Summary of findings for the main comparison. Phonics training versus control (no training or alternative training) for English‐speaking poor readers.

Phonics training versus control (no training or alternative training) for English‐speaking poor readers
Patient or population: English‐speaking poor readers Setting: English‐speaking countries Intervention: phonics Comparison: control (no training or alternative training)
Outcomes	Illustrative comparative risks (SMD* 95% CI*)		Relative effect (95% CI)	N° of participants (studies)	Quality of the evidence (GRADE)*	Comments
	Assumed risk	Corresponding risk
	Control (no training or alternative training)	Phonics training
Mixed/regular word reading accuracy Assessed with: various scales Follow‐up: immediate	—	The mean score in the intervention groups was 0.51 standard deviations higher (0.13 higher to 0.90 higher)	—	701 (11 studies)	⊕⊕⊝⊝ Lowa	A standard deviation of 0.51 represented a moderate effect between groups. Phonics training "may improve" outcome (Ryan 2016).
Non‐word reading accuracy Assessed with: various scales Follow‐up: immediate	—	The mean score in the intervention groups was 0.67 standard deviations higher (0.26 higher to 1.07 higher)	—	682 (10 studies)	⊕⊕⊝⊝ Lowa	A standard deviation of 0.67 presented a moderate effect between groups. Phonics training "may improve" outcome (Ryan 2016).
Irregular word reading accuracy Assessed with: various scales Follow‐up: immediate	—	The mean score in the intervention groups was 0.84 standard deviations higher (0.30 higher to 1.39 higher)	—	294 (4 studies)	⊕⊕⊕⊝ Moderatea,c	A standard deviation of 0.84 presented a large effect between groups. Phonics training "probably improves" outcome (Ryan 2016).
Mixed/regular word reading fluency Assessed with: various scales Follow‐up: immediate	—	The mean score in the intervention groups was 0.45 standard deviations higher (0.19 higher to 0.72 higher)	—	224 (4 studies)	⊕⊕⊕⊝ Moderateb	A standard deviation of 0.45 presented a moderate effect between groups. Phonics training "probably improves" outcome (Ryan 2016).
Non‐word reading fluency Assessed with: various scales Follow‐up: immediate	—	The mean score in the intervention groups was 0.39 standard deviations higher (0.10 higher to 0.68 higher)	—	188 (3 studies)	⊕⊕⊕⊝ Moderateb	A standard deviation of 0.39 presented a moderate effect between groups. Phonics training "probably improves" outcome (Ryan 2016).
Reading comprehension Assessed with: various scales Follow‐up: immediate	—	The mean score in the intervention groups was 0.28 standard deviations higher (0.07 lower to 0.62 higher)	—	343 (5 studies)	⊕⊕⊝⊝ Lowa	A standard deviation of 0.28 presented a small effect between groups. Phonics training "may improve" outcome (Ryan 2016).
Spelling Assessed with: various scales Follow‐up: immediate	—	The mean score in the intervention groups was 0.47 standard deviations higher (0.07 lower to 1.01 higher)	—	158 (3 studies)	⊕⊕⊝⊝ Lowa	A standard deviation of 0.47 presented a moderate effect between groups. Phonics training "may improve" outcome (Ryan 2016).
SMD: standardised mean difference. Different studies used different continuous measures. Thus, effect sizes are reflected by size of phonics training effect as indexed using SMDs. The results are expressed as standard deviation (SD) units. As a general rule, 0.2 SMD represents a small effect size, 0.5 a moderate effect size, and 0.8 a large effect size. CI: confidence interval. GRADE: Working Group grades of evidence High quality: we are very confident that the true effect lies close to the that of the estimate of the effect. Moderate quality: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low quality: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of effect. Very low quality: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.

^aDowngraded two levels due to very serious imprecision: very wide confidence intervals (greater than 0.6; Schünemann 2011b).
 ^bDowngraded one level due to serious imprecision: wide confidence intervals (0.3 to 0.6; Schünemann 2011b).
 ^cUpgraded one level due to large effect: SMD greater than 0.8 (Ryan 2016).