Cross-validation study of class III beta-tubulin as a predictive marker for benefit from adjuvant chemotherapy in resected non-small-cell lung cancer: analysis of four randomized trials

Abstract

Background: The IALT, JBR.10, ANITA and Cancer and Leukemia Group B 9633 trials compared adjuvant chemotherapy with observation for patients with resected non-small-cell lung cancer (R-NSCLC). Data from the metastatic setting suggest high tumor class III beta-tubulin (TUBB3) expression is a determinant of insensitivity to tubulin-targeting agents (e.g. vinorelbine, paclitaxel). In 265 patients from JBR.10 (vinorelbine-cisplatin versus observation), high TUBB3 was an adverse prognostic factor and was associated (nonsignificantly) with 'greater' survival benefit from chemotherapy. We explored this further in additional patients from JBR.10 and the other three trials.

Patients and methods: TUBB3 immunohistochemical staining was scored for 1149 patients on the four trials. The original JBR.10 cut-off scores were used to classify tumors as TUBB3 high or low. The prognostic and predictive value of TUBB3 on disease-free survival (DFS) and overall survival (OS) was assessed by Cox models stratified by trial and adjusted for clinical factors.

Results: High TUBB3 expression was prognostic for OS [hazard ratio (HR)=1.27 (1.07-1.51), P=0.008) and DFS [HR=1.30 (1.11-1.53), P=0.001). TUBB3 was not predictive of a differential treatment effect [interaction P=0.20 (OS), P=0.23 (DFS)]. Subset analysis (n=420) on vinorelbine-cisplatin gave similar results.

Conclusions: The prognostic effect of high TUBB3 expression in patients with R-NSCLC has been validated. We were unable to confirm a predictive effect for TUBB3.

Figures

Figure 1.

Survival distributions for patients with TUBB3-high or -low tumors in the validation cohort of 1149 patients with R-NSCLC who participated in four clinical trials of adjuvant chemotherapy versus observation. The HRs, 95% CIs and P value estimations reported were obtained from adjusted Cox models stratified by trial. (A) Overall survival is inferior in TUBB3-high patients; (B) disease-free survival is inferior in TUBB3-high patients. TUBB3, class III β-tubulin; HR, hazard ratio; CI, confidence interval.

Figure 2.

Forest plots of predictive analysis results from individual trials included in the validation cohort for patients treated with chemotherapy versus observation according to TUBB3 status, along with the results from the patients in the previously published JBR.10(1) study [7]. HRs of the interaction between TUBB3 and chemotherapy effect for each trial, the validation set, the hypothesis set, and the overall population are reported with their 95% confidence intervals, the corresponding number of events and patients. The number of events/patients given for each line corresponds to the sum of the number of events/patients of the four following groups of the corresponding trial or group of trials of this line: control group with low tubulin; control group with high tubulin; chemotherapy group with low tubulin; chemotherapy group with high tubulin. The interaction HR is equal to the HR of chemotherapy effect (chemotherapy compared with control) for the high tubulin group (HR high) divided by the HR for the low tubulin group (HR low); if this ratio is equal to 1 then the effect of chemotherapy is the same for the two tubulin groups. In this case, there is no interaction between tubulin group and chemotherapy effect. For instance, in Figure 2A, the HR of 0.81 given for the pooled analysis of the validation set is the ratio of the HR of deaths given in the upper part of Table 3 left column): 0.83 divided by 1.03. (A) Overall survival for all patients. (B) Disease-free survival for all patients. (C) Overall survival for the subset randomly allocated to cisplatin–vinorelbine or observation. (D) Disease-free survival for the subset randomly allocated to cisplatin–vinorelbine or observation. TUBB3, class III β-tubulin; HR, hazard ratio.