Six RCTs (n=1,716) were included.
Five RCTs scored 7 out of 8 points for quality and one (which reported end points for less than 85% of participants) scored 6. In all studies intention-to-treat analysis was used and prognostic factors were balanced between the groups at baseline. None of the 3 smaller studies included in the review achieved their planned power, owing to accrual problems
There was a statistically significant benefit for the IP group compared with the IV group in progression-free survival (HR 0.792, 95% CI: 0.688, 0.912, p=0.001; 4 RCTs, n= 1,052) and overall survival (HR 0.799, 95% CI: 0.702, 0.910, p=0.0007; 6 RCTs, n=1,716). There was no evidence of statistically significant heterogeneity or publication bias, and it was calculated that 6 negative studies would be required to negate the significance of the results of overall survival. Sensitivity analyses did not change the statistical significance of any of the results.
Toxicity (grade 3 or 4).
When the data were pooled, there was a statistically significant increase in the IP group in the risk of grade 3 or 4 gastrointestinal symptoms (OR 1.95, 95% CI: 1.17, 3.24, p=0.01) and fever (OR 1.7, 95% CI: 1.02, 2.84, p=0.04; 4 RCTs). In the IV group there was a statistically significant increase in the risk of ototoxicity (OR 0.38, 95% CI: 0.19, 0.73, p=0.004; 2 RCTs). An assessment of potential publication bias showed that an additional 18 negative studies would be required to change the significance of the results for gastrointestinal toxicities, while 4 negative studies would be required in the case of fever.
There was no statistically significant difference between the groups in the incidence of leukopenia (6 RCTs), haemoglobin toxicity (4 RCTs), platelet toxicity (6 RCTs), neurotoxicity (5 RCTs) or treatment-related death (3 RCTs).
Catheter-related complications were not reported consistently in the included studies. Such complications occurred among 19.5% of the 205 women in the IP arm of one RCT, the commonest problems being infection (n=21) and blockage (n=10). Another RCT reported that 8.7% of the 56 women in the IP arm experienced catheter blockage.
Heterogeneity of classifications precluded pooling of data for some haematological toxicities.
Quality of life.
The only RCT to report this outcome (n=415) reported significantly higher toxicity and quality-of-life disturbances in the IP arm during treatment. However, quality-of-life scores were equivalent in both groups one year later.