Eight trials (3,736 participants) with nine pair-wise comparisons were included in the review. Seven of the eight trials were considered of high quality (Jadad score >2). Four trials were published and four were in abstract form. Median follow-up ranged from 8.3 to 20.5 months, where reported.
Efficacy: Compared with observation or placebo, switch maintenance significantly improved overall survival (HR 0.85, 95% CI 0.79 to 0.92; six trials with no significant heterogeneity) and continuous maintenance was associated with a trend suggesting improved overall survival (HR 0.88, 95% CI 0.74 to 1.04; three trials with no significant heterogeneity). The interaction test suggested that there was no difference between continuous and switch maintenance therapy (interaction P=0.77). Compared with observation or placebo, both maintenance therapies were associated with significantly greater progression-free survival (switch HR 0.67, 95% CI 0.57 to 0.78, six trials with significant heterogeneity and continuous HR 0.53, 95% CI 0.43 to 0.65; two trials with no significant heterogeneity). The effect of continuous maintenance therapy on progression-free survival was not significantly different from switch therapy (interaction P=0.12). There was no evidence of statistically significant differences in overall survival and progression-free survival between switch maintenance therapy with cytotoxic agents and that with tyrosine kinase inhibitor agents.
Safety: Compared with observation or placebo, continuous maintenance therapy was associated with significantly greater rates of neutropenia (OR 14.6, 95% CI 2.6 to 82.4) and anaemia (OR 4.1, 95% CI 1.3 to 12.6), but not thrombocytopenia. Compared with observation or placebo, switch maintenance therapy was associated with significantly greater rates of alanine aminotransferase level (OR 6.4, 95% CI 1.2 to 35.1), diarrhoea (OR 5.8, 95% CI 1.3 to 25.4), rash (OR 13.4, 95% CI 2.0 to 89.0), fatigue (OR 8.4, 95% CI 1.9 to 37.2) and infection (OR 8.4, 95% CI 1.1 to 65.7). There was no evidence of significant differences between switch therapy and control in rates of neutropenia, anaemia, vomiting and anorexia, but the numbers of patients who experienced these adverse events were greater in the switch therapy group. No significant heterogeneity identified from Χ², but some adverse events had Ι² values greater than 25%.
There was no evidence of publication bias. Sensitivity analyses did not markedly change the estimates.