Risk of Thromboembolic Events and Anti-EGFR Agents
Our search yielded a total of 192 potentially relevant studies with C, P, erlotinib or gefitinib. Initially, 179 trials were excluded for at least one of the following reasons: absence of data on VTEs and ATEs, duplicate trials, phase I trials, nonrandomized trials, review articles, observational studies, case reports, editorials, letters and commentaries. Finally, 13 trials were considered highly relevant for the meta-analysis (phase II and III trials reporting VTEs and ATEs in the toxicity section of the publication). See supplemental Figure S1 (available at Annals of Oncology online) for flow diagram of trial's selection progress according to PRISMA guidelines.
The baseline characteristics of each trial are presented in Table 1. The most frequently reported adverse events were pulmonary embolism and venous thrombotic events, while only four trials reported cardiac events. Drug dosage and schedule were those currently approved (P 6 mg/kg every 14 days, C 400 mg/m loading dose followed by 250 mg/m weekly, erlotinib 150 mg/day, and gefitinib 250 mg/day) except for those of P in Vermorken and Okines trials, where a dose of 9 mg/kg was administered every 21 days.
A total of 7611 patients was available for the meta-analysis; 3778 of these patients were assigned to treatment arms and 3833 to placebo or control arms for the analysis of the RR.
A total of 7073 patients (11 trials) was considered for the incidence analysis of VTEs. VTEs occurred in 186 of 3508 patients, showing an incidence of 5% (compared with 3.7% in control arms). The incidence of VTEs was calculated separately for MoAbs and oral tyrosine kinase inhibitors (TKIs) (Table 2). The incidence was 5.9% for MoAbs and 2.6% for TKIs. The difference in the incidence of VTEs between MoAbs and TKIs trials was statistically significant (difference 3.3%; P < 0.001). The trials were also stratified by underlying malignancy (gastrointestinal versus lung). The incidence was then calculated and was 6.5% in gastrointestinal cancer patients and 3.3% in lung cancer patients. The difference between the two groups was statistically significant (P < 0.001).
A total of 3030 patients (five trials) was considered for the incidence analysis of ATEs. ATEs occurred in 69 of 1509 patients, showing an incidence of 4.5% (3.4% in control arms; P equals 0.11) (Table 3). The incidence was 4.6% for MoAbs and 1.9% for TKIs in one trial only. The difference in the incidence of ATEs between MoAbs and TKIs trials was not statistically significant (2.7% for MoAbs; P equals 0.37).
Eleven randomized studies were available to calculate the RR of VTEs in patients assigned to anti-EGFR agents versus controls in the same trial. The results are presented in supplemental Figure S2 (available at Annals of Oncology online). The meta-analysis showed that the summary RR of VTEs in experimental versus control arms was 1.32 (95% CI 1.07–1.63; P equals 0.01), suggesting a 32% higher risk of developing VTEs with anti-EGFR agents compared with controls. The test for heterogeneity was not significant [χ equals 7.64; df equals 10, (P equals 0.66); I equals 0%]. The risk difference was 1% [P equals 0.01; heterogeneity: χ equals 11.03, df equals 10, (P equals 0.36); I equals 9%]. All events except nine are of high grade (RR equals 1.36 for grades 3–4).
After stratifying patients by their underlying malignancy (supplemental Table S1, available at Annals of Oncology online), an RR of VTE of 1.31 (95% CI 0.84–2.04) was reported in lung cancer patients and an RR of 1.25 (95% CI 0.97–1.61) was reported in gastrointestinal cancer patients (P not significant for both). The correspondent RR reported in head and neck trials (n equals 2) was 2 (P equals 0.06). After stratifying patients by the class of drug used, an RR of VTE of 1.34 (95% CI 1.07–1.68) was reported in MoAbs-treated patients (P equals 0.01; heterogeneity: χ equals 3.70, df equals 7, (P equals .81); I equals 0%) and an RR of VTE of 1.16 (95% CI 0.61–2.18) was reported in oral TKIs-treated patients (P equals 0.65) (supplemental Figure S2 and Table S1, available at Annals of Oncology online). The risk difference for VTEs in MoAbs meta-analysis was 2% (P equals 0.01; heterogeneity: χ equals 2.81, df equals 7, (P equals 0.9); I equals 0%). After analysis of Crino' trial that reported only five cases of superficial phlebitis, we decided to perform the analysis after excluding this study. After exclusion of this trial, the overall result did not change (RR 1.37; 95% CI 1.37; P equals 0.04). The test for subgroup differences between MoAbs and TKIs studies (χ equals 0.28, df equals 1 (P equals 0.60), I equals 0%) remained positive (RR 1.34 for MoAbs, P equals 0.01; RR 1.64 for TKIs, P equals 0.17).
An exploratory analysis of RR of VTE calculated including trials with improved outcome associated with experimental arms (only C and P trials) showed a RR of 1.61 (95% CI 1.15–2.27; P equals 0.006). A similar result is obtained after inclusion of trials that included a cisplatin-based combination, known for its vasculotoxic properties (RR 1.69; 95% CI 1.06–2.07; P equals 0.03).
Additionally, a meta-regression was carried out to test whether the RR of VTEs varied as function of difference in length of the experimental and control treatments. Since in 3 studies data about length of treatment were not reported, 8 of 11 studies were included in the analysis (see Table 1). Results indicated that the RR tended to be higher in studies in which the experimental treatment was longer than the control treatment. However, this effect was not statistically significant (β equals 0.18; P equals 0.31).
Five randomized trials were available to calculate the RR of ATEs. The meta-analysis showed that the summary RR of ATEs in experimental versus control arms was a nonsignificant 1.34 (95% CI 0.94–1.9; P equals 0.11). The test for heterogeneity was not significant [χ equals 6.93; df equals 4, (P equals 0.14); I equals 42%]. After stratifying patients by the class of drug used, an RR of ATEs of 1.38 (95% CI 0.76–2.51) was reported in C- and P-treated patients (P equals 0.29), while an RR of 3.12 (95% CI 0.13–74.76) was reported in one trial with oral TKIs-treated patients (P equals 0.48). The risk of ATEs in head and neck cancer trials (n equals 2/5 trials) was 2.39 (95% CI 1.24–4.62; P equals 0.01) [heterogeneity: χ equals 1.48, df equals 1, (P equals 0.22); I equals 32%].
No evidence of publication bias was detected for the incidence or the RR of VTEs in this study by either Begg or Egger's tests (RR of VTEs: Begg's test P equals 0.5; Egger's test P equals 0.35). The absence of a 'dominant' study driving the results was demonstrated by the 'one-study removed' procedure that generated an overall similar RR estimate. Analogously, the trim and fill analysis did not show a publication bias for VTEs events.
Results
Our search yielded a total of 192 potentially relevant studies with C, P, erlotinib or gefitinib. Initially, 179 trials were excluded for at least one of the following reasons: absence of data on VTEs and ATEs, duplicate trials, phase I trials, nonrandomized trials, review articles, observational studies, case reports, editorials, letters and commentaries. Finally, 13 trials were considered highly relevant for the meta-analysis (phase II and III trials reporting VTEs and ATEs in the toxicity section of the publication). See supplemental Figure S1 (available at Annals of Oncology online) for flow diagram of trial's selection progress according to PRISMA guidelines.
The baseline characteristics of each trial are presented in Table 1. The most frequently reported adverse events were pulmonary embolism and venous thrombotic events, while only four trials reported cardiac events. Drug dosage and schedule were those currently approved (P 6 mg/kg every 14 days, C 400 mg/m loading dose followed by 250 mg/m weekly, erlotinib 150 mg/day, and gefitinib 250 mg/day) except for those of P in Vermorken and Okines trials, where a dose of 9 mg/kg was administered every 21 days.
A total of 7611 patients was available for the meta-analysis; 3778 of these patients were assigned to treatment arms and 3833 to placebo or control arms for the analysis of the RR.
Incidence of VTEs and ATEs
A total of 7073 patients (11 trials) was considered for the incidence analysis of VTEs. VTEs occurred in 186 of 3508 patients, showing an incidence of 5% (compared with 3.7% in control arms). The incidence of VTEs was calculated separately for MoAbs and oral tyrosine kinase inhibitors (TKIs) (Table 2). The incidence was 5.9% for MoAbs and 2.6% for TKIs. The difference in the incidence of VTEs between MoAbs and TKIs trials was statistically significant (difference 3.3%; P < 0.001). The trials were also stratified by underlying malignancy (gastrointestinal versus lung). The incidence was then calculated and was 6.5% in gastrointestinal cancer patients and 3.3% in lung cancer patients. The difference between the two groups was statistically significant (P < 0.001).
A total of 3030 patients (five trials) was considered for the incidence analysis of ATEs. ATEs occurred in 69 of 1509 patients, showing an incidence of 4.5% (3.4% in control arms; P equals 0.11) (Table 3). The incidence was 4.6% for MoAbs and 1.9% for TKIs in one trial only. The difference in the incidence of ATEs between MoAbs and TKIs trials was not statistically significant (2.7% for MoAbs; P equals 0.37).
Relative Risk of VTEs and ATEs
Eleven randomized studies were available to calculate the RR of VTEs in patients assigned to anti-EGFR agents versus controls in the same trial. The results are presented in supplemental Figure S2 (available at Annals of Oncology online). The meta-analysis showed that the summary RR of VTEs in experimental versus control arms was 1.32 (95% CI 1.07–1.63; P equals 0.01), suggesting a 32% higher risk of developing VTEs with anti-EGFR agents compared with controls. The test for heterogeneity was not significant [χ equals 7.64; df equals 10, (P equals 0.66); I equals 0%]. The risk difference was 1% [P equals 0.01; heterogeneity: χ equals 11.03, df equals 10, (P equals 0.36); I equals 9%]. All events except nine are of high grade (RR equals 1.36 for grades 3–4).
After stratifying patients by their underlying malignancy (supplemental Table S1, available at Annals of Oncology online), an RR of VTE of 1.31 (95% CI 0.84–2.04) was reported in lung cancer patients and an RR of 1.25 (95% CI 0.97–1.61) was reported in gastrointestinal cancer patients (P not significant for both). The correspondent RR reported in head and neck trials (n equals 2) was 2 (P equals 0.06). After stratifying patients by the class of drug used, an RR of VTE of 1.34 (95% CI 1.07–1.68) was reported in MoAbs-treated patients (P equals 0.01; heterogeneity: χ equals 3.70, df equals 7, (P equals .81); I equals 0%) and an RR of VTE of 1.16 (95% CI 0.61–2.18) was reported in oral TKIs-treated patients (P equals 0.65) (supplemental Figure S2 and Table S1, available at Annals of Oncology online). The risk difference for VTEs in MoAbs meta-analysis was 2% (P equals 0.01; heterogeneity: χ equals 2.81, df equals 7, (P equals 0.9); I equals 0%). After analysis of Crino' trial that reported only five cases of superficial phlebitis, we decided to perform the analysis after excluding this study. After exclusion of this trial, the overall result did not change (RR 1.37; 95% CI 1.37; P equals 0.04). The test for subgroup differences between MoAbs and TKIs studies (χ equals 0.28, df equals 1 (P equals 0.60), I equals 0%) remained positive (RR 1.34 for MoAbs, P equals 0.01; RR 1.64 for TKIs, P equals 0.17).
An exploratory analysis of RR of VTE calculated including trials with improved outcome associated with experimental arms (only C and P trials) showed a RR of 1.61 (95% CI 1.15–2.27; P equals 0.006). A similar result is obtained after inclusion of trials that included a cisplatin-based combination, known for its vasculotoxic properties (RR 1.69; 95% CI 1.06–2.07; P equals 0.03).
Additionally, a meta-regression was carried out to test whether the RR of VTEs varied as function of difference in length of the experimental and control treatments. Since in 3 studies data about length of treatment were not reported, 8 of 11 studies were included in the analysis (see Table 1). Results indicated that the RR tended to be higher in studies in which the experimental treatment was longer than the control treatment. However, this effect was not statistically significant (β equals 0.18; P equals 0.31).
Five randomized trials were available to calculate the RR of ATEs. The meta-analysis showed that the summary RR of ATEs in experimental versus control arms was a nonsignificant 1.34 (95% CI 0.94–1.9; P equals 0.11). The test for heterogeneity was not significant [χ equals 6.93; df equals 4, (P equals 0.14); I equals 42%]. After stratifying patients by the class of drug used, an RR of ATEs of 1.38 (95% CI 0.76–2.51) was reported in C- and P-treated patients (P equals 0.29), while an RR of 3.12 (95% CI 0.13–74.76) was reported in one trial with oral TKIs-treated patients (P equals 0.48). The risk of ATEs in head and neck cancer trials (n equals 2/5 trials) was 2.39 (95% CI 1.24–4.62; P equals 0.01) [heterogeneity: χ equals 1.48, df equals 1, (P equals 0.22); I equals 32%].
Publication Bias
No evidence of publication bias was detected for the incidence or the RR of VTEs in this study by either Begg or Egger's tests (RR of VTEs: Begg's test P equals 0.5; Egger's test P equals 0.35). The absence of a 'dominant' study driving the results was demonstrated by the 'one-study removed' procedure that generated an overall similar RR estimate. Analogously, the trim and fill analysis did not show a publication bias for VTEs events.
SHARE
![Pediatric Supportive Care (PDQ®): Supportive care - Patient Information [NCI]-Overview](https://cdnus.oss-us-west-1.aliyuncs.com/libs/images/NewsJpg/cate_6/072_215x160.jpg)
