Stage 3 lung cancer

Introduction: Since 1974, a tumor size of 3 cm in diameter has been regarded as the prognostic threshold in the staging of bronchogenic carcinoma.

Objective: To study the prognostic behavior of surgical-pathologic tumor size in non-small cell lung cancer (NSCLC) with complete resection.

Design: Four-year multi-institutional prospective study from 1993 to 1997.

Patients: Consecutive cases of NSCLC in pathologic stages IA-IB (pIA-pIB) treated surgically with complete resection in hospitals belonging to the Bronchogenic Carcinoma Co-operative Group of the Spanish Society of Pneumology and Thoracic Surgery (GCCB-S).

Methods: The Schoenfeld procedure was used to identify different prognostic groups, considering 1 cm as the measurement unit.

Results: Based on the 1,020 cases evaluated, four prognostic groups were identified: 0 to 2 cm (group A; n = 147), 2.1 to 4 cm (group B; n = 448), 4.1 to 7 cm (group C; n = 336), and > 7 cm (group D; n = 89). At 5 years, survival was 0.63 (95% confidence interval [CI], 0.58 to 0.68), 0.56 (95% CI, 0.53 to 0.59), 0.49 (95% CI, 0.46 to 0.52), and 0.38 (95% CI, 0.32 to 0.44) for groups A, B, C, and D, respectively. Differences between paired groups (log-rank) were significant: 0.0074 between groups A and B, 0.0048 between groups B and C, and 0.0034 between groups C and D.

Conclusions: In initial stages (pIA-pIB) of NSCLC, the 3-cm value was not found to behave as a prognostic threshold; in this study, four surgical-pathologic tumor size groups were identified with strong prognostic differences: from 0 to 2 cm, from 2.1 to 4 cm, from 4.1 to 7 cm, and > 7 cm.

Key words: lung cancer; lung neoplasms; size; surgery

Abbreviations: CI = confidence interval; GCCB-S = Bronchogenic Carcinoma Co-operative Group of Spanish Society of Pneumology and Thoracic Surgery; NSCLC = non-small cell lung cancer

**********

Since 1974, (1) 3 cm has been regarded as the only tumor size to establish a prognostic threshold in the staging of bronchogenic carcinoma. Even though the staging classification has been updated repeatedly (the last update being in 1997), (2) criteria for tumor size has remained unchanged for the last 25 years.

There is a need to conduct further research on other clinical or molecular biological prognostic factors in patients with lung cancer; therefore, it becomes essential for the basic anatomic classification of tumors to be founded on solid grounds. A new prognostic factor could only be considered useful, and included as part of the tumor classification, if it has the ability to modify the prediction and accuracy of the TNM anatomic classification. (3)

The TNM classification is simplified to make its use easier; however, that simplicity may lead to decreased prognostic certainty or, in other words, a loss of prognostic discrimination. Repeatability or validation of the prognostic values for the main pathologic stages in a surgical population with non-small cell lung cancer (NSCLC) has been demonstrated in the United States, Japan, Germany, and Spain. (4) Nonetheless, and as it was to be expected, it has now been acknowledged that the consideration of TNM descriptors was more predictive of prognosis than the assessment of tumor stages. (5)

The Bronchogenic Carcinoma Co-operative Group of the Spanish Society of Pneumology and Thoracic Surgery (GCCB-S) has recently evaluated its experience in the prognostic analysis of prethoracotomy clinical tumor size in lung cancer. (6) This study was able to identify four different prognostic groups of tumor size based on radiography. However, since this clinical staging was based on a population that in the end required surgery, it involved some problems: the greater the tumor size, the greater the probability to have invaded mediastinal lymph nodes at thoracotomy, even in cases of small NSCLC; or the higher frequency of exploratory thoracotomies or incomplete resections in larger tumors, among other limitations. (7-10) The goal of this study is to evaluate tumor size as a prognostic factor in patients with pT1-T2N0M0 NSCLC undergoing complete resection in a nonselected surgical population.

MATERIALS AND METHODS

Study Subjects

All the patients included prospectively in this study had NSCLC in initial stages (pIA-pIB) and underwent thoracotomy with complete resection in hospitals pertaining to the GCCB-S (11) between October 1993 and September 1997, both inclusive. Similar criteria for the functional operability of patients and oncologic operability of the tumor were used in all the GCCB-S hospitals. (12) The participating GCCB-S centers had a wide variety of activities, including a representative range of number of beds, type of activity (university and nonuniversity, community, public, and private ownership), and number of interventions per year (range, 8 to 100 interventions). The sample was complete, as verified by the inclusion in the registry of all patients undergoing complete surgical resection. Patients with death due to operative mortality or patients receiving neoadjuvant therapy were excluded from the analysis. The total number of NSCLC patients operated on with any pathologic stage and any type of surgery was 2,300. The final number of cases included in this study was 1,020. Data were collected prospectively, in real-time, over the 4-year study period, using a unified single self-copying form that included the original and a copy (the original form was filed at the hospital and a copy omitting affiliation data was sent to the GCCB-S headquarters). At the GCCB-S Registry, each classificatory component for the different T categories (tumor size, pleural tumor involvement, or tumor involvement of other endothoracic structures) was considered differently and on an individual basis for each patient. The diagnostic procedure employed was marked using a previously agreed-on code and the procedure showing the best resolution recorded in the registry. The surgical-pathologic size was obtained by measuring the greatest diameter of the fresh surgical specimen.

Surgical-pathologic stage N0 was assigned when there was no lymph node involvement after systematic nodal dissection or when sampling of at least four lymph node stations was performed (stations 2, only on the right side, 4, 7, and 10 ipsilateral to the tumor). (13) These criteria were similar to those proposed in recent recommendations, (14) such as the need to obtain six hilar-mediastinal lymph nodes, in order to define pN0. For the purpose of classifying the presence or absence of mediastinal lymph node involvement, a randomized study (15) demonstrated that sampling had a value similar to that of mediastinal lymph node dissection. The GCCB-S operational definition for standard complete resection requires the absence of tumor involvement of resection margins, extracapsular involvement of resected mediastinal lymph nodes, positive biopsy finding in unresected mediastinal lymph nodes, and the absence of tumor pleural effusion.

Internal and external audits were performed to review the ratio between the number of patients undergoing surgery and the number of cases included in the GCCB-S Registry (standard > 95%), and to review the presence and validity of the data collected and recorded for each case (standard > 75%), including the consistency of tumor staging. The criterion for the validity of survival data was established on the existence of a known follow-up of, at least, 85% of the case-patients registered in each hospital. In the hospitals that did not meet all these conditions, the cases corresponding to the anomalous period were excluded. Finally, correct data transmission from the paper record to the computer database was verified on two separate occasions.

These procedures were designed to control the following aspects: selection bias of surgical cases, registered cases over the total number of surgical cases, sample size, type of hospital, prognostic migration due to the prolonged period of case recruitment, classification with low or deficient degrees of certainty, contamination by data from incomplete series or erroneous data, and loss of adequate long-term follow-up.

Analysis

Tumor size has been analyzed in different studies (16,17) using different methods and perspectives. Notwithstanding, it is a known fact that compilation of observations of supposedly continuous variables is bound to include errors that affect both the accuracy and validity of such estimations. The terminal digit preference phenomenon and rounding up have been cited as typical examples. (18,19) The statistical reliability, the way the variable is distributed, and the accurate classification of the study subjects can be seriously affected as a result of this phenomenon. This study verified the distribution of the tumor size variable in order to choose the best method of analysis by observing the mentioned digit preference.