Six RCTs were included with 1,219 participants.
Tumour response rate was higher in patients assigned to 5-FU CI than in patients assigned to 5-FU bolus (22% (CR = 3% and PR = 19%) versus 14% (CR = 2% and PR = 12%); overall OR 0.55, 95% CI: 0.41, 0.75, p = 0.0002). This was equivalent to a risk reduction of 45% with a standard error of 12%.
Median duration of tumour response was 7.1 months in the 5-FU CI arm (95% CI: 5.7, 8.5 months) and 6.7 months in the 5-FU bolus arm (95% CI: 5.7, 8.5 months).
In the group of trials that used a biochemical modulation of 5-FU by LV (2 trials) the difference between continuous 5-FU plus LV and bolus 5-FU plus LV failed to reach statistical significance (tumour response OR = 0.82, 95% CI: 0.33, 2.07).
Overall survival was also higher in patients assigned to 5-FU CI (overall hazard ratio (HR) 0.88, 95% CI: 0.78, 0.99, p = 0.04). The median survival duration was 12.1 months (95% CI: 11, 13.1 months) in the 5-FU CI group versus 11.3 months (95% CI: 10.5, 12 months) in the 5-FU bolus group. The number of patients still alive in the 5-FU CI group versus the 5-FU bolus group were, respectively, 99 versus 102 at 2 years, 39 versus 23 at 3 years, and 16 versus 6 at 4 years.
When 5-FU was modulated by LV (2 trials) overall survival did not appear to be better for patients assigned to continuous 5-FU plus LV and 5-FU bolus plus LV (HR = 1.03, 95% CI: 0.77, 1.38) which was not statistically significant.
Multivariate analysis showed that randomised treatment and performance status were the only two significant predictors of tumour response, whereas the same plus primary tumour site were independent significant predictors of survival (patients with rectal cancer did somewhat better).
Grade 3 or 4 hematologic toxicity was more frequent in patients assigned to 5-FU bolus (31% versus 4%, p < 10(-16)), whereas hand- foot syndrome was more frequent in the 5-FU CI group (34% versus 13%, p < 10(-7)).