2. Methods used to impute standard deviations for changes values.

Study	Group	Outcome measure	P value	Imputation	Imputed SDs	Implied correlation	Plausibility of imputation
Driva 19851	Intervention	PDI	> 0.05	Impossible	10.94	0.69	Plausible
		MDI	< 0.01	Possible	10.94	0.78	Plausible
	Control	PDI	> 0.05	Impossible	10.94	0.82	Plausible
		MDI	> 0.05	Impossible	10.94	0.79	Plausible
Honig 19782	Intervention	PDI	0.10	Possible	21.22	‐0.26	Plausible
		MDI	0.01	Possible	15.14	‐0.27	Plausible
	Control	PDI	> 0.05	Impossible	21.22	0.20	Plausible
		MDI	> 0.05	Impossible	15.14	0.57	Plausible
Kimmons [pers comm]3	Intervention	PDI	0.36	Possible	9.73	0.79	Plausible
		MDI	0.27	Possible	3.90	0.97	Plausible
	Control	PDI	0.36	Possible	9.73	0.79	Plausible
		MDI	0.27	Possible	3.90	0.97	Plausible
Lozoff 19824	Intervention	PDI	‐	Impossible	7.4	0.92	Plausible
		MDI	‐	possible	7.4	0.93	Plausible
	Control	PDI	‐	Impossible	7.4	0.94	Plausible
		MDI	‐	possible	7.4	0.998	Plausible
Lozoff 19875	Intervention	PDI	> 0.80	Impossible	‐	‐	‐
		MDI	> 0.69	Impossible	‐	‐	‐
	Control	PDI	> 0.80	Impossible	‐	‐	‐
		MDI	> 0.69	Impossible	‐	‐	‐
Idjradinata 19936	Intervention	PDI	< 0.001	Possible	17.91	‐0.59	Plausible
		MDI	< 0.001	Possible	25.09	‐1.84	Implausible
	Control	PDI	< 0.001	Possible	17.91	‐0.51	Plausible
		MDI	‐	Impossible	25.09	‐1.70	Implausible

MDI: Mental Development Index; PDI: Psychomotor Development Index; SD: standard deviation.

¹ P values resulted from paired t‐tests (for within group changes). The SD for the change MDI in the intervention group (10.94) was used for the SDs of the change in PDI in both groups and the change MDI in the control group in the analyses, which were impossible to be imputed.

² P values resulted from paired t‐tests (for within group changes). The SD for the change PDI in the intervention group (21.22) and for the change in MDI (15.14) were used for the SDs of the change in PDI and MDI in the control group, respectively, which were impossible to be imputed.

³ P values derived from a repeated measures analysis of variance were treated as if resulting from unpaired t‐tests (for the between group differences in changes).

⁴ Here the SDs for the change in MDI were imputed from the reported power of the study to detect differences of specified sizes in MDI given the observed SDs of the changes (although not giving the values of these SDs), rather than from P values. The SDs for the change MDI were used for the SDs of the change in PDI in the analyses, which were impossible to be imputed.

⁵ P values derived from a repeated measures analysis of variance were treated as if resulting from unpaired t‐tests (for the between group differences in changes).

⁶ P values were derived from a repeated measures analysis of variance. For PDI, P values were treated as if resulting from unpaired t‐tests (for the between group differences in changes). For MDI, P values were treated as if resulting from paired t‐tests (for within group changes). The SD for the change MDI in the intervention group (25.09) was used for the SDs of the change in MDI in the control group, which was impossible to be imputed. However, the estimated SDs for change in MDI would imply correlations between pre‐ and post‐test scores substantially in excess of ‐1, which is impossible. It was, therefore, decided instead to estimate SDs from these reported scores and standard errors, following methods described by Follmann (Follmann 1992), assuming a correlation of 0 between pre‐ and post‐treatment test scores. This provided an estimate of the SD for the change in MDI of 15.26 for the control group and 14.90 for the intervention group and of the SD for the change in PDI of 14.21 for the intervention group and 14.58 for the control group.