Table 1.

Conditional logistic regression analysis of matched severe malaria–community control pairs

Groups	Number of pairs
	Cases	Controls			Odds ratio (95% CI) [P value]
		−α/−α	−α/αα	αα/αα	−α/−α	−α/αα
Severe malaria group (249 pairs)	−α/−α	72	35	4	0.40  (0.22–0.74)	0.66  (0.37–1.2)
	−α/αα	48	28	17	[0.003]	[0.2]
	αα/αα	22	15	8
Clinical subgroups
Severe anemia (155)	−α/−α	37	16	3	0.34  (0.16–0.73)	0.84  (0.41–1.7)
	−α/αα	36	20	12	[0.004]	[0.6]
	αα/αα	15	10	6
Coma (56)	−α/−α	19	10	0	0.90  (0.19–4.3)	1.27  (0.31–5.2)
	−α/αα	9	6	5	[0.90]	[0.7]
	αα/αα	2	4	1
Hypoglycemia (13)	−α/−α	2	2	0	1.75  (0.023–135)	4.90  (0.078–307)
	−α/αα	5	2	1	[0.8]	[0.4]
	αα/αα	1	0	0
Acidosis (42)	−α/−α	19	4	0	0  (0–0.028)	0  (0–0.14)
	−α/αα	9	3	0	[0.0006]	[0.005]
	αα/αα	3	2	2
Hyperlactatemia (62)	−α/−α	22	9	1	0.25  (0.057–1.1)	0.32  (0.070–1.4)
	−α/αα	9	6	4	[0.06]	[0.1]
	αα/αα	5	4	2

Conditional logistic regression analysis was performed with case-control status as the outcome variable and age and ethnicity included in the regression model. The middle columns show case-control pairs according to genotype for α⁺-thalassemia. In the middle columns, only pairs discordant in genotype (shown in bold type) contribute to the analysis. An approximate assessment of the degree of protection afforded by α⁺-thalassemia can be obtained from these raw data. For example, considering the 26 homozygote/normal pairs in the severe malaria group, the homozygous member of the pair was the control in 22 and the case in only four pairs whereas an approximately equal distribution would be expected if homozygous α⁺-thalassemia conferred no protection against severe malaria. However, more accurate estimates are obtained by using all the discordant pairs in a single analysis, because the three comparisons of homozygote vs. normal, heterozygote vs. normal, and homozygote vs. heterozygote are mutually dependent. The odds ratios shown represent the risk of developing severe malaria for homozygotes and heterozygotes relative to normals. To investigate the protective effect of α⁺-thalassemia against individual manifestations of severe malaria, conditional logistic regression analysis was repeated selecting only the pairs in which the case had a particular severe manifestation of malaria. Median (range) age was 2.1 (0.2–12.4) years for children with severe anemia, 3.5 (0.8–11.7) for coma, 3.0 (0.5–7.0) for hypoglycemia, 2.4 (0.5–9.5) for acidosis, and 2.6 (0.2–9.9) for hyperlactatemia. Many children had more than one severe manifestation of malaria (12). A sole severe manifestation of malaria was detected in 145 of 179 children in whom it was possible to assess all severe manifestations (103 with severe anemia, 24 with coma, 3 with hypoglycemia, 14 with acidosis, and 1 with hyperlactatemia).