Table 2. The association between risk of malaria and ML ratio is independent of age and antibodies to parasite blood-stage antigens.

		PARASITE NEGATIVE (N  =  207)		PARASITE POSITIVE (N  =  81)
Analysis type	Variable	IRR (95% CI)	P value	IRR (95% CI)	P value
Univariate	ML ratio	0.8 (0.55, 1.17)	0.3	2.8 (1.48, 5.12)	0.001
	Age	0.9 (0.88, 0.96)	0.0002	0.8 (0.69, 0.96)	0.02
	AMA1 antibodies	0.3 (0.17, 0.64)	0.001	0.2 (0.10, 0.55)	0.0007
	MSP2 antibodies	1.4 (1.01, 1.81)	0.04	0.8 (0.46, 1.56)	0.6
	MSP3 antibodies	0.8 (0.29, 2.01)	0.6	0.3 (0.17, 0.69)	0.002
	Parasite schizont extract	1.5 (1.20, 1.78)	0.0001	1.2 (0.72, 2.02)	0.5
Multivariate	ML ratio	0.9 (0.61, 1.20)	0.4	2.2 (1.16, 4.28)	0.02
	Age	0.9 (0.87, 0.96)	0.0002	0.8 (0.75, 0.96)	0.01
	AMA1 antibodies	0.4 (0.17, 0.91)	0.03	0.3 (0.13, 0.83)	0.02
	MSP2 antibodies	1.1 (0.75, 1.49)	0.7	1.1 (0.58, 2.09)	0.8
	MSP3 antibodies	0.5 (0.26, 1.08)	0.08	0.4 (0.22, 0.85)	0.01
	Parasite schizont extract	1.7 (1.35, 2.04)	0.000002	1.5 (1.00, 2.40)	0.05

Presented are incidence rate ratios (IRR) and 95% confidence intervals (CI) from negative binomial regression models predicting the total number of malaria episodes between sampling at May 2008 baseline survey and 31^st December 2011 using ML ratio, age, antibodies to AMA1, MSP2 and MSP3 and to parasite schizont extract, used routinely as a control for previous parasite exposure in antibody assays [8]. Univariate analysis is done using each variable in turn whereas all variables are included in the multivariate model and the respective results from each variable shown. As with the complete dataset on which the antibody data is based a statistically significant interaction between parasite positive/negative status at baseline and ML ratio was evident (IRR  =  2.7, 95%CI 1.32, 5.47, P  =  0.006).