Study design and participants
This was an individually randomized unmasked controlled trial (RCT) conducted between July 2015 to November 2016. The first 552 infants enrolled in the primary trial [21] were included for evaluation of child developmental outcomes. This cohort of LBW infants was a very selected cohort consisting majorly of stable late preterm or term small for gestational age infants without any problems at birth. Although the inclusion weight was 1500–2250 g, yet for those weighing between 1500 and 1800 g referral was facilitated for hospital care following Government of India guidelines. The infants weighing between 1500 and 1800 g were considered for inclusion only if the families refused to take the baby to the hospital, or if the baby was taken to hospital but was either not admitted, or admitted and discharged before s/he became 72 h old and not started on KMC. Infants unable to feed, with difficulty in breathing, with less than normal movements or gross congenital malformations, those for whom KMC was initiated in hospitals, and those whose mothers did not intend to stay in the study area for the next 6 months or did not consent to participate were excluded. The weight category was determined based on our formative research findings [22] that suggested most babies with birth weight > 2250 g wriggle out of KMC position before the neonatal period. The lower cut-off 1500 g was considered to avoid including infants who would have been at a high-risk of complications and would have required hospital care. The study was conducted in rural and semi-urban populations of around 2 million in Haryana, North India. In the study sites, around 40% were home births and around one-fourth of all babies were born with LBW. Details of the study settings have been published elsewhere [23].
Enrollment, randomization and allocation
Ethical clearances were obtained from the Institutional Ethics Review Committee and the WHO Ethics Review Committee. State approvals were also available. Pregnant women were identified by a door-to-door pregnancy surveillance team every three months. Identified pregnant women were followed-up regularly till delivery, with the frequency of contacts being higher in the third trimester. Newborns were visited at home and weighed as early as possible. A digital hanging weighing scale (AWS-SR-20; American Weigh Scales, Cumming, GA, USA) was used for weight measurement. Gestational age was documented from ultrasound report, hospital records or maternal recall, whichever was available, in the given order of preference. After screening as per inclusion and exclusion criteria, in the eligible mother-infant dyads, a study worker obtained written informed consent in the local language from caregivers prior to enrollment.
The unit of randomization was the mother-infant dyad. The randomization list was prepared by an independent statistician using random permuted blocks of variable size. Allocation of study identification number was done by an off-site randomization coordinator using serially numbered opaque sealed envelopes (SNOSE), kept under lock and key. Similar procedures were followed for participants enrolled in intervention or in the control group. If a dyad was allocated to the intervention (ciKMC) group, the randomization coordinator subsequently informed the intervention delivery team. We attempted to ensure that the study team responsible for outcome assessment and study investigators were not aware of the group allocation till the end of the study.
Intervention delivery
The ciKMC intervention comprised of promotion and support of skin-to-skin contact and exclusive breastfeeding by the intervention workers and supervisors. Mother-infant dyads allocated to the ciKMC group were visited at home by a trained intervention delivery team consisting of a pair of workers as soon as possible after enrollment to explain and initiate KMC and support its practice. The team home-visited daily for the first 3 days, then on days 5 and 7, twice in the second week and once each in the 3rd and 4th week to provide support and solve any problems related to practicing KMC. During home visits, the team observed the mother practicing KMC, enquired about skin-to-skin contact and breastfeeding in the preceding 24-h period, and supported the mother and family to solve any problems or overcome barriers to effective KMC. They counselled that skin-to-skin-contact be done for as long as possible during day and night, preferably for 24 h a day, with the assistance of other family members. Visits continued till 28 days of age or until the baby wriggled out and no longer accepted SSC, whichever was earlier. The intervention delivery was designed based on previous formative research [22]. All infants in the intervention and control groups received Home Based Post Natal Care (HBPNC) visits by government health workers (Accredited Social Health Activists; ASHAs) as implemented through the health system [24]. Specific details of the intervention have been published elsewhere [23].
Outcomes
The outcomes were cognitive, language, motor, socio-emotional development and infant temperament scores at 6 and 12 months of infant age.; maternal depressive symptoms at 6 weeks and 6 months of infant age; maternal sense of competence at 6 weeks and 12 months of infant-age; mother-infant attachment at 6 weeks of infant-age and assessment of the home environment at 12 months of infant-age.
Outcome ascertainment
Information on skin-to-skin-contact (number of days and average hours per day), as reported by the mother, in both the groups was ascertained by a trained outcome ascertainment team at the end of neonatal period. Developmental outcomes were ascertained in the study clinic by trained psychologists, who were unaware of allocation.
Bayley Scales of Infant and Toddler Development, 3rd Edition (BSID-III) was used to ascertain child development (cognition, language, motor and socio-emotional performance) at 6 and 12 months of age corrected for gestational age [25]. We adapted BSID-III for use in the study setting. For the adaptation, the test items were reviewed by the team of psychologists and public health experts in terms of cultural relevance, and subsequently, modifications were identified, discussed and incorporated. While conducting the adaptations, care was taken to match the style of the original item. For items that required translation in the local language i.e. Hindi, the translation was done by psychologists fluent in the local language and with a thorough understanding of the cultural context. An individual fluent in English language, and not a part of the study team, performed the back-translation. Prior to the start of the formal testing, the adapted materials were piloted on approximately 15–20 infants who were not a part of the trial.
The infant temperament scale was used as adapted in the MAL-ED study [26]. This 47 item-scale covered six domains i.e. activity, positive emotionality, negative emotionality, sociability, attention and soothability, where higher scores reflect more difficult temperament. Maternal depressive symptoms were assessed using the Patient Health Questionnaire (PHQ)-9; higher scores reflecting more depressive symptoms. The PHQ-9 is the depression module of the self-administered version of the PRIME-MD diagnostic instrument. The 9 items of PHQ-9 tool are based on the DSM-IV diagnostic criteria [27].
Maternal sense of competence was assessed using “maternal self-efficacy scale” that consists of 10 questions with four point scale responses; higher scores reflecting better maternal self-efficacy [28]. The maternal postnatal attachment scale was used to assess mother-infant attachment. This scale consists of 19 items with higher scores reflecting better attachment [29]. Home environment was assessed using “Pediatric Review of Children’s Environmental Support and Stimulation (PROCESS)” questionnaire. It consisted of three components: clinical observation, parent questionnaire and toy list. Higher scores reflect better stimulation and support to infants [30]. The study questionnaires were adapted according to local cultural context, translated in local language (Hindi), pre-tested and validated for use.
Sample size
To examine a difference of 0.25 SD between the intervention and the control group for cognitive, language, motor and socio-emotional outcomes at 80% power, 260 infants in each group i.e. 520 infants were required. 552 infants were enrolled, assuming a 10% loss to follow up.
Statistical analysis
All analysis was done using STATA version 14.0 (Stata Corp, Texas, USA). Intention to treat analysis was performed. The distribution of continuous data was examined using histograms and skewness and kurtosis coefficients calculated. Mean (SD; standard deviation) or median (IQR; inter-quartile range) were calculated for continuous variables and proportions for categorical variables. Distribution of baseline data on household, maternal and paternal, birth-related and infant characteristics were compared across the intervention and control groups. Chi-square test was used to compare proportions; independent t-test to compare mean and Mann-Whitney U test to compare median. The prematurity adjusted composite BSID-III scores for cognitive, motor, language and social-emotional domains were calculated using the raw scores and scaled scores [25].
To examine the effect of ciKMC on the outcomes considered, univariable linear regression analysis was done as the initial step. This was followed by multivariable linear regression wherein potential confounding variables were included in the model. The choice of variables to be adjusted for was based on biological plausibility and/or on the statistical significance (p < 0.20) of their association with the outcome(s) of interest in the univariate analysis. Potential interaction between ciKMC and other variables, especially sex of the infant and wealth quintile, was examined by including interaction terms in the multivariable regression models.
Additionally, as an a priori decision, the continuous BSID-III and Infant Temperament Scale scores were categorized into quartiles to examine the effect of ciKMC using ordinal logistic regression. Odds ratios with 95% confidence intervals were used to report associations.
As an exploratory analysis, we also conducted a post-hoc equivalence testing of means of cognitive, language and motor scores across the two study groups. The purpose of this equivalence testing was to provide evidence to support the contention that ciKMC does not do harm in terms of neurodevelopment as there could be an argument that ciKMC improves survival at the cost of poorer neurodevelopment. We used Statgraphics Centurion Version 18.0 (http://www.statgraphics.com/centurion-xviii) statistical analysis software to perform the post-hoc equivalence testing using two one-sided tests of equivalence (TOST). The lower (ΔL) and upper (ΔU) equivalence limits were set as − 3.0 and + 3.0 points respectively (1 SD =15 points for BSID; 3.0 points equate to 0.20 SD). Equivalence limits were set based on the discussion among the study investigators and with clinical psychologists. A difference of more than 3 points in the composite scores was considered to be clinically relevant. We defined “equivalence” as: ΔL ≤ μ1-μ2 ≤ ΔU where μ1 - μ2 represent the difference in mean scores among the two study groups. Null hypothesis considered was H0: μ1 - μ2 < ΔL or μ1 - μ2 > ΔU.
At a p-value of < 0.05, null hypothesis is rejected and “equivalence” is considered to be present.