Table 3.

The effect of each blood pack type supplied for transfusion

			Adjusted estimate (95% CI) for packed cells versus whole blood packs	Adjusted estimate (95% CI) for settled cells versus whole blood packs	Overall Wald test p value
Time to event analyses (hazard ratio)
28-day mortality*			0·99 (0·48 to 2·04)	1·12 (0·64 to 1·95)	0·92
180-day mortality*			1·11 (0·66 to 1·85)	1·05 (0·75 to 1·46)	0·91
	Readmissions
		All cause†	1·05 (0·66 to 1·65)	0·85 (0·68 to 1·06)	0·30
		Anaemia†	1·49 (0·87 to 2·57)	0·90 (0·69 to 1·16)	0·18
		Haemoglobinuria†	0·61 (0·12 to 3·02)	1·08 (0·70 to 1·67)	0·76
		Malaria†	0·59 (0·25 to 1·39)	0·70 (0·47 to 1·02)	0·13
	Time to discharge from hospital†		0·94 (0·81 to 1·10)	0·86 (0·79 to 0·94)	0·0024
Mean change in haemoglobin analyses
8 h haemoglobin, g/dL (20 mL/kg)‡			−1·3 (−1·5 to −1·0)	−1·1 (−1·2 to −0·9)	<0·0001
8 h haemoglobin, g/dL (30 mL/kg)‡			−1·4 (−1·6 to −1·1)	−1·5 (−1·7 to −1·3)	<0·0001
180-day haemoglobin,‡ g/dL			0·0 (−0·4 to 0·4)	0·1 (−0·1 to 0·2)	0·57
Second transfusions (odds ratio)
Odds of second transfusion§			2·32 (1·30 to 4·12)	2·97 (2·18 to 4·05)	<0·0001

All models were adjusted for site and each model was checked for heterogeneity of effect by volume randomisation.

^*Analysed using Cox regression models.

^†Competing risks regression models. For readmission, the competing event was death post-discharge and for time to discharge, the competing event was death during initial admission.

^‡Linear regression models.

^§Logistic regression model.