This meta-analysis followed the recommendations proposed by Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA).21

We did not restrict the search strategy to a specific language. To be included, a publication needed to be a randomized and parallel controlled trial assessing the effects of multisensory stimulation combined to standard care compared to standard care alone on length of hospital stay or weight gain of healthy stable hospitalized infants born at less than 37 weeks of gestational age.

The multisensory interventions, which combined two or more sensory stimuli, considered in this review were the following: the ATVV, the TKS, and the KMC.

The length of hospital stay was defined as the number of days from the randomization to hospital discharge or the NICU discharge. The weight gain was divided into three measures: a) final weight, defined as the weight at discharge from hospital or at the end of the study (grams), b) daily weight gain from randomization until hospital discharge or the end of the study (days), c) total weight gain, defined as the difference between the start and final weight (grams).

Exclusion criteria were: a) quasi-experimental studies in an attempt to obtain more valid conclusion for this review, b) studies that did not report outcomes for preterm infants separately from those for full-term infants, c) studies that assessed preterm infants at risk for developmental delay, including neurological disorders or environmental exposure risks, d) studies that applied interventions that did not focus on sensory stimulation, e) studies that applied the sensory stimulation as a secondary treatment, f) studies that applied multisensory intervention only once, g) studies that applied the reduction of external stimuli, h) studies that applied breastfeeding interventions, oral-motor stimulations, and massage therapy using only tactile stimulation because these interventions are based on the stimulation of only one sensory system, and i) studies that applied therapeutic touch involving non-touch or energy-balancing techniques.

The following electronic databases were searched: PubMed, Scopus, Web of Science, Embase, SciELO Citation Index (Web of Science), CINAHL, Cochrane Library, and LILACS (BIREME). The search occurred from the inception date of each database until May 6, 2022. The search strategy used combined medical subject heading (MeSH) terms and text words related to population (preterm and infant), intervention (multisensory, multimodal, auditory-tactile-visual-vestibular (ATVV), massage, tactile-kinesthesic, kangaroo care, skin to skin contact), location of intervention (hospital, NICU), and type of study (clinical trial, randomized).

An example of the combination of terms used with PubMed is: ((((((Preterm*) OR (Premature)) OR (Premature Birth[MeSH Terms])) AND ((((Infant*) OR (Infant[MeSH Terms])) OR (Newborn)) OR (Infant, Newborn[MeSH Terms]))) AND (((((((((ATVV) OR (Tactile-kinesthesic)) OR (“tactile kinesthesic”)) OR (massage)) OR (massage[MeSH Terms])) OR (kangaroo)) OR (Kangaroo-Mother Care method[MeSH Terms])) OR (skin-to-skin)) OR (“skin to skin”))) AND (((((hospital) OR (hospitalized)) OR (hospitals[MeSH Terms])) OR (“Neonatal Intensive Care Units”)) OR (Intensive Care Units, Neonatal[MeSH Terms]))) AND ((((Randomized) OR ("clinical trial")) OR (Randomized Controlled Trials as Topic[MeSH Terms])) OR (clinical trial[MeSH Terms]).

We also searched for additional articles in the list of the references from 15 review studies17,18,20,22-33 found in the electronic search and the list of references from the included studies. All duplicated papers were removed, and then two reviewers (GPR, BABR) independently selected the studies according to inclusion criteria. In the case of any disagreement, a third reviewer (ANS) arbitrated. We used the State of the Art through Systematic Review (START) software34 for study selection.

Two review authors (GPR, BABR) independently extracted and compiled the study data using a spreadsheet from the Excel program. A third reviewer (ANS) resolved disagreements. We extracted the following information from the included studies: study country, the characteristics of the participants (number of subjects, age, and sex), the outcome studied, the characteristics of treatment (technique/approach, frequency and dosage, delivery method, and duration), and results. The mean and standard deviation of data of interest were extracted, when available. When data were not available, we contacted the corresponding author to request information by e-mail.

We evaluated the risk of bias of included studies using the Cochrane Handbook for Systematic Reviews of Interventions.35 We verified the following features: randomization process, deviations from the intended interventions, missing outcome data, measurement of the outcome, and selection of the reported results. Each item and the overall bias were classified as low risk, high risk, or some concerns.7 Two review authors (GPR, ANS) independently rated the risk of bias. In the case of any disagreement, the authors discussed and reached a consensus.

We used the Grading of Recommendations Assessment, Development and Evaluation (GRADE) method to verify evidence synthesis, assessing: risk of bias, inconsistency of results, indirectness, imprecision, and publication bias.36 The level of evidence for risk of bias was downgraded by one level if 25% to 50% of the included studies were classified as high risk of bias or most studies were classified as some concerns, and downgraded by two levels if more than 50% of the included studies were classified as high risk of bias. The level of evidence for inconsistency was downgraded by one level if the I2 test was greater than 50%, or there was a high variation on the size of the effect between the included studies, or if there was high heterogeneity regarding the content of the intervention between the included studies; and downgraded by two levels if there was a combination of these factors. The level of evidence for indirectness was downgraded by one level if more than 50% of participants varied from the population of interest. The level of evidence for imprecision was downgraded by one level if the confidence intervals were too wide or the sample size was less than 400 participants.37 The level of evidence for publication bias was downgraded by one level if publication bias was identified by visual inspection of funnel plots for outcomes that included more than 10 studies. The level of evidence was classified as: high (enough evidence in the estimate of the effect), moderate (the true effect is close to the estimate of the effect), low (the confidence of the effect is limited), and very low evidence (little confidence of the effect estimate). A high or moderate level of evidence represented a strong recommendation for use of the intervention.38 A summary of findings table was produced using GRADEpro software (McMaster University, Hamilton, Canada).

Meta-analysis was conducted using RevMan 5.1 (Cochrane Collaboration, Oxford, England). Effect sizes are reported as the mean difference (MD) with 95% confidence interval (CI) for each study and outcome. Pooled treatment effects were calculated across trials by using a random-effects model.39 We verified heterogeneity among studies using the I2 statistic.39

The funding sources played no role in the design, conduct, or reporting of this study.

The search resulted in 63 randomized clinical trials that met the full inclusion criteria.8,40-101 Fifty-four were included in the meta-analysis (Supplementary material Figure 1).

Studies included medically stable preterm infants. The sample size varied from 20 to 488. Most of the included studies evaluated preterm infants with gestational age between 28 and 35 weeks. Studies included hospitalized infants from birth to two months old (Table 1).

Forty-three (68%) studies assessed length of hospital stay. Twenty-three (37%), 30 (48%), and 11 (17%) studies assessed, respectively, the final weight, the daily weight gain, and the total weight gain (Table 1).

Eleven (17%) studies applied multimodal stimulation.8,40-49 Nine studies applied the ATVV, which included sensory stimuli such as female voice or soft lullaby song, light stroking all over the body, eye-to-eye contact or black and white cards, and rocking. One study applied the auditory-vestibular stimulation, using rocking waterbeds, taped simulated heartbeat, and a woman's voice. Two studies applied the STMT intervention that includes the use of a water mattress (vestibular), massage (tactile), active changes of position (kinaesthetic), soft music and verbal contact (auditory), and eye-to-eye contact (visual). The control group received the standard care of each hospital. Three studies specified the application of kangaroo care as part of their standard care, and two studies the application of visual stimulation. All the infants in the multimodal stimulation group also received standard care (Table 1).

Twenty-eight (44%) studies applied TKS50-74,98,99,101 which is a form of infant massage that includes kinesthetic movement. The tactile portion consisted of the application of moderate strokes to the infant. The kinesthesic part consisted of extension and flexion movements of the upper and lower limbs. One study combined TKS with classical music, and one study applied oil during massage. The control group received the standard care of each hospital. Two studies specified the application of kangaroo care as part of their standard care, two studies the application of gentle touch, and one the application of classical music. All the infants in the TKS group also received standard care (Table 1).

Twenty-four (38%) studies applied the KMC or skin-to-skin contact.77-97,100 The main component of KMC is skin-to-skin contact. During the skin-to-skin contact, the parent holds the infant in an upright position, closer to their chest. The KMC also includes breastfeeding. One study associated KMC with massage and touch. One study combined KMC with music, and one study with a calming touch. Both kangaroo and control groups received the standard care of each hospital. The infants from the control groups were kept inside an incubator or a warmer room in 11 studies. The parents applied the traditional holding to the infants of the control group in five studies. In two studies, the parents from the control group were educated about the kangaroo method and had the option to apply or not (Table 1). All the infants in both groups also received standard care.

The risk of bias of the 63 eligible trials are described in supplementary material Figure 2. We found that 12 studies (19%) were classified as low risk, 30 (48%) as some concerns, and 21 (33%) as high risk.

The main methods limitations across studies were that 45 studies (71%) did not report the randomization process or did not use or inform about concealed allocation, and 22 (35%) did not inform about missing outcome. Because of the type of interventions, it was not possible to blind therapists and participants. Most studies did not blind assessors or provide information about blinding. Because the outcome length of hospital stay reflects decisions made by the intervention provider, we classified the studies as presenting some concerns regarding the influence of the non-blinding of the assessor. Because weight gain is an observer-reported outcome not involving judgment, we classified the studies as presenting low risk regarding the influence of the non-blinding of the assessor. Studies had different designs, convenience samples, and included small sample sizes (supplementary material Figure 2).

Fifty-three (84%) studies provided means and standard deviations to input in the meta-analysis. We estimated the standard deviation based on means and p values for the length of hospital stay in three studies,47,71,74 and for the daily weight gain in three studies.47,72,74

Auditory–tactile–visual-vestibular intervention

We included 10 studies8,40-43,45-49 that compared the effects of ATVV combined to standard care to only standard care. There is low-quality evidence (downgraded due to risk of bias and inconsistency) that adding ATVV to standard care may increase total weight gain (MD = 72.7 g, 95% CI: 68.3, 77.2; n = 114, 2 trials, I2 = 0%) at post intervention when compared to standard care alone (Fig. 1, Table 2). In addition, there is low-quality evidence (downgraded due to risk of bias and imprecision) that ATVV combined with standard care may not reduce length of hospital stay (MD = −0.6 days, 95% CI: −2.2, 1.1; n = 923, 9 trials, I2 = 55%) compared to standard care alone and there is very low-quality evidence (downgraded due to risk of bias, inconsistency, and imprecision) that ATVV combined with standard care may have little to no effect on daily weight gain (MD = 0.9 g, 95% CI: −1.1, 3.0; n = 227, 2 trials, I2 = 0%) (Fig. 1, Table 2) compared to standard care alone. Only one study assessed the final weight.

Fig. 1.

Forest plot for effect of the Auditory–Tactile–Visual-Vestibular Intervention (ATVV) combined with standard care compared with standard care alone on the length of hospital stay, the final weight, the daily weight gain, and the total weight gain.

(0.64MB).

Table 2.

Summary of the evidence: certainty assessment, summary of findings, and level of evidence for each outcome and intervention.

	Outcome	Certainty assessment						Number of patients		Anticipated absolute effects		Level of evidence
		Infants (studies)	Risk of bias	Inconsistency	Indirectness	Imprecision	Publication bias	Multisensory	Standard care	Risk with standard care	Risk difference with multisensorial
ATVV	Length of hospital stay (days)	923 (9 RCTs)	seriousa	seriousb	not serious	not serious	None	455	468	0	MD 0.6 lower (2.2 lower to 1.1 higher)	⨁⨁○○
												LOW
	Daily weight gain (grams)	227 (2 RCTs)	very seriousc	seriousb	not serious	seriousd	None	114	113	0	MD 0.9 higher (1.1 lower to 3.0 higher)	⨁○○○
												VERY LOW
	Total weigth gain (grams)	114 (2 RCTs)	seriousa	not serious	not serious	Serious	None	64	53	0	MD 72.7 higher (68.3 higher to 77.2 higher)	⨁⨁○○
												LOW
TKS	Length of hospital stay (days)	570 (13 RCTs)	very seriousc	seriousb	not serious	not serious	none	283	287	0	MD 1.7 lower (3.2 lower to 0.2 lower)	⨁○○○
												VERY LOW
	Final weight (grams)	633 (11 RCTs)	seriousa	not serious	not serious	seriousd	none	390	393	0	MD 69.31 higher (4.65 lower to 143.28 higher)	⨁⨁○○
												LOW
	Daily weight gain (grams)	682 (14 RCTs)	seriousa	seriousb	not serious	not serious	None	338	344	0	MD 3.3 higher (1.7 higher to 4.9 higher)	⨁⨁○○
												LOW
	Total Weigth gain (grams)	182 (5 RCTs)	very seriousc	not serious	not serious	seriousd	None	93	89	0	MD 85.6 higher (35.5 higher to 135.6 higher)	⨁○○○VERY LOW
KMC	Length of hospital stay (days)	1830 (17 RCTs)	seriousa	seriousb	not serious	not serious	None	929	901	0	MD 0.6 lower (2.1 lower to 0.8 higher)	⨁⨁○○LOW
	Final weight (grams)	798 (10 RCTs)	not serious	not serious	not serious	not serious	None	406	392	0	MD 8.5 lower (25.8 lower to 42.9 higher)	⨁⨁⨁⨁HIGH
	Daily weight gain (grams)	841 (10 RCTs)	seriousa	seriousb	not serious	not serious	none	428	413	0	MD 3.2 higher (2.2 higher to 4.3 higher)	⨁⨁○○LOW

ATVV, Auditory–Tactile–Visual-Vestibular Intervention; KMC, kangaroo mother care; MD, mean difference; RCTs, randomized controlled trials; TKS, tactile-kinesthesic stimulation.

a

25% to 50% of the included studies were classified as high risk of bias or most studies were classified as some concerns.

b

I2 result greater than 50% or high variation on the effects size or high heterogeneity regarding the content of the intervention.

c

More than 50% of the included studies were classified as high risk of bias.

d

Confidence intervals too wide or the sample size < 400.

Tactile–kinesthesic stimulation

We included 32 studies50-64,66-74 that compared the effects of TKS combined to standard care to only standard care. There is very low-quality evidence (downgraded due to risk of bias and inconsistency) that adding TKS to standard care may reduce days at the hospital (MD = −1.7 days, 95% CI: −3.2, −0.2; n = 570, 13 trials, I2 = 1%), low-quality evidence (downgraded due to risk of bias and inconsistency) that it may increase daily weight gain (MD = 3.3 g, 95% CI: 1.7, 4.9; n = 682, 14 trials, I2 = 83%), and very low-quality evidence (downgraded due to risk of bias and imprecision) that it may increase total weight gain (MD = 85.6 g, 95% CI: 35.8, 135.6; n = 182, 6 trials, I2 = 42%) at post intervention when compared to standard care alone (Fig. 2, Table 2). In addition, there is a low-quality evidence (downgraded due to risk of bias and imprecision) that the TKS combined with standard care may have no effect on final weight when compared to standard care alone (MD = 57.3 g, 95% CI: −6.9, 121.5; n = 783, 12 trials, I2 = 82%) (Fig. 2, Table 2).

Fig. 2.

Forest plot for effect of the tactile-kinesthesic-stimulation (TKS) combined with standard care compared with standard care alone on the length of hospital stay, the final weight, the daily weight gain, and the total weight gain.

(1.11MB).

Kangaroo method care or skin-to-skin contact

We included 20 studies75-93,95 that compared the effects of KMC combined to standard care to standard care alone. There is low-quality evidence (downgraded due to risk of bias and inconsistency) that adding KMC to standard care may increase daily weight gain (MD = 3.2 g, 95% CI: 2.2, 4.3; n = 841, 10 trials, I2 = 64%), at post intervention when compared to standard care alone (Fig. 3, Table 2). We found there was low-quality evidence (downgraded due to risk of bias and inconsistency) that KMC combined with standard care may have no effect on length of hospital stay (MD = −0.6 days, 95% CI: −2.7, 0.8; n = 1830, 18 trials, I2 = 88%) compared to standard care alone and there is high-quality evidence that KMC combined with standard care does not increase final weight (MD = 8.5 g, 95% CI: −25.8, 42.9; n = 798, 10 trials, I2 = 28%) compared to standard care alone (Fig. 3, Table 2). Only one study assessed the total weight gain.

Fig. 3.

Forest plot for effect of the kangaroo mother care (KMC) combined with standard care compared with standard care alone on the length of hospital stay, the final weight, the daily weight gain, and the total weight gain.

(1.05MB).

The summary of findings and the overall quality of evidence for the primary outcomes are described in Table 2.