Fifty-six patients were enrolled between August 2006 and April 2008 at eight sites in the Phase II Consortium. The data are reported as of November 2009. Three patients were ineligible after starting treatment (brain metastases, alteration in pathologic diagnosis, colitis) and were included in the toxicity analysis but not the primary end point analysis. Median age was 63 years (range, 31 to 87 years). A majority of patients were female (30 patients, 57%) and white (48 patients, 91%). At study entry, 52 patients (98%) had a performance status of 0 or 1, 43 patients (81%) had cholangiocarcinoma, and 10 patients (19%) had gallbladder cancer. Metastatic disease was seen in 58% of patients, with the liver being the most common site of metastasis (13 patients; ).
Baseline Characteristics of Eligible Patients (N = 53)
Efficacy and Patient Outcome
Fifty-three patients completed a total of 327 cycles of treatment (median, 4 cycles; range, 1 to 33 cycles). The primary end point was confirmed tumor response. Forty-nine patients were evaluable for assessing response. Nine patients achieved a best response of PR while six patients (12%; 95% CI, 6% to 27%; five at University of Wisconsin Carbone Cancer Center, one at Mayo Clinic [Rochester, MN]) had prolonged responses confirmed 4 weeks after their initial response was observed. Each PR was reviewed and confirmed by an independent investigator. Twenty-five and 15 patients achieved a best response of SD and PD, respectively. Among the six patients with confirmed PRs, median duration of response was 8.4 months (95% CI, 6.0 to 11.7 months). At the time of data cutoff, three patients remained on study medication, having received 18, 29, and 33 cycles of therapy. Eighty-seven percent of patients progressed with a median time to disease progression of 4.4 months (95% CI, 3.0 to 7.8 months). Median OS was 9.9 months (95% CI, 7.2 to 13.6 months; , ).
Time to progression and overall survival.
Grade 4 adverse events which were at least possibly related to study treatment were experienced by four patients (8%), including cerebral ischemia and thrombosis (two patients each). Fourteen patients (27%) experienced grade 3 adverse events considered at least possibly related, including rash/desquamation (three), anorexia (three), fatigue (three), hyponatremia (three), nausea (three), ALT (one), bilirubin (one), diarrhea (one), dizziness (one), hypertension (one), nail changes (one), prothrombin time (one), and alkaline phosphatase (one). Details on toxicity are provided in . The most common of these events included (number of patients experiencing grade 1, 2, or 3): rash/desquamation (16, 21, three), diarrhea (15, six, one), fatigue (18, six, two), nausea (seven, four, two), anorexia (six, five, two), and oral mucositis (10, two, zero).
Maximum Severity of Adverse Events (N = 53)
Two patients (4%) died during treatment of causes felt to be unrelated to study treatment. A 79-year-old male died suddenly on day 21 of the first cycle, experiencing grade 4 cerebral ischemia (possibly related) and grade 1 blood bilirubin increase (possibly related). Bevacizumab was held on day 15. This death was considered unlikely to be related to study treatment. A 69-year-old male died within 30 days of initiating treatment, possibly due to disease progression. Adverse events before death included unrelated grade 3 anorexia, dyspnea, epigastric and back pain, and nausea; treatment-related grade 2 rash/desquamation combined; and unrelated grade 1 vomiting. This patient received full doses of both study agents during treatment.
Evaluable tissue was submitted from 26 patients, four of whom had a confirmed response. Characteristics of this subset were comparable to the larger study population in terms of age, sex, performance status, and enrolling site. Fewer patients had gallbladder tumors (8% v 19%) or poorly differentiated tumors (4% v 15%) in the subset with available tissue (Appendix Table A1, online only).
Mutation on EGFR vIII, ≤ 16 C>A repeats for the EGFR intron 1 polymorphism, a G>G genotype measured by EGFR-Q787 single nucleotide polymorphisms, and wild-type (ie, nonmutant) KRAS measured by 38G primer I were hypothesized to be positively correlated with patient outcome (ie, confirmed response, tumor progression, and survival). Although only four patients experienced a confirmed response in this group and the results were not statistically significant, the data appeared to be in the direction of our hypotheses (Appendix Tables A2 and A3, online only), albeit with a limited sample size. VEGF expression did not change significantly from baseline between the responding and nonresponding patients (Appendix Table A2 and Appendix Fig A1, online only).
Desirable results (responders v nonresponders) were most often observed in patients having low repeats (≤ 16) for the EGFR intron 1 polymorphism (75% v 55%) or G>G K-ras Q38 genotype (100% v 82%). In univariate models, EGFR-Q787 genotype of G>G trended toward lower hazard rates for TTP (hazard ratio [HR], 0.7; 95% CI, 0.3 to 2.0; P = .56) and survival (HR, 0.6; 95% CI, 0.2 to 1.7; P = .34), although it was not statistically significant. VEGF 38Q genotype of G>G (ie, nonmutant) may be associated with a lower hazard rate for TTP (HR, 0.4; 95% CI, 0.1 to 1.4; P = .13). Conversely, EGFR vIII mutation is suggestive of worsened TTP (HR, 2.0; 95% CI, 0.6 to 7.1; P = .27) and survival (HR, 1.7; 95% CI, 0.5 to 6.1; P = .39; Appendix Table A3).