Table 3. Sensitivity analysis of odds ratio^a vs hazard ratio^b for estimating mortality in chemotherapy studies with long-term follow-up.

Number of studies analysed	Studies excluded	Summary measure	Point estimate	95% CI
20	None	Study-level OR	1.05	0.92–1.19
18	Engert, 2007; Strauss et al, 2008	Study-level HR	1.04	0.95–1.14
16	Engert, 2007 Osterborg et al, 2005; Aapro et al, 2008a; Strauss et al, 2008	Study-level OR	1.04	0.89–1.21
16	Engert, 2007; Osterborg et al, 2005 Aapro et al, 2008a; Strauss et al, 2008	Study-level HR	1.04	0.94–1.15
16	Engert, 2007 Osterborg et al, 2005 Aapro et al, 2008a; Strauss et al, 2008	Patient-level HR	1.03	0.95–1.11

CI=confidence interval; HR=hazard ratio; OR=odds ratio.

^aORs were calculated for 20 studies with long-term follow-up data, for 16 studies where primary data were available and for 18 studies with either a reported or calculated HR. The OR was based only on the number of deaths.

^bHRs were calculated for 16 studies where primary data were available. Both study-level and patient-level analyses are provided.