Validation of the lung cancer staging system revisions using a large prospective clinical trial database (ACOSOG Z0030)

Abstract

OBJECTIVES

A new revision of the international lung cancer staging system has been recently introduced. The revisions are largely focussed on the T descriptor. We sought to test the validity of this new system on a separate prospectively collected cohort of patients from a recent multicentre trial of early-stage lung cancer.

METHODS

We reviewed the prospectively collected data from 1012 patients undergoing pulmonary resection for early-stage lung cancer in the ACOSOG Z0030 trial. TNM descriptors and overall staging were assessed using both the sixth and seventh editions of the American Joint Committee on Cancer and the Union Internationale Contre le Cancer (AJCC/UICC) lung cancer staging system. Survival results were analysed according to both staging allocations.

RESULTS

Using the proposed criteria, the number of patients by stage in the sixth and seventh edition allocations, respectively, were as follows: IA (432, 431); IB (402, 303); IIA (39, 167); IIB (94, 70); IIIA (26, 40); IIIB (19,0); there were no stage IV patients by either version. Overall, 180 (18%) patients had a change in the stage group from the sixth to seventh edition versions with 76 (8%) being downstaged and 104 (10%) being upstaged. In the sixth edition staging system based on pathological stages, median survivals in years were as follows: IA, NA; IB, 7.7; IIA, 4.0; IIB, 3.6; IIIA, 2.6 and IIIB, 2.4. Five-year survivals were: IA, 76.4%; IB, 62.0%; IIA, 47.8%; IIB, 40.4%; IIIA, 31.3% and IIIB, 44.4%. In the new system, median survivals in years were as follows: IA, NA; IB, 8.2; IIA, 4.4; IIB, 3.6 and IIIA, 1.8. Five-year survivals were: IA, 76.9%; IB, 65.0%; IIA, 48.5%; IIB, 42.9% and IIIA, 30.6%. Survival analysis and Kaplan–Meier survival curves showed more monotonic progression, distinction and homogeneity within groups in the seventh edition.

CONCLUSIONS

This study provides an external validation of the recently revised lung cancer staging system using a large multicentre database. The seventh edition of the AJCC/UICC lung cancer staging system appears to be an improvement over the preceding system.

INTRODUCTION

Lung cancer remains the leading cause of cancer death in the USA [1]. Staging, in that it provides information regarding prognosis and appropriate treatment options, is an essential component in the management of all cancers including lung cancer.

Recently, the International Staging Committee of the International Association for the Study of Lung Cancer (IASLC) introduced revisions to the lung cancer staging system or the so-called seventh edition revisions [2]. The prior edition (sixth edition), published in 2002, had no significant changes from the 1997 version and was based on a cohort of 5319 surgically treated patients from North America from 1975 to 1988 [3]. The seventh edition revisions are based on data from 68 463 patients with non-small-cell lung cancer from 20 countries collected from 45 different databases dating from 1990 to 2000.

The majority of revisions in the seventh edition focussed on the T (tumour) descriptor. In fact, no changes were made to the N (nodal) descriptor and the only changes in the M (metastasis) descriptor were to separate the intrathoracic metastases (contralateral nodules and malignant pleural effusions) from the extrathoracic metastases. The former is classified as M1a, whereas the latter become M1b [4]. This study provides the first large multicentre external validation of the recently revised lung cancer staging system.

MATERIALS AND METHODS

We used the database from the ACOSOG Z0030 trial, which randomized patients with early-stage lung cancer to either lymph node sampling or lymph node dissection [5]. From June 1999 to February 2004, 1111 patients were randomized into the Z0030 trial. Eligibility requirements included patients older than 18 years of age, an Eastern Cooperative Oncology Group performance score lower than 3 and a tissue diagnosis of a clinically resectable T1 or T2, N0 or nonhilar N1, M0 non-small-cell lung cancer (squamous cell carcinoma, large cell carcinoma or adenocarcinoma, including bronchoalveolar carcinoma). Eligible patients had to be candidates for complete resection of the carcinoma by means of pneumonectomy, lobectomy, bilobectomy or anatomical segmentectomy, with or without sleeve resection. Exclusion criteria for the ACOSOG trial included patients who had T3 or T4 tumours, those who were treated with pulmonary wedge excision and those who received prior chemotherapy or radiation therapy for their cancer. After randomization, a retrospective review found 155 (13.9%) patients to be ineligible for participation. The reason for ineligibility was minor (e.g. timing violation) in 67 of these patients. The remaining 88 patients were excluded for major violations including incorrect stage in 29, inadequate lymph node sampling in 14, benign disease in 6, insufficient documentation in 5 and other reasons in 34. Due to limited follow-up on the 88 ineligible subjects, they were excluded from the current study. An additional 11 subjects were excluded due to inadequate data for the staging review. The remaining 1012 patients form the basis of this study.

Each case was evaluated using both the seventh and sixth edition TNM classifications [2, 3]. A summary of the classification changes is included in Table 1. This study was approved by the institutional review board of the Mayo Foundation.

Cumulative survival probabilities were estimated using the Kaplan–Meier method. The log-rank test was used to compare the survival of groups. A survival concordance index was calculated assessing the fit of the survival model for both the sixth and seventh staging classifications. In all cases, P-values <0.05 were considered statistically significant.

RESULTS

There were 1012 patients with lung cancer; 526 (52%) males and 486 (48%) females who underwent surgery in the ACOSOG Z0030 trial. Mean age was 67.4 ± 9.2 (range 23.7–89.6). The median follow-up was 80 (range 0–121 months). The distribution of the type of surgical procedure was as follows: segmentectomy: 70 (6.9%), lobectomy: 754 (74.5%), bilobectomy: 42 (4.2%), pneumonectomy: 42 (4.2%), combination: 100 (9.9%).

The allocation of patients by the TNM descriptor and stage grouping for both the sixth and seventh edition staging systems is summarized in Tables 2–4. Overall, 180 (18%) patients had a change in the stage group from the sixth to seventh edition versions with 76 (8%) being downstaged and 104 (10%) being upstaged.

Overall survivals for both sixth and seventh editions T-stage and overall staging systems are summarized in Table 5. The survival concordance index, which assesses the fit of the survival model, was 0.62 for the sixth edition stage and 0.63 for the seventh edition. Survival curves by overall stage can be seen in Figs 1 and 2.

DISCUSSION

Lung cancer treatments are based on staging and changes in staging greatly affect these treatments strategies [6]. Although the current (sixth edition) TNM Classification of Malignant Tumours by the UICC is an excellent staging system, various problems have been indicated in many aspects for lung cancer. The next edition of the TNM Classification of Malignant Tumours, the seventh, will try to solve this. In this regard, the IASLC established its Lung Cancer Staging Project in 1998 to bring together the large databases available worldwide and to formulate recommendations for the seventh edition that would be intensively validated. Analysis of data from 81 015 cases worldwide was performed. The IASLC proposed a revision of the current staging system that was presented at the 12th World Conference on Lung Cancer. In addition, the details were published in the Journal of Thoracic Oncology [2, 7–10]. The new staging system was based on an intensive and validated analysis of the largest database available to date.

Given this imminent change and also for the evaluation of new treatment strategies, there is the need for the migration of current data into the new staging system in most institutions [11].

In this study, we adopted the changes proposed by the IASLC and reclassified the patients of the prospectively collected ACOSOG Z0030 database according to the new stage grouping and analysed the new stage groupings and assessed survival differences across the stage groups (T-stage and overall).

The seventh edition of the TNM did not make any changes in the N-stage so there were no differences in this respect in staging when migrating patients from the sixth to seventh edition. We focussed on early stages (IA–IIB) as N2 and T4 patients were initially excluded from the study, although N2 and T4 patients were found after surgery.

Both staging systems showed distinctive survivals between groups with the log-rank values <0.001. When analysing the median survival for every stage, the primary differences in survival were seen in early stages (IA–IIB) where most of the patients in this study were included.

For a cancer staging system to be useful, it must provide accurate prognostic data that will ultimately assist in determining appropriate treatment options. To achieve this optimally, the staging system must fulfil three criteria.

First, it must be monotonic. This, in mathematical terms, describes sequences, the successive members of which either consistently increase or decrease but do not oscillate in relative value. In simple terms, as it relates to cancer staging, the curves on a Kaplan–Meier graph should not cross each other. The survival concordance index was 0.62 for the sixth edition and 0.63 for the seventh edition. In the sixth staging edition, the Kaplan–Meier curves for each were not completely distinct and crossed at numerous time points (Fig. 1). With the seventh edition, distinction was maintained and crossings were not seen until after 6 years when the sample size became limited (Fig. 2).

Secondly, there should be a distinction between groups. In other words, the prognosis for one-stage group should be statistically different from the others. We found statistical differences between IA vs IB and IB vs IIA in the seventh edition, while significant differences were only observed for IA vs IB in the sixth edition (Table 5). The lack of differences in other stages is probably due to the limited sample size in the higher stages.

Thirdly, the results should be homogenous within groups; those within a specific stage grouping should generally have little variation in prognosis. We see how in the sixth edition, the difference in years of survival between stages IB and IIA is 3.7 years while in the seventh it is 3.8 years; between IIA and IIB in the sixth it is 0.4 years and in the seventh, 0.8 years, between IIB and IIIA in the sixth it is 1 and in the seventh, 1.8 years. These results translate a higher homogeneity in survival in the seventh edition staging.

Although Z0030 is a prospective clinical trial, data to migrate patients to the seventh edition had to be abstracted retrospectively and thus the study carried limitations of population-based studies. It was focussed on early-staged patients; therefore, the number of patients included in advanced stages is lower.

In conclusion, for the pathological staging proposed by the IASLC, the revisions that focussed on the T-classification were reflected well by the improvements for stages I to II. Survival curves and overall survival in the seventh edition patients proved to be more monotonic, distinctive and homogeneous than in the sixth edition. The new pathological staging system was considered valid based on the multicentre database.

Funding

This project was supported by the American Association for Thoracic Surgery Graham Memorial Traveling Fellowship.

Conflict of interest: none declared.

ACKNOWLEDGEMENTS

The authors would like to acknowledge the expert assistance in manuscript preparation of Kristin A. Burke.

REFERENCES