Kangaroo Mother Versus Traditional Care for Newborn Infants ≤2000 Grams: A Randomized, Controlled Trial

Study Population (Inclusion and Exclusion Criteria)

The study population comprised live newborn infants with birth weights ≤2000 g, born at Clı́nica San Pedro Claver (CSPC), which is one of the largest tertiary care obstetrics facility in Colombia, delivering 13 500 to 15 000 live infants per year. Almost a fourth of them have birth weights ≤2500 g and 8% of the total are <2000 g. An infant could be included in the study if he/she had a mother or a relative (living in the same household) able to understand and willing to follow the general program instructions (care for a premature baby and compliance with a 1-year follow-up schedule) according to a semistructured interview, and also eligible for KMC (see eligibility criteria for KMC in Study Interventions). At this time, evaluation of mothers' willingness to perform the specifics of the kangaroo intervention was not made. Exclusion criteria included being referred to another institution, plans to leave Bogotá in the near future, life-threatening or major malformations, early-detected major conditions arising from perinatal problems (severe hypoxic–ischemic encephalopathy, pulmonary hypertension, etc), and parental or family refusal to comply with the follow-up program, or, for those assigned to the kangaroo group, refusal to comply with the specifics of the intervention. This limitation in the design had the potential for introducing an imbalance between study groups, but none of the mothers assigned to kangaroo refused to participate.

Sample Size

According to our previous study,²⁴ death rate among eligible patients under traditional care at the CSPC was ∼7%. To detect a twofold increase in the risk of dying, a sample of 215 subjects per group would be needed (α 0.05, two-tailed test, power 80%). Assuming a death rate among controls as high as 10%, at least 328 subjects per group were needed to detect a twofold increment in the risk of dying. We expected that only 70% of the newborn infants <2000 g would be eligible. Therefore, we planned to evaluate at least 940 infants ≤2000 g to identify a sufficient number of eligible patients.

Variables

Primary outcomes were mortality and growth (weight, height, and head circumference). Mortality was measured as the proportion of randomized children dying by 40 to 41 WCA among children with known life status at 40 to 41 WCA. Secondary outcomes were 1) length of hospital stay in survivors, measured as the sum of all days spent in hospital from eligibility to 40 to 41 WCA, in living children with complete follow-up; 2) overall incidence of infection measured as all infectious episodes requiring antibiotic treatment, regardless of whether treated in the inpatient or outpatient setting; 3) incidence of severe infection measured as nosocomial infections requiring systemic antibiotics or infections detected after discharge and requiring hospital admission for treatment; and 4) proportion of infants totally or partially breastfed, at 40 to 41 WCA. Partial breastfeeding was defined as the complement of breast milk with formula ordered by study personnel and covering up to 30% of the estimated nutritional requirements of the infant, as well as when the mother reported administration of nonprescribed formula in addition to breast milk. WCA 40 to 41 was chosen because it represents the time when LBW infants reach the conceptional age of term delivery. The study interventions compared here are expected to support LBW infants while they mature enough to face extrauterine life in similar conditions and with similar care requirements as those born at term.

Recruitment and Allocation

An observer especially assigned by the study reviewed the records of the delivery room and admission book daily and identified all newborn infants with birth weight or weight at admission ≤2000 g born at or referred to CSPC. Infants admitted to the neonatal care unit (NCU) as well as infants not admitted to the NCU (ie, those placed in a crib in the mother's hospital room) were followed and assigned a consecutive identification number.

Infants admitted to the NCU were followed by an attending physician, who assessed eligibility for the KMC intervention. Those children not admitted to the NCU were assessed for eligibility before mother and baby were discharge.

Once an infant was identified as eligible, he or she was randomly assigned to KMC or to the control intervention. Stratified block randomization lists were prepared in advance using a random numbers list (permutations of the number 16). There were four randomization lists according to birth weight: ≤1200 g, 1201 to 1500 g, 1501 to 1800 g, and 1801 to 2000 g. Blocks of four infants (two assigned to KMC and two to control) were prepared using a random number table.

Once the attending physician decided that a child was eligible, the person managing the allocation list (N.C.) was informed of the identification number of the eligible infant. The person managing randomization was aware of only the identification number, weight at birth (to consult the appropriate list according to weight strata), and whether the infant was a twin or triplet. The unit of randomization was the product of each delivery rather than individual infants. Therefore, twins or triplets were allocated to the same treatment to which the first of them who was eligible had been allocated. Each product of a delivery was allocated to the treatment corresponding to that registered in his/her weight stratum list according to the order in which each one had been declared eligible. In cases of twins or triplets, to consult the appropriate allocation list, the information on weight of the first eligible member of the set was used.

For infants assigned to KMC, parents were interviewed by a study nurse and invited to participate. If they accepted, the mother and infant were transferred to the Clı́nica del Niño, another Social Security Hospital where the ambulatory adaptation to KMC took place. Once in the Kangaroo Mother Program at the Clı́nica del Niño, parents were trained in KMC, and the adaptation of the mother and infant pair to kangaroo positioning and feeding was monitored.The mother and infant pair could stay at the program as long as necessary to demonstrate that the adaptation was appropriate.

Informed consent was not asked of parents of infants assigned to the control intervention. This procedure, proposed by Meinert and Tonascia²⁸ (randomization before seeking consent to participate), was chosen for the following two reasons. First, early discharge is very appealing to parents, and it is very likely that a sizable proportion of families would have asked to be assigned to KMC. This is not a theoretic possibility; such a condition could introduce a significant contamination bias in the study. In fact, a kangaroo vs traditional trial in Guatemala had to be terminated because of contamination of the control intervention. Second, according to the Ethics Committee (Javeriana University, Bogotá), there were no ethical problems because those assigned to the control intervention received the usual care currently provided at the participant institution. At that time, the intervention was regarded as the best available care for LBW infants.

Study Interventions

The interventions used were a modified version of the Rey-Martı́nez KMC. Infants <2000 g at birth are discharged as soon as they overcome major adaptation problems to extrauterine life, receive proper treatment for infection or concomitant condition, suck and swallow properly, and achieve a positive weight gain regardless of actual weight or gestational age. Mothers are used as “incubators” and as the main source of food and stimulation; babies are kept 24 hours a day in a strict upright position, in skin-to-skin contact, firmly attached to the mother's chest. In that way, the infant's temperature remains within the normal range by the mother's body heat. Babies must be breastfed regularly, and premature' formula supplements are administered to guarantee adequate weight gain if necessary. Infants are examined daily until weight gain reaches at least 20 g per day. Infants remain in the kangaroo position until they no longer accept it by demonstrating discomfort, pushing out limbs, and crying and fussing when mothers try to return them to the upright position.

Control (traditional care) intervention included keeping the infant in an incubator at the San Pedro Claver Hospital until he or she is able to regulate temperature and is thriving (ie, appropriate weight gain). Infants are discharged according to current hospital practice at CSPC (usually not before his or her weight is ∼1700 g). At the time of the study, the current practice at the San Pedro Claver NCU was to restrict parents' access to their babies severely.

Data Collection and Follow-up

After discharge, both groups had access to the same ambulatory care and follow-up. To minimize contamination bias, kangaroo infants' visits took place in the morning and control infants' visits in the afternoon. Ambulatory care included administration of metoclopramide²⁹ and supplemental vitamins (A, D, E, C) to all infants and prophylactic aminophylline to infants <34 WCA at entry to the study.

Mothers' interviews and clinical chart reviews provided baseline data. Baseline information and follow-up data were collected prospectively at the neonatal unit and at scheduled visits. Data were entered into a computerized data base, after which both manual and computerized data cleaning activities were performed. Given that baseline data concerned variables and events occurring before randomization, observers collecting baseline information were unaware of the study group to which the infant belonged. Clinicians assessing infants at follow-up could not be blinded, given that kangaroo mothers were holding their infants in the kangaroo position at time of the interviews and clinical exam.

Statistical Analysis

Comparisons between study groups for discrete variables were performed using χ² or Fisher's exact tests. Continuous variables were compared by t test or nonparametric tests when appropriate. Adjustments for potential confounders and exploration for interactions were accomplished using ANOVA and/or multiple linear regression for continuous variables; robust regression (RR) ANOVA/ANCOVA and multiple RR²⁶ for continuous variables with nonnormal distributions; and logistic regression for dichotomous outcome variables. The analysis was performed from the intention-to-treat perspective (effectiveness as opposed to efficacy), which means that all subjects were analyzed according to the group to which they were allocated, regardless of compliance with treatment or contamination of the intervention.