Effect of sleep position in term healthy newborns on sudden infant death syndrome and other infant outcomes: A systematic review

Background Though recommended by numerous guidelines, adherence to supine sleep position during the first year of life is variable across the globe. Methods This systematic review of randomized trials and observational studies assessed the effect of the supine compared to non-supine (prone or side) sleep position on healthy newborns. Key outcomes were neonatal mortality, sudden infant death syndrome (SIDS), sudden unexpected death in infancy (SUDI), acute life-threatening event (ALTE), neurodevelopment, and positional plagiocephaly. We searched MEDLINE via PubMed, Cochrane CENTRAL, EMBASE, and CINAHL (updated till November 2021). Two authors separately evaluated the risk of bias, extracted data, and synthesised effect estimates using relative risk (RR) or odds ratio (OR). The GRADE approach was used to assess the certainty of evidence. Results We included 54 studies (43 observational studies and 11 intervention trials) involving 474 672 participants. A single study meeting the inclusion criteria suggested that the supine sleep position might reduce the risk of SUDI (0-1 year; OR = 0.39, 95% confidence interval (CI) = 0.23-0.65; 384 infants), compared to non-supine position. Supine sleep position might reduce the risk of SIDS (0-1 year; OR = 0.51, 95% CI = 0.42-0.61; 26 studies, 59332 infants) and unexplained SIDS/severe ALTE (neonatal period; OR = 0.16, 95% CI = 0.03-0.82; 1 study, 119 newborns), but the evidence was very uncertain. Supine sleep position probably increased the odds of being 0.5 standard deviation (SD) below mean on Gross Motor Scale at 6 months (OR = 1.67, 95% CI = 1.22-2.27; 1 study, 2097 participants), but might have little to no effect at 18 months of age (OR = 1.16, 95% CI = 0.96, 1.43; 1 study, 1919 participants). An increase in positional plagiocephaly at 2-7 months of age with supine sleep position is possible (OR = 2.77, 95% CI = 2.06-3.72; 6 studies, 1774 participants). Conclusions Low- to very low-certainty evidence suggests that supine sleep position may reduce the risk of SUDI (0-1 year) and SIDS (0-1 year). Limited evidence suggests that supine sleeping probably delays short-term ‘gross motor’ development at 6 months, but the effect on long-term neurodevelopment at 18 months may be negligible.

The association of infant sleep position with neurodevelopment outcomes was assessed by four studies (Davis 1998, Dewey 1998, Dwyer 1999, and Majnemer 2006. [1][2][3][4] Davis 1998 compared the mean ages of eight motor milestones attainment of prone and supine sleepers (recorded by parents until 6 months of age). Dewey 1998 compared the social, communication, fine and gross motor, and total developmental scale scores (based on the Denver Developmental Screening Test; DDST) at 6 and 18 months in infants sleeping prone, side, and supine. Dwyer 1999 derived the adjusted odds ratios for several motor and cognitive milestones (assessed as binary outcome with yes/no question) in infants at 12 weeks comparing different infant sleep positions. Majnemer 2006 compared the mean scores on Alberta Infant Motor Scale (AIMS) and Peabody Developmental Motor Scale (PDMS) at recruitment (4 and 6 months) and on PDMS and Battelle Developmental Inventory at 15 months follow up in prone and supine sleepers. 4 Seven observational studies assessed plagiocephaly in infants with different tools: using Argenta's assessment tool at 2-3 months (Ballardini 2018; Mawji 2014), and visual and anthropometric examinations (Hutchison 2003(Hutchison , 2004(Hutchison , 2009 at 4-7 months; Leung 2017, van Vlimmeren 2007 at 2-3 months). We could not include data from Leung 2018 in meta-analysis because the study reported odds ratio of plagiocephaly for supine-sleep maximum (a continuous variable, defined as number of hours infant slept in supine position). 5

Appendix S3: Risk of bias in included studies
Of 48 observational studies/ non-randomized trials, 46 were considered to be at serious risk of bias ( Figures S1 and S2), mostly due to confounding and misclassification bias: 28 studies were at serious risk of bias for confounding because they either did not address the potential confounders (like maternal education, socio-economic status) or did not adjust for these confounders. Selection bias was serious in 15 studies either because the controls were not matched to cases for geographical area or age (10 studies) or the selection of participants could have related to outcome after exposure status (4 studies). 12 casecontrol studies were at serious risk of bias for misclassification of interventions due to recall bias expected in case interviews conducted remote to the event. 18 studies were at serious risk of bias for deviation from intended interventions. These studies noted exposure status as usual sleeping position for "SIDS cases" and no information was available on adherence to the exposure status. All six randomized crossover trials were judged to be at some concern of bias owing to inadequate information on randomization process (Figures S3 and S4).
Appendix S4: Subgroup analyses for SIDS (0-1 year) We explored the possible causes for significant heterogeneity noted for the outcome of SIDS (I 2 =64%) by conducting subgroup analyses by study design, study period and location (Figures S8, S9 and S10). These causes could partly explain the heterogeneity.
Appendix S5: Effect of sleep position on physiological parameters The intervention trials (n=11) evaluated the effect of infant sleep positions on respiration and hemodynamics in the neonatal period. These studies were mostly crossover in design; hence the same infant underwent supine and prone positioning in the sleep state, the latter being recorded either by daytime polysomnography or clinical observation.
Physiological parameters were reported by both randomized (n=6) and non-randomized (n=5) crossover studies. The participants in were term healthy neonates, except for Ma 2015, which included 9 preterm infants out of 35 participants. 6 One study (Lucchini 2015) provided p-values only and could not be included in the meta-analysis. Most of the studies compared supine with prone sleep position, except Poets 2009 which compared supine vs. side sleep position. All participants were studied in a wellcontrolled environment in medical universities ensuring their safety and accurately recording physiological parameters (by avoiding artefacts due to motion, arousals, body movements etc). Some trials also evaluated the effect of sleep state and postnatal age on these parameters by recording the outcomes in active sleep (AS) and quiet sleep (QS) in both the sleeping positions at various ages (2-4 weeks, 2-3 months and 5-6 months). The outcomes studied in these trials included stroke volume and cardiac output by electrical velocimetry, skin blood flow using Laser Doppler, systemic vascular resistance index, rate of desaturation events (<80%/h), minute ventilation and end-tidal carbon dioxide levels, arousal from sleep, heart rate (HR) responses following provoked arousal, HR responses to nonarousing trigeminal stimulation and baroreflex sensitivity.
Tables S1 and S2 list some of the important physiological parameters for the comparison of supine with prone sleep position in newborns. Most newborns were evaluated in quiet and active sleep states during 2-4 weeks of life, except three studies (Ma 2015, Rossor 2018, and Wu 2017) that had evaluated newborns in the first week of life. The studies reported mean values of the parameters and their standard deviations and had indicated if the differences were significant (p-value <0.05 for paired t-tests and repeated measures ANOVA). Since the study design was crossover, it was not possible to estimate the correlation factors and hence the effect sizes of the individual studies. Therefore, we could not pool these results.
Seven studies provided baseline heart rate in the neonates in supine and prone positions. Four studies reported significantly higher heart rate in prone sleep position, compared to supine position. There was no significant difference in Ma 2015 and Rossor 2018, while heart rate was higher in supine position in one study (Fister 2020; p-value >0.05). The increased heart rate in prone position may indicate a response to decreased stroke volume to maintain the cardiac output.  7 A survey by Mitchell 2007 found a substantial increase in the proportion of infants in New Zealand sleeping supine from 24% in 1992 to 72% in 2005, which could have accounted for 39%-48% decrease in SIDS mortality during this period. 8 Mitchell 2012 evaluated the impact of SIDS prevention campaign and calculated that over 3000 lives were saved over years (1990 to 2008) in New Zealand, and similarly, more than 17,000 and 40,000 lives were saved in England & Wales and the United States, respectively. 9 Muller-Nordhorn 2021 examined the relationship of immunization coverage and SUDI rates (1996-2015) using Poisson regression, adjusting for sleep position and poverty. 10 The authors reported crude and adjusted risk ratios of 0.90 (0.89-0.92) and 0.95 (0.90-1.00), respectively, for supine sleep position on the risk of SUDI, using multiple imputation due to missing data on sleep position.   Explanations a. Most of the pooled effect provided by studies "C". b. Statistical heterogeneity (I2≥60% or Chi2≥0.05). c. Evident asymmetry in funnel plot d. The included study used unadjusted OR and was considered as very serious risk of bias e. Less than 300 newborns in continuous outcomes or less than 400 newborns in dichotomous outcomes. f. Less than 30 events g. The pooled effect provided by study "B". h. Wide confidence interval crossing the line of no effect. Figure S1. Risk of bias "traffic light" plots: review authors' judgements about each risk of bias item for each included study (observational studies) Figure S2. Risk of bias "weighted bar plots": review authors' judgements about each risk of bias item presented as percentages across all included studies (observational studies) Figure S3. Risk of bias "traffic light" plots: review authors' judgements about each risk of bias item for each included study (randomized crossover trials) Figure S4. Risk of bias "weighted bar plots": review authors' judgements about each risk of bias item presented as percentages across all included studies (randomized crossover trials)