The science behind the 7th edition Tumour, Node, Metastasis staging system for lung cancer



    Corresponding author
    1. Department of Thoracic Medicine, The Prince Charles Hospital
    2. UQ Thoracic Research Centre, School of Medicine, The University of Queensland, Brisbane, Queensland, Australia
    Search for more papers by this author
    • The first two authors contributed equally to this paper.


    1. Department of Thoracic Medicine, The Prince Charles Hospital
    2. UQ Thoracic Research Centre, School of Medicine, The University of Queensland, Brisbane, Queensland, Australia
    Search for more papers by this author
    • The first two authors contributed equally to this paper.


    1. Department of Thoracic Medicine, The Prince Charles Hospital
    2. UQ Thoracic Research Centre, School of Medicine, The University of Queensland, Brisbane, Queensland, Australia
    Search for more papers by this author
  • IAN A. YANG,

    1. Department of Thoracic Medicine, The Prince Charles Hospital
    2. UQ Thoracic Research Centre, School of Medicine, The University of Queensland, Brisbane, Queensland, Australia
    Search for more papers by this author

    1. Department of Thoracic Medicine, The Prince Charles Hospital
    2. UQ Thoracic Research Centre, School of Medicine, The University of Queensland, Brisbane, Queensland, Australia
    Search for more papers by this author

  • The Authors: Dr Steven Leong, MBBS (Hons), FRACP, is a Thoracic and Sleep Physician with an interest in early lung cancer diagnosis and advanced bronchoscopy. He is currently a research fellow at The Prince Charles Hospital, Brisbane. Dr Henry Marshall, BM, MRCP, FRACP is a Thoracic Physician at The Prince Charles Hospital with a special interest in thoracic malignancy and an NHMRC Postgraduate Medical Research Scholar at the University of Queensland Thoracic Research Centre. His PhD thesis is based on lung cancer screening using low-dose CT. Dr Ian Yang MBBS (Hons), PhD, FRACP, Grad Dip Clin Epi., is a Consultant Thoracic Physician at The Prince Charles Hospital, Associate Professor and Head of the Northside Clinical School, School of Medicine, The University of Queensland, Brisbane. His clinical work is in the field of thoracic medicine, and his research focuses on gene-environmental interaction in COPD, asthma, lung cancer and air pollution. He is involved in the training of higher degree students in translational research into lung diseases. Associate Professor Rayleen Bowman, MBBS, FRACP, Grad Cert Educ PhD, is clinician-researcher with a major interest in the pathogenesis, diagnosis and treatment of pulmonary and pleural malignancies. As a full-time consultant thoracic physician, she treats many patients with lung cancer and mesothelioma, and is expert in the clinical management of these patients. She has extensive experience in the clinical and molecular aspects of lung cancer and mesothelioma. Her PhD students and Honours students are pursuing topics in these fields. She is currently Chief Investigator A of project grants awarded by the Dust Diseases Board and Cancer Australia. Professor Kwun Fong, MBBS(Lon), FRACP, PhD, is a Thoracic and Sleep Physician at The Prince Charles Hospital, Brisbane and a Professor with the School of Medicine at the University of Queensland, practising general thoracic and sleep medicine with a special interest in thoracic malignancies. He is involved in laboratory based molecular and genomic research and translational trials of diagnosis and treatment in lung cancer and lung diseases. He was the inaugural Chair of the Australasian Lung Cancer Trials Group and serves as Chair of the Australian Lung Foundation's Lung Cancer Consultative Group, Chair of Cancer Australia Lung Cancer Advisory Group, Co-Editor of the Cochrane Lung Cancer Review Group and Associate Editor for the Journal of Thoracic Oncology.


Henry Marshall, The Prince Charles Hospital, Rode Road, Chermside, Brisbane, Qld 4032, Australia. Email:


The Tumour, Node, Metastasis (TNM) system for classifying lung cancer is the cornerstone of modern lung cancer treatment and underpins comparative research; yet is continuously evolving through updated revisions. The recently published Union for International Cancer Control 7th Edition TNM Classification for lung cancer addresses many of its predecessor's shortcomings and has been subject to rigorous evidence-based methodology. It is based on a retrospective analysis of over 80 000 lung cancer patients treated between 1990 and 2000 carried out by the International Association for the Study of Lung Cancer. The dataset was truly international and included patients treated by all modalities. Extensive internal and external validation of the findings has ensured that the recommendations are robust and generalizable. For the first time, a single classification system has been shown to be applicable not only to non-small cell lung cancer, but also to be of prognostic significance in small cell lung cancer and bronchopulmonary carcinoid tumours.

We review the history of the Union for International Cancer Control TNM staging system, the changes in the most recent 7th edition and the strength of the scientific basis motivating these changes. Limitations of the current staging edition are explored, post-publication independent validation studies are reviewed, and the future of TNM staging for lung cancer is discussed.


The Tumour, Node, Metastasis (TNM) classification for staging, a description of the anatomical extent of disease, was developed in the 1940s and has been the cornerstone of lung cancer management since its adoption by the Union for International Cancer Control (UICC).

The objectives of the TNM system are fivefold: to help the clinician plan treatment; guide prognosis; assist in treatment evaluation; provide a ‘common language’ for exchange of information; and contribute to the continuing investigation of human cancer.1

The first UICC ‘handbook’ was published in 1968; the 6th edition (UICC-6) has served as the classification since 2002; the most recent iteration of the ‘Classification of Malignant Tumours’, the 7th edition (UICC-7), was released in January 2009.2 Since 1987 the UICC and American Joint Committee on Cancer (AJCC) have unified their classifications to enable a standard set of definitions worldwide.

From the 2nd through to the 6th edition, the TNM classification was based on pioneering work by Mountain and colleagues, which introduced many of the T, N and M descriptors that we are familiar with today. The original dataset comprised 2155 patients (1712 non-small cell lung cancer (NSCLC) 368 small cell lung cancer (SCLC) and 75 undifferentiated patients), most of whom had been treated surgically.3 Over the years the database grew to include 5319 patients including 623 with SCLC4. Four thousand three hundred fifty-one (83%) of the patients had been treated at The University of Texas–MD Anderson Cancer Centre from 1975 to 1988; the other 968 patients had been treated by the National Cancer Institute Cooperative Lung Cancer Study Group from 1977 to 1982.

Until now, the international lung cancer staging system has been based on this relatively small, unvalidated series of patients treated mainly at a single institution since the 1970s. The gradual shifting of histological subtypes, developments in diagnostic technology and staging (e.g. CT, PET and endobronchial ultrasound), evolving treatments and concerns over the geographic and historical generalizability of the data, especially for non-surgical patients and women, have prompted the medical community to re-evaluate the basis of lung cancer staging.

To this end, the International Staging Project on Lung Cancer was proposed by members of the International Association for the Study of Lung Cancer (IASLC) in 19965 and approved by the IASLC board in 1998. The IASLC, with an international membership representing all disciplines involved in lung cancer care, was well placed to take the lead in developing and analysing an international database of lung cancer patients, which would provide the evidence base for revision of the TNM classification. The Staging Project was undertaken with the co-operation of the UICC, the AJCC and the joint Japanese societies involved in lung cancer and funded by unrestricted grants from Eli Lilly & Co. and the AJCC.

The IASLC convened an International Staging Committee (ISC) to oversee the project. In particular, the ISC wished to address the issues of: (i) validation of individual T, N and M descriptors; (ii) validation of stage groups; and (iii) the generalizability of these findings across geographic location and treatment modality.

During 2007, the IASLC submitted its proposals to the UICC and published a series of papers outlining its findings and recommendations.6–11 The UICC accepted the proposed revisions (Tables 1,2) leading to the 7th edition. This paper will review the findings of the IASLC with particular reference to the underpinning scientific basis.

Table 1.  Summary of 7th edition Tumour, Node, Metastasis classification for lung cancer (includes non-small cell and small cell carcinoma and carcinoid) Thumbnail image of
Table 2.  Tumour, Node, Metastasis classification for lung cancer 7th edition
  1. Overall group stages (includes non-small cell and small cell carcinoma and carcinoid). Changes from the Union for International Cancer Control-6 are in bold (adapted from Sobin et al.2)

Occult carcinomaTXN0M0
IAT1a, bN0M0
T1a, bN1M0
IIIAT1a,b, T2a,bN2M0
T3N1, N2M0
T4N0, N1M0
Any TN3M0
IVAny TAny NM1a or M1b

Data collection and management

The ISC created seven Subcommittees to manage each element of the project: one each for the T, N and M Descriptors; Small-Cell Lung Cancer; a new Nodal Chart; Prognostic Factors; and Validation and Methodology. An analysis of bronchopulmonary carcinoid tumours was also undertaken. The subcommittees collaborated with the Cancer Research and Biostatistics Office, Seattle, Washington, a non-profit organization that had been chosen as the data-management and analysis centre because of its extensive experience in large multicentre studies.

Retrospective staging and outcome data from 46 pre-existing databases in over 19 countries were submitted to the Cancer Research and Biostatistics Office. Data came from three types of database: consortia/surgical series; clinical trials; series/registries. These databases had been developed for reasons other than informing a revision of the TNM classification. Patients were staged in accordance with previous iterations of the TNM classification and the data varied widely in the type and amount of detail provided. All patients had been treated between 1990 and 2000, a timeframe during which there had been no major changes in lung cancer staging and CT had been widely used. It also allowed up to 5 years of follow-up for each patient.

One hundred thousand eight hundred sixty-nine patient details were submitted. Eighty-one thousand fifteen (80.3%) patients met the inclusion criteria and were accepted for analysis. The inclusion criteria were: new cancer diagnosis (not recurrent disease); adequate follow-up for survival calculations; histological subtyping and complete baseline clinical (cTNM) and/or pathological (pTNM) staging. There were 67 725 patients with NSCLC and 13 290 patients with SCLC. Due to the volume of information, an audit of the data integrity was not performed.


All analyses were performed in the NSCLC cohort and then applied to SCLC and carcinoid tumours. Only the characteristics of the NSCLC patients are described here; SCLC and carcinoid populations are described separately.

The patients were drawn from a wide geographic base, originating from: Europe 40 059 (59%); North America 12 178 (18%); Asia 10 216 (15%); Australia 5272 (8%); but not from South America or Africa. Thirty-seven per cent were submitted from consortia/surgical series, 24% from clinical trials and 38% from series/registries. All treatment modalities were represented: 36% were treated by surgery only; 21% with chemotherapy only; 11% with radiotherapy only; 9% received best supportive care or no treatment; and 23% received multimodality treatment. Data reporting on the use of adjuvant treatment for surgical patients was encouraged but not mandatory.12

Median follow-up for the 17 754 patients alive at last contact was 5.3 years. Ninety-five per cent of the NSCLC patients were followed until death or at least 2 years and 88% until death or 5 years.

Of the NSCLC population, 53 640 (79.2%) were clinically staged, mostly with CT of the chest and upper abdomen; details of staging procedures were not generally analysed. Other tests to determine the extent of disease could include mediastinoscopy but not thoracotomy. Thirty-three thousand nine hundred thirty-three (50.1%) were pathologically staged and 20 006 (29.5%) had both clinical and pathological staging. PET data were not available as it was not in widespread use internationally at this time.

Endpoints, statistical methods and validation

Survival, calculated by the Kaplan–Meier method, was the chosen endpoint for each element and measured from date of entry for clinically staged data (i.e. date of diagnosis for registries, date of registration for protocols) and date of surgery for pathologically staged data through to the date of death or last contact. Survival curves were compared visually using all patients analysed and also comparing the subsets of pTNM and cTNM. ‘Unstable’ curves resulting from groups with small numbers of patients were generally not included in order not to skew the final recommendations.7

Each finding was intensively validated. Results were compared across subgroups of database type and geographic region to ensure the direction and magnitude of effects were consistent. Statistical comparisons were not carried out as the numbers often varied widely between data sources, compromising their usefulness.

As the IASLC database was not strictly speaking a population-based dataset, it was open to potential selection bias. The Surveillance, Epidemiology and End Results Program (SEER) database13 was chosen as the population-based dataset for external validation with the aim that IASLC results should be reproducible within SEER data. The SEER database catchment covers approximately 28% of the US population and documents the ‘best’ stage, that is, pathological stage if tissue was obtained or clinical stage if not. Patients from 1998 to the end of 2000 were chosen as they provided the most detailed information on disease extent.

To internally and externally validate the prognostic value of the proposed stage groupings on survival, adjacent groups were compared using hazard ratios (HRs) calculated with Cox proportional hazards regression. Adjustments were made for cell type, gender, age and region if adequate numbers of cases were available. Where numbers of patients allowed (i.e. for T size and overall stage), analyses were performed in a training set of two-thirds of the patients and validated in the remainder.


The validity of the UICC-6 T descriptors have been questioned by a number of publications demonstrating survival differences between subsets of T1 and T2, locations of intrapulmonary metastases in relation to the primary tumour, pleural disease and other T4 descriptors, as well as overlap in prognosis between groups (e.g. IB and IIA).9,14

The IASLC investigated the prognostic value of UICC-6 T descriptors in N0M0 tumours. Although analysis of several T descriptors was proposed, insufficient clinical and pathological data meant that only size, ipsilateral and contralateral nodules and pleural dissemination could be analysed in statistically valid groups. As a result, recommendations were limited to these descriptors.

Analysis was based on 18 018 UICC-6 M0 and 180 UICC-6 M1 (ipsilateral other lobe nodules) patients with complete cTNM and/or pTNM and adequate T descriptors to support the assigned T stage; 5760 were clinically staged and 15 234 were pathologically staged. None had received neoadjuvant treatment. Sixty-eight per cent (68%) and 64% had cN0 or pN0 nodal staging respectively and 4% and 0.5% had cN3 or pN3 respectively. Complete resection (R0) was achieved in 85% of patients (89% in pN0 patients) but all patients were included in analysis regardless of resection status (i.e. R1 microscopic residual disease, R2 macroscopic residual disease and RX uncertain residual disease15).

Tumour size

Tumour size analysis was based on 7480 pT1-2N0R0 tumours; 4891 patients were randomly selected as the ‘learning set’ used to create proposed cut points; the remaining 2589 patients were used for validation. The running log–rank statistic of each hypothetical tumour size cut point was plotted against tumour size. Cut points with the highest log–rank statistic values were chosen as best prognostic discriminators and rounded to the nearest centimetre.

Cut points at 2, 3, 5 and 7 cm provided the most discriminatory power in terms of overall survival and showed statistically significant differences between adjacent size groups. This led to the recommendation of splitting T1 and T2 into subgroups. These cut points remained statistically significant when applied to the validation set, and held when applied to pT1-3N0-3 any R patients. There was no difference in HRs for survival between tumours >7 cm and UICC-6 T3 tumours in either clinically or pathologically staged patients, prompting the upstaging of tumours >7 cm from UICC-6 T2 to UICC-7 T3. The prognostic value of the T size groupings was independent of histological subtype but diminished with worsening nodal status.

Additional tumour nodules

Analysis of patients with UICC-6 pT3, pT4 and pM1 (due to ipsilateral other lobe nodules) revealed five distinct prognostic groups: pT3 (n = 1224, median overall survival (mOS) 24 months); pT4 same lobe nodules (n = 363, mOS 21 months); pT4 pleural dissemination (n = 245, mOS 18 months); pM1 ipsilateral other lobe nodules (n = 180, mOS 18 months); and pT4 invasion (n = 340, mOS 15 months). When analysed in the pN0R0 subset, no statistically significant survival difference was seen between successive groups, possibly because of small numbers. However, when analysed in the pN0, any R and any N, any R subsets significant survival differences were demonstrated between pT4 same lobe nodules and pT4 invasion groups (P = 0.0029), with the former having survival akin to pT3 tumours. In contrast, survival was not significantly different between pT4 invasion versus pM1 ipsilateral other lobe nodules (P =  0.4115). This data drove the recommendations for downstaging ipsilateral other lobe nodules from M1to T4 and same lobe nodules from T4 to T3.

Pleural disease

Pathological data did not show significant survival differences between pT4 invasion versus pT4 pleural dissemination (P = 0.2904), however, as the clinical finding of pleural disease contraindicates attempted resection, comparison of clinical staging for this descriptor was more appropriate and also provided larger numbers of patients for analysis. Comparison of 471 patients with cT4 pleural dissemination and 418 patients with cT4 invasion revealed significantly different 5-year and median survival (2% and 8 months versus 14% and 13 months, respectively, P < 0.0001) suggesting that pleural disease should be upstaged from T4 to M1a.

Validation of T descriptor changes

The prognostic effects of subdividing T1 and T2 into T1a/T1b and T2a/T2b, respectively, were consistent across most contributing databases and geographical regions as well as the SEER database. Size cut points were independent of histology. In the SEER database, surgically managed patients with pT4 same lobe nodules demonstrated even better survival than those with T3 invasion further supporting reclassification from T4 to T3. Downstaging ipsilateral other lobe nodules from M1 to T4, upstaging pleural dissemination from T4 to M1a, and upstaging T2 > 7 cm to T3 were valid when applied to the SEER database.

Discussion of non-size based T descriptors is out of the scope of this review. Although prognosis is affected by vascular and visceral pleural invasion, this effect may be dependent on tumour size.16–21 Similarly, prognosis seems to vary depending on which thoracic structures are invaded.22,23 Research on the prognostic value of atelectasis has shown mixed results.24,25 Planned prospective data collection for UICC-8 will better determine the prognostic value of these descriptors.

Summary of T descriptor changes

  • 1Divide T1 into T1a (≤2 cm) and T1b (>2 to ≤3 cm).
  • 2Divide T2 into T2a (>3 to ≤5 cm) and T2b (>5 to 7 cm).
  • 3Upstage tumours >7 cm from T2 to T3.
  • 4Downstage ipsilateral same lobe nodules from T4 to T3.
  • 5Downstage ipsilateral different lobe nodules from M1 to T4.
  • 6Upstage pleural dissemination from T4 to M1a.

Independent validation of T descriptor changes

Several large surgical series and population registries have retrospectively validated the subdivision of T1 and stage changes for intrapulmonary metastases; results for subdivision of T2 and upstaging of T > 7 cm have been less consistent (Table 3). Ruffini et al.28 reported significant survival differences between T1a and T1b, but not between T2a and T2b in 1805 patients treated between 1994 and 2007. Tumours >7 cm had similar survival to other T3 tumours and shorter survival than T2 tumours. Survival of ipsilateral same lobe nodules was on par with T3 tumours and better than ipsilateral other lobe nodules, which demonstrated survival similar to T4 tumours. A study of 1632 Japanese patients29 found a significant survival difference between T2a and T2b but for uncertain reasons, T3 (>7 cm) N0M0 tumours had significantly better 5-year survival than T2bN0M0 and combined IIB disease. Independent analysis of the SEER database found that ipsilateral same lobe nodules had survival more similar to T2b than T3 tumours26 suggesting that this group could be further downstaged to T2b. Although the IASLC database contained limited patients from China, an independent Chinese cohort has validated the T descriptor changes: Li et al. analysed data from 325 stage I patients treated at the Shanghai Chest Hospital between 1998 and 2003.33 There were significant survival differences between all tumour size subsets (P = 0.025), driven mainly by differences between T2 subsets (P = 0.031).

Table 3.  Summary of independent studies validating the Union for International Cancer Control-7 T descriptors
Author, country, years analysedTotal patientsComparison (5 year survival unless otherwise stated)Findings
  1. All studies are surgical series unless otherwise stated. P-values refer to log–rank test of equality of survival hazard functions unless otherwise stated.

William et al.,26, USA, 1998–2003 (population based registry)89 240Same lobe nodule vs T2b0.966
Same lobe nodule vs T3<0.0001
Ou et al.,27 USA, 1999–2003 (population based registry)23 583Kaplan–Meier curves separated into three groups : same lobe nodule; T4 invasion and M1 contralateral nodules; and pleural/pericardial dissemination and distant metastasis<0.0001
M1a pericardial effusion vs M1a contralateral pulmonary nodules<0.0001
M1a pericardial effusion vs M1a pleural dissemination0.0379
M1a pericardial effusion vs M1b distant metastasis0.609
Ruffini et al.,28 Italy, 1994–20071 805T1a vs T1b (any N, any M)0.006
T2a vs T2b0.38
T3 vs T2 > 7 cm0.2
T3 vs same lobe nodule0.94
same lobe nodule vs T4 (invasion only)0.05
Suzuki et al.,29 Japan, 1990–20071 632T1a vs T1b (N0M0)0.009
Kameyama et al.,30 Japan, 1984–20071 532T1a vs T1b0.0026
T1b vs T2a0.0027
T2a vs T2b0.0062
Sakakura et al.,18 Japan, 1982–20001 245T1a + T1b vs T2a (any N, any M)<0.0001
T2a vs T2b + T3 (>7 cm)<0.0001
Lyons et al.,31 Argentina, 1985–2007414T1a vs T1b (N0)0.027
Ji et al.,32 China (abstract only, years not stated)327T1a vs T1bAll P < 0.0001
T2a vs T1b
T2b vs T2a
Li et al.,33 China, 1998–2003325T1a vs T1b0.6488
T1a vs T1b vs T2a vs T2b vs T3 (>7 cm)0.031
Ye et al.,34 USA, 1991–2004291T1a vs T1b (N0M0)0.023
T1a vs T1b (disease free survival)0.014


Volume of nodal disease (i.e. number of involved nodes), not just anatomical location, has never been used for lung cancer staging but has been part of TNM staging for breast, gastric and colorectal carcinoma for some time.2 Staging of nodal disease for lung cancer has also been criticized for not addressing the true heterogeneity of (especially) N2 disease, for example, N2 nodes invading the mediastinum could be considered analogous to T4 disease.35 The N Committee explored whether prognostic stratification could be improved through subdivision by anatomical location or bulk of nodal disease, for example, specifying N1 peribronchial versus N1 perihilar or single N2 versus multiple N2 nodes.10

Nodal stage

Thirty-eight thousand two hundred sixty-five cM0 patients had information on clinical N staging and 28 371 surgically managed patients had information on pathological N staging. Clear differences in overall survival were seen for each category: median survival was 40 months for cN0 falling to 23, 14 and 9 months for cN1, cN2 and cN3, respectively (P < 0.0001 between consecutive groups). The differences were predominantly driven by clinically staged patients treated surgically and were not seen in the subset of 15 451 patients managed non-surgically (median survival 13 months for N0 and 9 months for N3 disease), presumably due to the combination of medical comorbidity and lower efficacy of non-surgical treatments. Significant survival differences were also seen across the 28 371 pN patients.

A subset of 2876 pN1-N2R0 stage patients who had not received induction therapy and who had sampling results for individual nodal stations were selected for further survival analyses. Sixty per cent of these patients were from Japan. No significant difference in survival was seen when patients were analysed by anatomical location of involved lymph nodes, ‘skip metastases’ (N2 disease in the absence of N1) or the number of involved lymph node stations, although a trend for decreased survival with a greater number of positive stations was seen in the pN1 and pN2 patients.

Analysis was hampered by differences between the Naruke lymph node map used in Japan36 and the Mountain Dresler modification of the American Thoracic Society map (MD-ATS)37 used elsewhere. The most important difference lay in the nomenclature of subcarinal lymph nodes along the inferior border of the main bronchi; the Naruke map classified them as N1, but the MD-ATS map classified them as N2, potentially leading to an overall stage classification of IIA or IIIA, respectively, for the same nodal involvement. Retrospective data collection rendered this difference irreconcilable and classification remained defined by geographic region of origin.

Nodal zones

In order to bypass the differences between the Naruke and MD-ATS lymph node maps and therefore permit analyses of a larger number of patients with N1 and N2 disease, individual intrathoracic lymph node stations (2–14) were grouped into six anatomical ‘zones’ (Table 4).38

Table 4.  The IASLC lymph node ‘zones’ Thumbnail image of

Analysis of patients by lymph node zones found the following statistically significant results: patients with left upper-lobe tumours and ‘skip’ AP zone metastases had better median survival than those with involved peripheral and hilar nodes (44 vs 24 months, P = 0.0427). This was not seen for right upper-lobe tumours with right paratracheal nodal metastases. Nodal status (any T stage) was then explored using subgrouping by zonal involvement: N1a (disease in a single N1 zone, n = 798); N1b (disease in multiple N1 zones, n = 173); N2a (disease in a single N2 zone, n = 740) and N2b (disease in multiple N2 zones, n = 281). A significant difference in survival was seen between each of these groups except N2a versus N1b (mOS 52, 31, 35 and 19 months, respectively; P-values between groups: N1a vs N1b, <0.0090; N1b vs N2a, 0.7137; and N2a vs N2b, <0.0001). However, when stratified by T stage these results could not be confirmed because the resulting subgroups contained too few patients for valid statistical comparison. Nodal zones will be re-evaluated in time for UICC-8.

As a separate project, the Nodal Map Subcommittee was tasked with producing a new, internationally accepted map38 to reconcile discrepancies between current versions and enable international consistency for future data collection. The new map now provides concise and anatomically distinct descriptions of lymph node stations and their boundaries. Importantly, it has been designed to be easily applied to clinical (radiological) as well as surgical staging. The new IASLC nodal map is a major step forwards; it will simplify and harmonize nodal staging, particularly with its ready applicability to radiological staging.

Summary of N descriptor changes

No changes were made to the current N descriptors as they showed distinct survival differences and the potential for subdivision by nodal zones could not be validated.

Independent validation of N descriptor changes

Several surgical series have retrospectively evaluated survival outcomes in NSCLC using the proposed nodal zone subgroups, that is, N1a (single N1 zone), N1b/N2a (multiple N1 zones/single N2 zone) and N2b (multiple N2 zones); all have generally supported the IASLC proposals for subclassification (Table 5).

Table 5.  Summary of independent studies validating the IASLC proposed nodal zone groups (all retrospective surgical series)
Study N0N1aN1bN2aN2b
Single N1 zoneMultiple N1 zonesSingle N2 zoneMultiple N2 zones
  1. 5yr S, 5-year survival; MST, median survival time, months.

Ruffini et al.28Patients, n115028920067
5yr S, %5535258
Lee et al.39Patients, n5421712319391
5yr S, %6442363617
Rena et al.40Patients, n57010918751
5yr S, %73433219
MST, (disease free)54362821
Takamochi et al.41Patients, n42770216932
MSTNot reached82484434
Demir et al.42Patients, n367101
Kim et al.43Patients, n15364

Ruffini et al.28 found stratification into zone groups yielded highly significant survival differentiation. Patients with skip metastases had 20% 5-year survival, which was similar to the N1b/N2a and N2b groups and statistically different from the N1a group. Another Italian study40 compared the prognostic value of UICC-6 and UICC-7 in 921 patients treated between 1990 and 2005. Survival between the zone groups (N0, N1a, N1b/ N2a and N2b) was significant in both univariate and multivariate analysis.

In a study from Korea,39 Lee et al. assessed 1032 stages I–III patients treated between 1990 and 2005. Stratifying patients into zone groups yielded significantly different survival before and after multivariate analysis. In a second Korean study,43 Kim et al. focused on 217 pN2 patients with median follow-up of 35 months. Fifty per cent of the patients received adjuvant treatment. Overall and disease-free survivals were significantly better in the single-zone N2a group than in the multi-zone N2b group. If the involved nodes were confined to a single N2 zone, single and multiple station disease within the zone made no difference to survival.

Takamochi et al. analysed 647 Japanese patients.41 Survival curves showed clear separation as N stage increased. Zone groups showed significant survival differences. Within each N stage, the location of involved zones had no effect on survival. After multivariate analysis, skip metastases remained a significant prognostic factor in N2 disease.

Demir et al.42 compared the ability of UICC-6 and -7 to predict survival in 490 patients with pN1 R0 disease between 1995 and 2006. Significantly worse 5-year survival was seen for multi-station versus single-station N1 disease, multi-zone versus single-zone N1 disease, and for hilar N1 versus peripheral N1 disease. However, in a multivariate analysis, only multi-station lymph node disease predicted poor prognosis (P = 0.05). The authors concluded different postoperative strategies for N1 subgroups should be investigated.


The M Descriptors Committee analysed data from patients in the following UICC-6 groups: same lobe nodules (T4); pleural dissemination (T4); ipsilateral other lobe nodules (M1); contralateral nodules (M1); and distant metastases (M1).8 Due to the recommendations of the T Descriptors subcommittee, this analysis included patients with ipsilateral other lobe nodules but excluded same lobe nodules.9

The M group analysis was based on 5592 cT4 and cM1 patients with a further 1004 patients included in the secondary analysis of ‘best stage’, resulting in a total of 6596 patients (1216 treated surgically, 5380 treated non-surgically). Patients contribution based on geographical region was as follows: Europe, 52%; North America, 34%; Asia, 11%; and Australia, 3%. To be included in analysis, T4 patients needed at least one T4 descriptor and M1 patients required description of the metastatic site(s). Nodules in the contralateral lung with the same histology as the primary were considered metastatic; nodules with different histology were considered synchronous primaries and the highest staged primary was used (provided it was M1 disease).

As discussed in the T Descriptors section, patients with cT4 pleural dissemination had significantly worse median, 1- and 5-year survival compared with other cT4M0 (any N) disease. Similar findings were demonstrated when comparing best stage malignant pleural dissemination with best stage T4 disease (mOS 10 months vs 13 months, HR 1.91 P < 0.0001).

The majority (98%) of patients with contralateral nodules were clinically staged and hence clinical staging was analysed. Although this group had significantly better survival than the pleural dissemination group, it was more similar to the pleural group than to either the UICC-6 T4 group or the UICC-6 M1 distant metastases group (P < 0.0001 for both comparisons). Consequently, the UICC-7 M stage was divided into M1a and M1b, with pleural dissemination and contralateral nodules assigned M1a classification.

Patients with distant metastasis had a worse prognosis than either pleural dissemination or contralateral nodule groups justifying M1b status. Forty-three per cent of patients had metastases at multiple sites; this was associated with a slightly worse median survival than patients with a single metastatic site (mOS 5 months vs 6 months, 1-year survival 20% vs 23%, P = 0.006). No survival difference was seen between the various single affected sites.

The recommended changes were internally valid; survival of patients with pleural dissemination or contralateral nodules was consistently worse than T4M0 patients and consistently better than distant metastases patients across all regions and database types. Survival for UICC-6 T4 and M1 was generally worse in the SEER database than the IASLC database but discriminatory prognostic value was preserved.

Summary of M descriptor changes

  • 1Upstage pleural dissemination from T4 to M1a.
  • 2Reclassify contralateral nodules from M1 to M1a.
  • 3Reclassify distant metastasis from M1 to M1b.

Independent validation of M descriptor changes

The largest independent validation of UICC-7 T4 and M1a/M1b descriptors was based on 23 583 NSCLC patients from the California Cancer Registry, the largest contiguous area population based registry in the world27 (Table 3). UICC-6 T4 and M patients were divided into seven groups: same lobe nodules (UICC-7 T3); T4 invasion (UICC-7 T4); ipsilateral different lobe nodules (UICC-7 T4); contralateral nodules (UICC-7 M1a); pleural dissemination (UICC-7 M1a); pericardial effusion (UICC-7 M1a); and distant metastasis (UICC-7 M1b). Same lobe nodules had superior 1 year and mOS compared with all other groups. Survival with ipsilateral different lobe nodules was similar to T4 invasion and better than contralateral nodules and pleural dissemination. Distant metastatic disease demonstrated the worst 1 year and mOS. Subdivision of UICC-6 M1 into UICC-7 M1a and M1b was justified based on comparison of HRs from Cox proportional hazards regression analysis. Malignant pericardial effusion showed survival more similar to distant metastases than pleural dissemination and the authors suggested this should be upstaged to M1b in UICC-8.


The IASLC altered UICC-7 stage groupings to accommodate changes to individual T, N and M descriptors (Table 2).6 Survival analysis was based on Kaplan–Meier estimates and prognostic groupings were compared with Cox regression models. Where the individual T, N and M subcommittees had recommended up- or downstaging of a descriptor, the Stage Grouping committee considered two different approaches: (i) retain the descriptor in its existing category with additional alphabetic subscript, for example, additional same lobe nodules, considered T4 in UICC-6, would be designated T4a; or (ii) allow descriptors to move to a category containing other descriptors with a similar prognosis. The first approach would enable backward compatibility with existing databases, however, over 180 TNM subsets would have been created, making it impractical to analyse and use. Therefore backward compatibility was sacrificed and the second approach was used. Best split points were determined from patients classified by ‘best stage’. Recursive partitioning and amalgamation analysis was used based on survival data from a randomly selected learning subset of 17 726 patients; the remaining 9133 patients were used for validation. Different combinations of T, N and M descriptors were analysed and HRs calculated using T1aN0 as the reference. Groups with statistically similar HRs were then assigned the same stage, resulting in changes to several stage groups.

The UICC-7 stage groupings resulted in a more even distribution of patients between stages and better delineation of median and 5-year survival between IB and IIA and between IIA and IIB groups compared with UICC-6. R2 values were higher in UICC-7 compared with UICC-6 for both clinically (26.77 vs 25.86) and pathologically (30.40 vs 29.44) staged disease. Survival in stage IV disease was paradoxically better than stage IIIB disease, probably because of the nature of presumed surgically resectable tumours in these categories (e.g. small effusions, additional tumour nodules found at surgery or isolated metastatic disease).

The proposed changes to stage groupings were applied to the validation group of 9133 patients. In general, the survival curves and HRs were comparable with those in the learning set when controlled for age, gender, region and histology.

UICC-6 and UICC-7 pathological stage groupings showed significant differences in survival for all groups except IV versus IIIB. UICC-7 clinical stage groupings showed significant differences between all stages whereas UICC-6 did not differentiate between IIA versus IB or IIB versus IIA.

When applied to the SEER database, both UICC-6 and UICC-7 stage groupings demonstrated prognostic stratification, however UICC-7 better delineated survival between stages IB and IIA whereas UICC-6 demonstrated greater separation of Kaplan–Meier curves between IIA and IIB at 5 years. Good prognostic differentiation was achieved with UICC-7 between stages IIIB and IV.

Overall, the UICC-7 group stages result in more even distribution of patients amongst all groups, predominantly due to increased patients in stage IIA. It addresses several weaknesses of the UICC-6 groupings, such as similar survival between groups IB and IIA, but as a consequence sacrifices backwards compatibility and could alter established treatment algorithms.

Independent validation of stage group changes

Between 12 and 34% of patients will change stage between UICC-6 and UICC-7.44,45 Most of these changes are due to the addition of T2aN1M0 (from UICC-6 stage IIB) and T2bN0M0 (from UICC-6 stage IB) tumours to stage IIA. UICC-7 stage IIIB also has reduced numbers due to reclassifying ipsilateral same lobe nodules from UICC-6 T4 to UICC-7 T3 and upstaging of pleural dissemination from UICC-6 T4 to UICC-7 M1a. Strand et al. calculated that stage migration could change management in up to 17.3% of patients, mainly by crossing from surgery alone (stages IA and IB) to surgery with adjuvant chemotherapy (stage II).46 Further trials are needed to determine whether treatment algorithms should be altered in light of these changes. Details of stage changes from four institutions are outlined in Table 6. Independent studies have generally validated new stage groupings, with survival decreasing in a stepwise fashion from stages I through IV.

Table 6.  Stage migration in studies comparing the Union for International Cancer Control (UICC)-6 and UICC-7 classification
StageSuzuki et al.29Klikovits et al.44Fukui et al.47Strand et al.46
% change% change% change% change
(n = 1632)(n = 145)(n = 1556)(n = 1885)
IB↓ 3.5%↓ 30%↓ 5.8%↓ 11.2
IIA↑ 7.1%↑ 283.3%↑ 8.4%↑ 16.3%
IIB↑ 1.2%↓ 70.7%↓ 2.5%↓ 5.9%
IIIA↑ 6.1%↑ 31.4%↑ 3.9%↑ 6.2%
IIIB↓ 12.8%↓ 36.4%↓ 5.2%↓ 2.9%
IV↑ 2%↓ 33.3%↑ 1.2%↓ 2.5%
Total % changeNot stated33.8% changed stage9% higher stage 9% lower stage17.3% ‘potential change in treatment’


Although small surgical series have validated the prognostic significance of TNM staging for SCLC, it has never been widely adopted for two main reasons: SCLC is rarely amenable to surgery (less than 5% of patients); and curative-intent radiotherapy is considered only if disease is confined to a single radiation field. For simplicity, the dichotomous ‘Limited’ and ‘Extensive’ disease classification (LD and ED) has been used, with LD referring to involvement of one hemithorax with or without regional lymph node metastases with or without ipsilateral pleural effusion.48 ED comprised everything beyond LD. This broad and heterogeneous classification may, however, hide subsets of patients who would benefit from more aggressive therapy such as newer chemo-radiation regimes or conformal radiotherapy techniques. Thus, the IASLC attempted to firstly assess the applicability of UICC-6 staging to SCLC, and secondly to apply the proposed UICC-7 descriptors.11,12

The IASLC database included 12 620 SCLC patients. Eight thousand eighty-eight had complete cTNM staging (3430 cM0 patients and 4530 cM1 patients); 349 had complete (R0) resection and full pTNM staging. Of the surgical patients, 262 had enough staging detail to allow UICC-7 classification. Over half of the patients were from Europe and 60% were from population-based registries or single-institution series/registries.

In patients who had undergone both clinical and pathological staging, correlation between T stage was reasonable but N staging showed discrepancies. The overall agreement between cTNM and pTNM was 58%: 30% of stage cIIIA and cIIIB patients were downstaged to ≤stage pII, and 19.7% of stage cI and cII were upstaged to ≥ stage pIIIA.

SCLC pathological stage

Survival analysis by pathological stage was hindered by small numbers and patient selection, particularly in higher T groups. In pT1-4N0-3 M0 disease, significant 5-year survival differences were only found between pT1 and pT2 (P = 0.0109) although only small numbers were included in the higher pT-staged groups. Application of UICC-7 demonstrated mOS for pT3 and pT4 of 18 and 9 months, respectively. Worsening nodal status predicted progressively shorter 5-year survival (N0 vs N1 P = 0.0357, N2 vs N1 P < 0.0001, N3 vs N2 P = 0.0371), consistent across all pTM0 groups. Insufficient pM patients prevented survival analysis of this group. Two hundred sixty-two patients could be classified by both UICC-6 and UICC-7, hence direct survival comparison was limited to this subgroup; UICC-7 demonstrated more clearly a progressive worsening of 1-year survival with increasing stage, however HRs were only significant between groups IIIA and IIB (HR 2.49 P = 0.0006 UICC-6 and HR 2.34 P = 0.0015 UICC-6). Application of UICC-7 to the SEER database revealed the same trend.

SCLC clinical stage

For SCLC patients clinically staged using UICC-6, significant 1- and 5-year survival differences were seen between T stage categories (N0-3 M0): T1 versus T2 P < 0.0001; T3 versus T2 P = 0.0185; T4 versus T3 P = 0.0055. Increasing N status was generally associated with worsening survival (ordered log–rank test <0.0001) although the difference between cN0 and cN1 was not significant (P = 0.76) and although of statistical significance, the median survival between cN2 and cN3 was actually quite similar (14 months vs 12 months). Comparison of sequential stage groupings showed significantly different HRs, except for stage IB, which had worse survival than IIA possibly due to the fact that IA and IIA groups contained T1 tumours, whereas IB and IIB groups comprised T2 and T3 tumours. Survival curves between patients with LD with effusion were intermediate between LD without effusion and ED disease (P < 0.0001), supporting an M1a designation. When classified by UICC-7, the same overall survival trend was seen although smaller numbers (n = 2464) limited statistical significance. The findings were validated in the SEER dataset but details were not provided.

Summary of SCLC changes

The IASLC proposed that UICC-7 TNM classification should be applied to patients with SCLC. In particular, stratification by stage should be incorporated into clinical trials for early-stage SCLC.

Independent validation of SCLC changes

To our knowledge, only one study has independently validated the use of TNM staging in SCLC. In 10 660 SCLC patients from California Cancer Registry diagnosed from 1991 to 2005, survival correlated with both UICC-6 and UICC-7 T descriptors and N2–3 disease had worse survival than N0-1 disease. Survival curves were better separated by UICC-7 classification than UICC6 in both univariate and multivariate analyses. Pleural effusion had similar survival to other M1a categories.49


Bronchopulmonary carcinoid tumours have traditionally been excluded from TNM staging however some researchers have found reasonable prognostic correlation when applying TNM staging.50 As part of the IASLC Staging Project, data from 513 completely staged carcinoid tumours from the IASLC database and 1619 from the SEER database were analysed to evaluate the applicability of UICC-7 in predicting survival.51

The cohort of surgically managed patients (n = 1437) from the SEER dataset (1990–2002) was used as the validation group. Cause of death data were generally available but typical versus atypical designations were not. Sixty per cent of patients were followed for 5 years or until death; 26% had 10 years of follow-up. Median tumour size was 2.0 cm but there were very few tumours >3 cm and most N0 carcinoids were between 1 and 1.5 cm. Seventy-eight per cent were stage I, 12% stage II, 6% stage III and 3% stage IV. Survival generally decreased with increasing T, N and M stage, however, there was no difference in the 5-year survival between T1 and T2 (any N) groups, or between T1a/T1b (93/92%) or T2a/T2b (90%) categories. T3 disease had significantly shorter survival than both T1 and T2 tumours (P < 0.0001), and T3 > 7 cm had slightly worse survival than other T3 disease. Five-year survival by N status was as follows: N0, 92%; N1, 81%; N2, 74%; N3, 0%. M1 disease had shorter survival than M0 disease (5-year survival, 57% vs 91% P < 0.0001).

The IASLC database contained data on 513 carcinoids, mostly from surgical centres. Seventy-six per cent of these patients had sufficient pTNM descriptors to allow meaningful analysis. Clinically staged carcinoid tumours were not analysed due to small numbers (n = 36). Only 50% of patients had typical versus atypical designations and information regarding cause of death was unavailable. The IASLC database followed general trends in survival and HRs that were found with the SEER data, although smaller numbers precluded analysis of more advanced disease.

The IASLC hopes to improve staging of carcinoid tumours by prospective data collection through an International Registry of Pulmonary Neuroendocrine Tumours.52

Summary of carcinoid tumour changes

The IASLC recommended that UICC-7 staging be applied to all bronchopulmonary carcinoid tumours.

Independent validation of carcinoid tumour changes

Two studies have assessed the appropriateness of TNM staging in carcinoid tumours. A study of 186 patients (164 typical, 22 atypical) who underwent surgical resection showed that UICC-7 staging, age>60, atypical histology, N stage and metastatic disease were all significant prognostic factors.53 A separate cohort of 110 resected carcinoids demonstrated a significantly different survival for all patients in relation to tumour size.54


The UICC-7 TNM staging system represents a major advance in lung cancer care. Refinements over previous iterations are based on the collation and analysis of the largest collection of lung cancer patients ever assembled, irrespective of treatment modality. The international staging system is now truly based on international data with patients sourced from 46 data repositories across more than 19 countries and four continents. Meticulous analysis of tremendous amounts of data has resulted in a raft of recommendations built on a strong scientific foundation. From the outset, data integrity and applicability has been paramount and the various subcommittees have certified their recommendations by applying multiple levels of both internal and external validation. The benefits of this approach are evidenced in the high level of concordance when independent authors apply the recommended changes to their own cohorts.

For all its successes, however, the UICC-7 TNM staging system has several limitations, the most significant of which is the retrospective collection of data from sources that were never designed to inform the staging system. Because of the enormous amount of resources required to devise and regulate an international prospective database, the IASLC realized that retrospective data collection from pre-existing databases was the only way to ensure recommendations were agreed upon and disseminated in a timely fashion for the UICC-7th edition. As a result, they had an abundance of information regarding certain TNM descriptors (such as tumour size), which had been routinely collected, but almost nothing on other descriptors that had either been deemed ‘not important’ by the individual databases, or were too difficult to accurately measure or confirm (such as extent of pleural invasion). Although these descriptors remain in UICC-7 staging their prognostic value remains uncertain. Another limitation on the choice of retrospective data was the time period chosen. This was prior to widespread use of PET scanning, which nowadays is routinely used. This questions the contemporaneousness of the data upon which the recommendations are based. Furthermore, retrospective data is well known for problems such as missing data and inaccuracies. Due to the global reach of this study and lack of both personnel and finances, the IASLC could not undertake its own data audit and was thus obliged to rely on the integrity of the data as supplied.

Non-anatomic factors

TNM classification for some cancers already includes non-anatomic prognostic factors such as age (thyroid), histological grade (soft tissue sarcoma, prostate) and serum markers (testicular and gestational trophoblastic tumours). Variables such as gender, age and performance status were found to have significant prognostic value within the limitations of the retrospective IASLC database55 and will be prospectively evaluated in the next revision. standardized uptake value (SUV), a normalized index of tracer uptake in 18FDG-PET scans, was found to have prognostic value in NSCLC in a meta-analysis of 21 studies56 (high SUV vs low SUV HR 2.08, 95% confidence interval: 1.69 to 2.56). It is likely, in the future era of ‘personalized’ lung cancer therapy, that somatic molecular and genomic data such as EGFR status57–59 will need to be incorporated into staging.

In an attempt to address at least some of these shortcomings, the IASLC has commissioned a prospective dataset to inform changes to UICC-8, which will be officially published in 2016.60 Because IASLC is unable to verify the accuracy of collected information, the authenticity of the data upon which UICC-8 is based will rely heavily on the goodwill of the medical community to accurately characterize lung cancer patients. Necessary data elements have already been proposed (available at with particular emphasis on those descriptors that eluded validation in UICC-7. Because data collection is occurring over a shorter time period (2009–2012), adequate numbers of patients will only be achieved by collaborative international representation, including from regions such as South America and Africa and countries like China and Russia, all of which were absent or under-represented in UICC-7.


The UICC 7th Edition TNM Classification System represents the culmination of a 10-year project designed to improve the scientific basis for lung cancer staging. It is underpinned by the largest international database of lung cancer patients ever assembled. Rigorous internal and external validation has ensured accuracy and universal applicability. However, staging is not a static process: it needs to continually evolve as lung cancer disease, investigation and treatment continue to evolve. The 7th iteration of the TNM Classification System has only been possible through the voluntary involvement of lung cancer centres worldwide but an ongoing international collaboration is needed if we are to continue to make progress in this challenging but fundamentally important area.