Twenty-nine trials were included for analysis, with a total of 12,079 patients. The number of patients in each trial was not reported.
Quality of trials: Mean quality score was 48.6% (+/- 16.2 standard deviation). Correlation between the two raters was good (kappa = 0.50, p=0.02).
Heterogeneity: Using the fixed model, the statistical test for heterogeneity showed no significant difference in any case. The chi-square value was 33.23 when combining all trials (p=0.02). When using the random model, six groupings showed heterogeneity. They were: pooling of all 29 trials, pooling of all trials with quality score >30, pooling of all trials with quality score >50 and including colorectal trial, pooling of all trials using 5-fluorouracil, pooling of colon trials with quality scores >50 and pooling of unicentric trials.
Effect sizes, Random Model (after heterogeneity adjustment):
The values given below are total odds ratio (95% confidence interval).
All trials pooled (29 trials): 0.90 (0.81,0.99).
Trials with quality score >30 pooled (25 trials): 0.87 (0.80,0.95).
Trials with quality score >50,including colorectal trial pooled (14):0.85 (0.71,1.04).
Trials using 5-fluorouracil pooled (14): 0.80 (0.76,0.85)
Trials using 5-fluorouracil plus other drugs pooled (9): 0.80 (0.76,0.85).
Trials with quality score >50, including only colon, pooled (8) 0.80 (0.70,0.90)
Trials using 5-fluorouracil and other drugs for >48 weeks, including only colon, pooled (5): 0.81 (0.69,0.94).
Trials with quality score >30, including only rectal, pooled (3): 0.64 (0.48,0.85)
Unicentric trials, pooled (5):1.06 (0.86,1.3)
Effect size weighted for quality score:
All trials pooled (29 trials): Mantel-Haenszel 0.04, Fleiss and Gross 0.06.
Trials with quality score >50 and including colorectal trial pooled (14): Mantel-Haenszel 0.06, Fleiss and Gross 0.06.
Trials using 5-fluorouracil pooled (14): Mantel-Haenszel 0.09, Fleiss and Gross 0.11.
Trials with quality score >50, including only colon, pooled (8): Mantel-Haenszel 0.08, Fleiss and Gross 0.07.
Trials using 5-fluorouracil and other drugs for >48 weeks, including only colon, pooled (5): Mantel-Haenszel 0.08, Fleiss and Gross 0.08.
Trials with quality score >30, including only rectal, pooled (3): Mantel-Henszel 0.2, Fleiss and Gross 0.22.
Unicentric trials, pooled (5): Mantel-Haenszel -0.05, Fleiss and Gross -0.04.
Sensitivity analysis:
Rectal cancer (3 trials, 695 patients): Effect size = 0.22, % increase of 5-year survival = 9%.
Colon cancer (25 trials, 11673 patients): = 0.05, % increase of 5-year survival = 2%.
Quality score >50, + 5-FU for more than 48 weeks: = 0.11, % increase of 5-year survival = 5%.
Quality score effects:
>=30 (25 trials, 11673 patients): Effect size = 0.05, % increase 5-year survival = 2%.
>=50 (16 trials, 7926 patients): Effect size = 0.07, % increase of 5-year survival = 3%.
>=60 (5 trials, 2749 patients): Effect size = 0.08, % increase of 5-year survival = 3.5%.
Unicentric (30<=, 5 trials, 663 patients): Effect size = -0.04, % increase of 5-year survival = -2%.