Giaccone et al
attenuated vaccinations. For these attenuated vaccinations, BCG was reconstituted in diluent, giving final solutions containing 5.0
106, 1.0 106, 5.0 105, 1.0 105, and 5.0 104 colony-
forming units for vaccinations one through five. Before each vac- cination performance status, clinical symptoms and adverse effects were evaluated and blood sampling for humoral response was performed. Two weeks after the last vaccination (week 12), patients were reassessed with full blood counts and serum chem- istry, chest x-ray, and ECG, and clinical signs of progressive disease were evaluated. Patients in the observation arm were treated ac- cording to best supportive care, but no cancer-specific therapy was allowed until documented progression of disease. First clinical assessment was performed 6 weeks after random assignment. At 12 weeks and thereafter, the assessments were the same as in the vaccination arm.
In both arms, a full radiological reassessment was performed at disease progression.
Quality of Life.
Quality of life (QoL) was assessed in both arms by using the European Organisation for Research and Treatment of Cancer (EORTC) QOL-C30 version 3.0 complemented with the Lung Module (QLC-LC13) before random assignment, at weeks 6, 12, and 24 and every 6 months thereafter until progression.
Humoral response was assessed before each vaccination and at weeks 2, 6, and 12 after the last vaccination.
The serologic response to GD3 was assayed by an enzyme- linked immunosorbent assay (ELISA) using purified GD3. The GD3 ganglioside was dissolved in methanol (1 g/mL) and ad- sorbed to ELISA plates (Corning-Costar, Corning, NY) by evapo- ration at 37°C. Plates were rehydrated in 300 L of 4% human serum albumin (Calbiochem, San Diego, CA) in phosphate- buffered saline; then 100 L of diluted serum was incubated overnight at 4°C. After several washes with high-salt phosphate- buffered saline plus 0.1% human serum albumin and Tween-20, 100 L of polyclonal rabbit antihuman immunoglobulin (Ig)M or polyclonal rabbit antihuman IgG (Dako, Copenhagen, Denmark) appropriately diluted (1:700) in washing solution was incubated for 2 hours at 37°C. After several wash cycles, 100 L of polyclonal goat antirabbit immunoglobulins alkaline phosphatase-labeled antibody (Dako) appropriately diluted (1:2,000) in washing solu- tion was added to each well and incubated for 2 hours at 37°C. After several washing cycles, the reaction was developed with paranitrophenylphosphate (Calbiochem) and incubated for an- other 60 minutes at 37°C. The optical density of each well was read at 405 nm. The absorbance of the preimmune sera was subtracted from that of the immune sera to give the corrected absorbance. To eliminate the effect of nonspecific antibodies, the sera were also tested on ELISA plates, which were processed identically but to which no GD3 had been added. The background absorbance of this parallel assay was subtracted from the absorbance of each serum. Serologic titer in ELISA was defined as the biggest dilution yielding a corrected absorbance of 0.100 or greater.
For humoral response, a responder was defined as a patient with prevaccine and postvaccine blood samples for whom the baseline corrected IgM and/or IgG anti-GD3 titers of at least two consecutive postvaccination samples were at least 1:50 within the same Ig subclass.
Study Design and Statistics
The study initially was designed to register patients immedi- ately after diagnosis and before starting induction therapy. The protocol also mandated specific types of chemotherapies, which was later amended to register and randomly assign patients only after the completion of all induction therapies and confirmation of a partial or complete response. The time frame for randomization had to be within 8 months from diagnosis, and randomization had to occur between 3 and 7 weeks after all induction therapy had been given.
The primary end point of the study was overall survival. Secondary end points were progression-free survival, safety, QoL, and humoral response. The study was coordinated by the EORTC Data Center in Brussels, which is where the randomization took place. The randomization was performed by using the minimiza- tion technique, stratifying for Karnofsky performance status (60% to 70% v 80%), complete versus partial remission to induction therapy, and institution. The study was approved by the Protocol Review Committee of the EORTC and the medical ethical com- mittees of all participating institutions.
The study was powered to detect an increase in median survival of 40%, from 15 to 21 months from randomization, with a power of 90% and a two-sided type I error of 5%. The targeted number of events was 376; for that, 500 patients were to be accrued in 4 years and followed for 2 extra years. As foreseen by the study protocol, one interim analysis was performed after 108 deaths were reported. An spending function with an O’Brien-Fleming boundary was used to determine the nominal significance level to be used. Results of this interim analysis were only declared to an independent data-monitoring committee, which recommended to pursue the trial to its final accrual.
All analyses were performed according to intent-to-treat principles. Overall survival and progression-free survival were estimated by using the Kaplan-Meier method; a two-sided log- rank test was used, and P .05 was considered statistically signif- icant. Univariate and multivariate analyses were performed on overall survival to study the impact of the following potential prognostic factors: age at random assignment ( 59 v 59 years of age); continent (North America v Europe v Australasia); sex; Karnofsky performance status (60-80 v 90-100); response to in- duction (complete v partial response); sequential versus concom- itant radiotherapy; no PCI versus PCI during induction; PPD test (negative/doubtful v positive); baseline sodium level (140 v
140 mmol/L); baseline calcium level (2.4 v 2.4 mmol/L);
baseline white blood cell level (5.2 109/L v 5.2 109/L); baseline platelets level (221 109/L v 221 109/L); and baseline lactate dehydrogenase level (grade 0 v 0). For the multivariate analysis, a Cox multivariate proportional-hazards model was used with a step-down (backward) variable-selection procedure (at the 5% level). Both univariate and multivariate analyses were stratified for treatment arm.
A generalized linear mixed-model approach was used to an- alyze the QoL data, with treatment, time, and their interaction as fixed effects. The QoL analysis population was defined as all pa- tients with a valid baseline questionnaire and at least one valid questionnaire during the first 24 weeks.
Accrual to the study opened in March 1998 and closed in October 2002, with 515 patients from 120 institutions in 17
JOURNAL OF CLINICAL ONCOLOGY
Downloaded from jco.ascopubs.org on February 15, 2015. For personal use only. No other uses without permission. Copyright © 2005 American Society of Clinical Oncology. All rights reserved.