Our systematic search identified studies published from 2007 to March 2022 covering the same electronic databases as the 2007 review: Pubmed, PsycINFO, Embase, the Cochrane library, and Blackwell Synergy. The search terms used are shown in Table 1. Following standard guidelines for systematic reviews (9) leading journals were hand searched as were the reference lists of those studies meeting the inclusion criteria. Due to lack of resources to undertake translation searches were restricted to English language material. We included all papers that identified any effects of swaddling on infant sleep and arousal even if this topic was not the primary purpose of the paper.

After screening, the first author worked independently to extract data from the included papers. All papers that met the inclusion criteria were read in full, with data extracted by hand using a customized template (Supplementary Appendix 1) (10). Mindful that “the effect of swaddling may be somehow mediated through the modified sleep characteristics in the infants” (11, 12) the data extraction template included factors associated with modified sleep behaviour, such as environmental temperature (9, 13, 14), nocturnal or diurnal setting (15–17) and feeding method (18). Study authors were contacted in cases where information was missing. An appraisal form was used (Supplementary Appendix 2) to assess the quality of each study in terms of internal and external validity.

An initial list of 171 studies was retrieved from the initial searches. Studies were assessed against the inclusion criteria with 115 discarded following title review. A further 43 were discarded following a review of abstracts, and 7 were discarded following review of full text. At each stage reviews were conducted by the first author and verified by the second author with discussion of any discrepancies. Where two publications reported on the same study (19, 20) (confirmed by contacting authors) the information extracted was combined as a single study (20), meaning 6 studies met the inclusion criteria (see Figure 1).

The six studies that met the inclusion criteria all used swaddling as their primary intervention and measured infant sleep and arousal outcomes, but varied in their approach to study design and setting. Richardson et al. (20), Meyer and Erler (21), and Narangerel et al. (22) each sought to evaluate the effects of swaddling on different stages of infant arousability, although Narangerel et al. (22) narrowed their focus to respiratory rate and oxygen saturation. The remaining studies, sought to ascertain the effects of swaddling on the temporal nature of infant sleep and arousal. Three studies (20–22) used clinical measures such as polysomnography to assess sleep and arousal while the other three studies (11, 23, 24) were conducted in a home setting and used participant self-report instruments (see Table 2).

The main findings identified from two clinical studies (20, 21) and one self-report study (24) were that swaddling of infants naïve to the intervention was associated with decreased arousal and increased sleep duration. These findings were consistent across nocturnal and diurnal sleep periods. However, two self-report studies did not report any difference in duration of crying or sleeping as a result of swaddling (23) or of any effect of swaddling upon “infant sleep disturbances” (11). The remaining clinical study (22) did not report any measurable effect of swaddling on SaO₂ or respiratory rates. The authors of two studies (22, 23) cited Type II Errors as explanations for not reporting significant outcomes.

Evidence hierarchy is often used to assess the quality of intervention studies (25). The first author independently applied the hierarchy to determine the appropriateness of study designs for assessing questions of “effectiveness”. In applying this approach to the question of swaddling effectiveness, two studies (11, 23) would be designated “poor quality” in terms of measures of effectiveness, being qualitative studies considered more susceptible to bias than the quasi-experimental studies (20, 22). The randomized trial (21), designated at the top of the hierarchy, would be deemed most robust. However, the objectives of each study are too dissimilar to synthesize in this manner. Narangerel et al. (22) is a particular outlier, focusing specifically on respiratory rate and oxygen saturation, while the other studies examined sleep and arousal more broadly, including maternal perception of swaddling effectiveness.

Varying degrees of confounding were evident across all studies. Only Richardson et al. (20) and Narangerel et al. (22) reduced contamination by accounting for environmental temperature, for instance, which is known to significantly modify an infant's cardiorespiratory and autonomic parameters (26). Moreover, only two studies (20, 24) ascertained infant feeding method. Whether an infant is breast or bottle fed is considered a strong moderating variable with regard to sleep fragmentation (27–29). Potential differences in such important baseline characteristics reduce internal validity (30).

The weakest study in relation to confounding was that of Narangerel et al. (22) which not only failed to allocate participants randomly but also failed to ascertain infant feeding method and whether the sleep studied was nocturnal, diurnal or both. An investigation of sleep and arousal cannot be effective when analyzed in isolation from these key factors (31, 32).

For most of the included studies it was unclear whether true randomization had been conducted reflecting previous critiques that randomization is inadequately described in most published studies (33–36). Additionally two studies (23, 24) used a recruitment technique that biased the pool of participants towards certain demographics, namely educated, western mothers giving birth in a hospital setting. One study (23) required participants to have “the resources to play a VHS tape/CD at home” and to “contact a research assistant by phone” before being discharged from hospital. Further, mothers who did not speak English or were unable to fill out a diary themselves were excluded. The generalizability of McRury and Zolotor (23) was further weakened by the relatively high attrition rate of 31%. The tendency to drop-out has been associated with ethnicity and low academic achievement (37–39).

With regard to allocation bias, those studies which utilized randomization (20, 21, 23, 24) also used comparison groups that were balanced at baseline and thus distributed possible confounding factors evenly. McRury and Zolotor (23) was the strongest study in this respect, being the only one to blind the researchers to control and intervention groups, and to compare demographic and baseline characteristics of the intervention and control participants.

Two studies (11, 22) allocated infants to predefined “natural” intervention and control groups. The outcomes of these studies cannot be confidently attributed to the intervention as their non-randomized nature may have encouraged a tendency to over-estimate effect sizes (40).

However, the greatest threat to the robustness of the synthesis involves the integrity of the intervention. This was problematic across all studies, and it is unclear whether the interventions were similar enough to synthesize: Narangerel et al. (22) for instance, gave a very brief description of the swaddle technique used, namely “two or three layers of cloth that enclosed the trunk and arms up to infant's neck”. Where swaddle technique was described, three (20–22) enclosed the infants' arms within the swaddle. This detail is important as previous research suggests infants swaddled with arms free experience the same startle frequency as unswaddled infants (7). Arm restraint therefore appears to be integral to the effectiveness of the swaddling intervention. The weakest studies, and those most at risk of Type III error, were those lacking any description of swaddle technique (11, 23). While Richardson et al. (20) and Meyer and Erler (21) describe their swaddle techniques in some detail, they along with all other studies, failed to measure or control for swaddle tightness–another important moderating variable discussed below. Richardson et al. (20) merely stated that the infants were “firmly” wrapped.

Significant outcomes for swaddling were only detected in non-routinely swaddled infants (20, 21). When samples involved only routinely swaddled infants significant changes in sleep and arousal were not detected (11). The application of routine vs. non-routine sleep interventions has important implications for infant physiology, and therefore health policy. In support of this finding non-routinely swaddled infants have been shown to exhibit higher arousal thresholds (41) and less spontaneous arousability in AS (20), both considered risk factors for SIDS (42). Future studies should further explore habituation to swaddling.

To date, a largely neglected area of infant sleep research is the interaction between interventions and sleep architecture (changes in sleep states). This review indicates that swaddling significantly reduces the frequency of infant sleep state changes, promoting quiet sleep and sleep efficiency (21).

Previous research has associated a deficiency in acethecholine brain receptors with insufficient arousal (43, 44). It is possible that swaddling, by decreasing the type and number of arousals, may exacerbate this. Furthermore, the arousal threshold from quiet sleep is greater than that from active sleep (21). By promoting greater duration of quiet sleep at a time when arousal mechanisms are underdeveloped, swaddling may impair an infant's ability to handle a regulatory crisis, such as prolonged apnea. Future research should therefore examine the effect of swaddling on the sequence and duration of infant sleep states.

Both studies conducted in sleep laboratories (20, 21) found swaddling to have measurable outcomes, and there is evidence from other sleep studies that the study environment is an important moderating variable (45). For instance, swaddled infants have been shown to demonstrate increased responsiveness to environmental auditory stress compared to the unswaddled condition (46). Another facet of environment, bedsharing, has also been shown to influence infant sleep and arousal (28, 29, 47). Bedsharing infants experience significantly less time in deep sleep and significantly more shifts in sleep stages than solitary sleeping infants (48). Despite these factors, none of the studies in the present review detailed whether bedsharing or co-sleeping (parent sharing room with infant) was a factor considered in their analysis.

Motor restraint is a key component of swaddling, fundamental to the intervention. It has been suggested that the sedative effects of swaddling are “triggered by very specific stimuli delivered above a certain threshold” (37, 49). This threshold may, in part, be determined by tightness. Indeed, the decreased arousability procured through swaddling is unlikely to be explained merely by the material's constant stimulation of the skin and thermal receptors, as one study in the review (20) controlled for these factors and still found decreased arousability in swaddled infants. Yet despite the potential importance of swaddle tightness, none of the studies attempted to measure or control it, weakening the value of the studies reviewed. Swaddle tightness is an elusive variable, not only across swaddling studies, but also within health policy. Australian guidelines (12) advise that swaddles “should be firm, to prevent loose wrapping becoming loose bedding. However, the wrap should not be too tight and must allow for hip and chest wall movement.” “New guidance from the American Academy of Pediatrics does not mention tightness at all (50). Earlier guidance provided slightly more detail: The swaddle should not be too tight or make it hard for your baby to breathe or move their hips” (51).

That swaddled infants demonstrate heightened responsiveness to environmental auditory stress (26, 46) supports the hypothesis that tightness may be an important factor in the effectiveness of swaddling. It is possible, for instance, that the tightness of the swaddle numbs tactile sensory receptors prompting a compensatory increase in auditory sensory reception.

There is also evidence to suggest that swaddle tightness may impact cardiorespiratory function. This has implications for SIDS, as abnormalities in cardiorespiratory function have been shown to impact the arousal response needed to regulate breathing (41, 52, 53). Unless very loose, swaddling increases thoracic compression (54) which leads to a reduction in oxygen saturation and a consequent increase in respiratory rate and cardiac activity (21). These physiological changes may explain why swaddling has been associated with an increased risk of acute respiratory infection (1, 55). The finding that swaddling has little or no effect on respiratory rate applies only to routinely swaddled infants (22) contrasting with earlier studies (56) and with the previous review (1). The implications of this are important as impaired respiratory and heart rates are associated with arousal difficulties, a key mechanism proposed in the triggering of SIDS (41). An investigation which explicitly examines swaddle tightness would therefore be of benefit, particularly with regard to the formation of policy guidelines for cardio-respiratory compromised infants.

Only six studies met the inclusion criteria, and no consistent outcomes were revealed. As the integrity of the intervention varied across all studies reviewed, including baseline characteristics, interpretation and transferability of the results should be treated with caution.