of pathologists who found significant differences for each finding by conclusion regarding primary or metastasis. For example, looking at the finding of “similar” for gland/acinus formation in the “difference by finding” columns, 14 out of 16 (88%) pathologists found that there was a difference in the distribution of conclusions of primary or metastatic; that is, they found that similar acinus formation was more likely to be associated with a conclusion of metastatic tumor. A binomial test was used to determine whether the number of pathologists was different from the result of a coin toss. A significant p value in cases in which the number of significant findings was small (e.g., three of 16 or zero of 17) was thought to be noninformative in choosing between primary or metastatic tumors, as most of the pathologists found no difference.
      Table 6Differences by Histologic Feature
      Histologic FeatureFindings% Differences in Feature by SPLC/IPM statusBinomial p Value
      Informative
       Gland/acinus formationDissimilar44% (7 of 16)0.804
      Similar88% (14 of 16)0.004
       Cell sizeVariable size56% (9 of 16)0.804
      Similar size81% (13 of 16)0.021
       MitosesDifferent mitoses37% (6 of 16)0.454
      Similar mitoses94% (15 of 16)0.0005
       Nuclear pleomorphismDifferent level53% (8 of 15)0.999
      Similar level87% (13 of 15)0.007
       Nucleolar appearancesDifferent56% (9 of 16)0.804
      Similar94% (15 of 16)0.0005
      Noninformative
       Lepidic growthInconsistent0% (0 of 17)<0.0001
      Consistent59% (10 of 17)0.629
       Alveolar clustersNo or inconsistent clusters19% (3 of 16)0.021
      Consistent clusters69% (11 of 16)0.210
       Necrosis patternDissimilar19% (3 of 16)0.021
      Similar81% (13 of 16)0.021
       Mucin contentDifferent12% (2 of 16)0.004
      Similar75% (12 of 16)0.077
       LymphocytosisDissimilar31% (5 of 16)0.210
      Similar81% (13 of 16)0.021
       Macrophage responseDissimilar21% (3 of 14)0.057
      Similar75% (12 of 16)0.077
       NeutrophilsDissimilar in both/all tumors14% (2 of 14)0.013
      Similar in both/all tumors62% (10 of 16)0.454
       Vascular invasionNone/inconsistent25% (4 of 16)0.077
      Present in both/all tumors17% (2 of 12)0.039
      Noninformative with one finding uncommon (<10%)
       Nuclear inclusionsNone/inconsistent37% (6 of 16)0.454
      Present in both/all tumors10% (1 of 10)0.012
       KeratinizationNone/inconsistent44% (7 of 16)0.804
      Present in both/all tumors0% (0 of 11)0.001
       Clear cell changeNone/inconsistent44% (7 of 16)0.804
      Present in both/all tumors0% (0 of 7)0.016
       GranulesNone/inconsistent27% (4 of 15)0.118
      Present in both/all tumors40% (2 of 5)0.999
       EmperipolesisNone/inconsistent50% (8 of 16)0.999
      Present in both/all tumorsNA
      NOTE: Boldface indicates statistically significance (p < 0.05). Based on number of binomial p values less than 0.05 for a difference in proportions of responses from a coin toss. Calculated from the proportion differences in metastasis/primary for each finding for each pathologist. The p value here is from a binomial test that determines whether the proportion of pathologists with a positive difference is different from a coin toss (i.e., 50%–50%).
      SPLC, separate primary lung cancer; IPM, intrapulmonary metastasis; NA, not available.

      Discussion

      Our study demonstrates that a panel of practicing pulmonary pathologists has good agreement in using CHA to distinguish SPLC from IPM. This supports the concept that morphology is a powerful tool in addressing this problem.
      Multiple tumor nodules are an increasingly frequent finding in clinical practice and highlight the complex biology of lung cancer, especially adenocarcinoma of the lung. Although historically the default staging approach has been to view multiple tumor nodules in the lung as IPM and to stage them as T3 if present in the same lobe, T4 if present in a different ipsilateral lobe, and M1a if present in contralateral lobes, there is increasing recognition that multiple tumor nodules may represent SPLCs. SPLCs, especially those with ground glass appearances on imaging, have been associated with improved survival in comparison with IPMs,
      • Detterbeck F.C.
      • Franklin W.A.
      • Nicholson A.G.
      • et al.
      The IASLC Lung Cancer Staging Project: background data and proposed criteria to distinguish separate primary lung cancers from metastatic foci in patients with two lung tumors in the forthcoming eighth edition of the TNM classification for lung cancer.
      • Detterbeck F.C.
      • Marom E.M.
      • Arenberg D.A.
      • et al.
      The IASLC Lung Cancer Staging Project: background data and proposals for the application of TNM staging rules to lung cancer presenting as multiple nodules with ground glass or lepidic features or a pneumonic type of involvement in the forthcoming eighth edition of the TNM classification.
      and determination of SPLC/IPM status will be important in new staging algorithms and clinical trials.
      • Bosman F.T.
      • Lehr H.A.
      Quality in pathology: how good is good? An introduction.
      However, it is noticeable that agreement was far from perfect, and consideration should be given to routine application of immunohistochemical and molecular testing to buttress histologic analysis. Studies of SPLCs report tumors either with EGFR mutations in only one tumor
      • Arai J.
      • Tsuchiya T.
      • Oikawa M.
      • et al.
      Clinical and molecular analysis of synchronous double lung cancers.
      or differing EGFR and KRAS mutations,
      • Girard N.
      • Deshpande C.
      • Azzoli C.G.
      • et al.
      Use of epidermal growth factor receptor/Kirsten rat sarcoma 2 viral oncogene homolog mutation testing to define clonal relationships among multiple lung adenocarcinomas: comparison with clinical guidelines.
      and comparative genomic analysis has shown a high level of concordance for recurrent alterations at first relapse in IPMs,
      • Vignot S.
      • Frampton G.M.
      • Soria J.C.
      • et al.
      Next-generation sequencing reveals high concordance of recurrent somatic alterations between primary tumor and metastases from patients with non-small-cell lung cancer.
      although the numbers in these studies are small. Others report that although comprehensive genotypic and morphologic assessment is feasible, it is not sufficient to establish their clonal relationship.
      • Schneider F.
      • Derrick V.
      • Davison J.M.
      • Strollo D.
      • Incharoen P.
      • Dacic S.
      Morphological and molecular approach to synchronous non-small cell lung carcinomas: impact on staging.
      Indeed, further studies are needed to assess this challenging problem by using a multidisciplinary approach involving clinical (including survival data), radiologic, and immunohistochemical features and molecular data, as well as histologic features. Such an approach may be additionally informative in future staging studies and clinical trials when this issue is being addressed.
      In addition, these data provide background information in creating machine algorithms for cellular diagnosis that are currently under development.
      • Yu K.H.
      • Zhang C.
      • Berry G.J.
      • et al.
      Predicting non-small cell lung cancer prognosis by fully automated microscopic pathology image features.
      • Luo X.
      • Zang X.
      • Yang L.
      • et al.
      Comprehensive computational pathological image analysis predicts lung cancer prognosis.
      The data suggest that concentration on nuclear features may be a way forward in the development of image analysis tests to identify tumor cells in mixed cell populations and support diagnosis of malignancy, dysplasia, second primary, or metastases.
      This study not only examined agreement on the judgment exercised by a panel of experts in regard to metastatic status but also evaluated those microscopic features that go into making this determination. Similar acinar patterns, cell size, mitotic rate, nuclear pleomorphism, and nucleolar appearances were significantly associated with the final decision regarding SPLC/IPM status, whereas alveolar clusters, cytoplasmic granules, lepidic growth, neutrophils, and vascular invasion were not. In parsing these data further, the statistical strength of these associations can be attributed to the direction of specific properties of a tumor. For example, similar acinus formation was associated with IPM but differing patterns of acinus formation did not associate with SPLC. When evaluated in this way, consistent cell size, similar mitotic rate, similar nuclear pleomorphism, and similar nucleolar abnormalities were all closely associated with metastasis, but the obverse was not true and inconsistency of these features was not associated with SPLC. Inconsistency of lepidic growth pattern, presence of alveolar clusters, necrosis pattern, mucin content, and neutrophil presence were associated with SPLC, whereas consistency of necrosis pattern and vascular invasion was associated with metastasis. In several categories, insufficient numbers of response were obtained to make valid comparisons and the survey responses to these questions were determined to be noninformative.
      There are several limitations to our study, including the lack of clinical follow-up, immunohistochemistry, and molecular testing results and our inability to determine the combined effects of the pathologic categories. It was not possible to do a meaningful multivariate analysis to study the combined effects of the different findings from different pathologic categories as the effective sample size was only 16 pathologists. There were also too many pathologic categories compared with the number of pathologists, which limits the power for a multivariate analysis. By studying the categories individually, we were not able to ascertain interactions between findings and conclusions by category. For example, even though we found that similar acinus formation was associated with a conclusion of metastasis, we could not state that similar acinus formation alone was sufficient for the pathologist to conclude that a second tumor was a metastasis of the first tumor. The pathologist undoubtedly takes a more general view of several pathologic findings to reach a conclusion regarding SPLC/IPM status. Despite this limitation, this study still provides an interesting picture of the importance of several nuclear features taken together relative to the other categories.
      We conclude that CHA shows good reproducibility between practicing lung pathologists. In addition to main tumor type and predominant patterns, nuclear pleomorphism, cell size, acinus formation, nucleolar size, and mitotic rate strongly correlated with pathologic staging status. Inconsistency between other features, such as lepidic growth pattern, also likely contribute to the reductionist evaluation that pathologists undertake when assessing multiple tumor nodules (Fig. 7). How this correlates with molecular analysis is a subject for further study.
      Figure thumbnail gr7
      Figure 7Comprehensive histologic assessment: stepwise assessment of tumor type, histologic patterns in adenocarcinoma, and cytologic features allows more accurate and reproducible distinction between second primary lung cancers and intrapulmonary metastases (separate tumor nodules). ADC, adenocarcinoma; LCC, large cell carcinoma; LCNEC, large cell neuroendocrine carcinoma; SQCC, squamous cell carcinoma.
      Adapted with permission from Girard et al.
      • Girard N.
      • Deshpande C.
      • Lau C.
      • et al.
      Comprehensive histologic assessment helps to differentiate multiple lung primary nonsmall cell carcinomas from metastases.

      References

        • Martini N.
        • Melamed M.R.
        Multiple primary lung cancers.
        J Thorac Cardiovasc Surg. 1975; 70: 606-612
        • Girard N.
        • Deshpande C.
        • Lau C.
        • et al.
        Comprehensive histologic assessment helps to differentiate multiple lung primary nonsmall cell carcinomas from metastases.
        Am J Surg Pathol. 2009; 33: 1752-1764
      1. Amin M.B. Edge S. Greene F. AJCC Cancer Staging Manual. 8th ed. Springer International Publishing, New York, NY2017
      2. Brierley J.D. Gospodarowicz M.K. Wittekind C. UICC TNM Classification of Malignant Tumours. 8th ed. Wiley Blackwell, Oxford, UK2017
        • Nicholson A.G.
        • Chansky K.
        • Crowley J.
        • et al.
        The International Association for the Study of Lung Cancer Lung Cancer Staging Project: proposals for the revision of the clinical and pathologic staging of small cell lung cancer in the forthcoming eighth edition of the TNM classification for lung cancer.
        J Thorac Oncol. 2016; 11: 300-311
        • Eberhardt W.E.
        • Mitchell A.
        • Crowley J.
        • et al.
        The IASLC Lung Cancer Staging Project: proposals for the revision of the M descriptors in the forthcoming eighth edition of the TNM classification of lung cancer.
        J Thorac Oncol. 2015; 10: 1515-1522
        • Rami-Porta R.
        • Bolejack V.
        • Crowley J.
        • et al.
        The IASLC Lung Cancer Staging Project: proposals for the revisions of the T descriptors in the forthcoming eighth edition of the TNM classification for lung cancer.
        J Thorac Oncol. 2015; 10: 990-1003
        • Detterbeck F.C.
        • Bolejack V.
        • Arenberg D.A.
        • et al.
        The IASLC Lung Cancer Staging Project: background data and proposals for the classification of lung cancer with separate tumor nodules in the forthcoming eighth edition of the TNM classification for lung cancer.
        J Thorac Oncol. 2016; 11: 681-692
        • Detterbeck F.C.
        • Franklin W.A.
        • Nicholson A.G.
        • et al.
        The IASLC Lung Cancer Staging Project: background data and proposed criteria to distinguish separate primary lung cancers from metastatic foci in patients with two lung tumors in the forthcoming eighth edition of the TNM classification for lung cancer.
        J Thorac Oncol. 2016; 11: 651-665
        • Detterbeck F.C.
        • Marom E.M.
        • Arenberg D.A.
        • et al.
        The IASLC Lung Cancer Staging Project: background data and proposals for the application of TNM staging rules to lung cancer presenting as multiple nodules with ground glass or lepidic features or a pneumonic type of involvement in the forthcoming eighth edition of the TNM classification.
        J Thorac Oncol. 2016; 11: 666-680
        • Travis W.D.
        • Asamura H.
        • Bankier A.A.
        • et al.
        The IASLC Lung Cancer Staging Project: proposals for coding T categories for subsolid nodules and assessment of tumor size in part-solid tumors in the forthcoming eighth edition of the TNM classification of lung cancer.
        J Thorac Oncol. 2016; 11: 1204-1223
      3. Travis W.D. Brambilla E. Burke A.P. Marx A. Nicholson A.G. WHO Classification of Tumours of the Lung, Pleura, Thymus and Heart. 4th ed. France IARC Press, Lyon2015
        • Travis W.D.
        • Brambilla E.
        • Noguchi M.
        • et al.
        International Association for the Study of Lung Cancer/American Thoracic Society/European Respiratory Society international multidisciplinary classification of lung adenocarcinoma.
        J Thorac Oncol. 2011; 6: 244-285
        • Franklin W.A.
        • Nicholson A.G.
        • Torkko K.
        • Travis W.D.
        • Detterbeck F.
        Pilot Internet survey of interobserver variability in pathology diagnoses of multiple tumor nodules.
        J Thorac Oncol. 2015; 10: S455
        • Bosman F.T.
        • Lehr H.A.
        Quality in pathology: how good is good? An introduction.
        Virchows Arch. 2016; 468: 1-4
        • Arai J.
        • Tsuchiya T.
        • Oikawa M.
        • et al.
        Clinical and molecular analysis of synchronous double lung cancers.
        Lung Cancer. 2012; 77: 281-287
        • Girard N.
        • Deshpande C.
        • Azzoli C.G.
        • et al.
        Use of epidermal growth factor receptor/Kirsten rat sarcoma 2 viral oncogene homolog mutation testing to define clonal relationships among multiple lung adenocarcinomas: comparison with clinical guidelines.
        Chest. 2010; 137: 46-52
        • Vignot S.
        • Frampton G.M.
        • Soria J.C.
        • et al.
        Next-generation sequencing reveals high concordance of recurrent somatic alterations between primary tumor and metastases from patients with non-small-cell lung cancer.
        J Clin Oncol. 2013; 31: 2167-2172
        • Schneider F.
        • Derrick V.
        • Davison J.M.
        • Strollo D.
        • Incharoen P.
        • Dacic S.
        Morphological and molecular approach to synchronous non-small cell lung carcinomas: impact on staging.
        Mod Pathol. 2016; 29: 735-742
        • Yu K.H.
        • Zhang C.
        • Berry G.J.
        • et al.
        Predicting non-small cell lung cancer prognosis by fully automated microscopic pathology image features.
        Nat Commun. 2016; 7: 12474
        • Luo X.
        • Zang X.
        • Yang L.
        • et al.
        Comprehensive computational pathological image analysis predicts lung cancer prognosis.
        J Thorac Oncol. 2017; 12: 501-509