This cohort study consisted of a prospective cohort of infants with propranolol treatment for IH, whose sleep behavior was assessed over time, and a comparison with an identically designed, historical cohort of untreated infants, whose sleep behavior was assessed in the same manner without any temporal matching. The propranolol cohort consisted of otherwise healthy infants treated with propranolol for a complicated IH. The indication for treatment was made independently from and prior to possible inclusion in this project. Infants were recruited prospectively from the vascular anomalies clinics of the University Children’s Hospital Zurich and the Department of Pediatric Surgery, Charité University Medicine, Berlin, from September 2018 through August 2020. The control group was composed of an equal number of untreated matched healthy infants, investigated in an earlier project on infant sleep using the same assessments and study design [10, 24]. Matching criteria included subjects’ sex and chronological age at both assessment timepoints at ages 3 and 6 months (maximum difference of equal or less than ± 10 days at each timepoint). Inclusion criteria were healthy infants aged 0–5.5 months at baseline, born at term (37–42 weeks of gestation), and being mainly breastfed at time of inclusion (at least 50% of daily nutrition intake) to match with the control group . Only vaginal birth was allowed in the control cohort, whereas no restrictions regarding birth mode were applied in the propranolol group. Exclusion criteria included CNS disorders, acute illness, evidence of brain damage, chronic pediatric disease, and a family background of narcolepsy or significant psychiatric disease. Low birth weight (< 2500 g), treatment with medications affecting the sleep-wake cycle (apart from propranolol), and travelling across more than one time zone less than 1 week prior to the measurements also led to exclusion.
Propranolol treatment was performed according to current guidelines [1, 26]. In case of intolerable side effects, patients were offered off-label treatment with atenolol, a hydrophilic cardio-selective beta-blocker (1 mg/kg/day, single daily dose) .
Ethical approval was obtained according to local standards (Cantonal ethics committee, BASEC 2018-01366, and 2016-00730), and study procedures were consistent with the declaration of Helsinki. Written parental consent was obtained in all subjects before data collection.
Assessments were scheduled at ages 3 and 6 months, within a 1-month window centered around the target age. For subjects initiating propranolol treatment after age 3 months, only one assessment at 6 months was performed. In those infants who stopped propranolol and were switched to atenolol, additional measurements on atenolol were performed within the first 2 weeks after atenolol initiation.
Demographic data and patient characteristics were assessed using patient charts and questionnaire data.
Sleep-wake behavior was quantified for 7–10 days at each timepoint, simultaneously acquiring actigraphy and a 24-h diary in the infant’s natural environment (usually at home). GENEActiv movement sensors (Activinsights Ltd, Kimbolton, UK, 43x40x13mm, MEMS sensor, 16 g, 30 Hz Frequency recording resolution) were attached to the infant’s left ankle in a modified sock (Fig. 1).
The 24-h diary was completed in 15-min resolution by caregivers in parallel to the actigraphy recording. Diary data included reporting on infant sleep, external movements occurring during sleep (e.g., sleeping in a stroller/car), feeding episodes, propranolol administration, and bedtimes (clock time) (Fig. 1).
Caregivers in both cohorts completed the Brief Infant Sleep Questionnaire (BISQ) during the measurements, which is a validated survey to assess infant sleep . To investigate the impact of the infants’ sleep behavior on the whole family, both parents’ sleep quality was assessed using the Pittsburgh Sleep Quality Index (PSQI) .
The behavioral development of infants in both cohorts was investigated with the parent-completed age-appropriate Ages-and-Stages questionnaire .
The Hemangioma Activity Score (HAS) was used to document IH treatment response in the propranolol cohort .
Actigraphy data was processed according to in-laboratory standards [10, 24]. A published algorithm was applied to identify infant sleep and wake periods . By applying a previously validated 6-step modification, a better fit with the 24-h diary was achieved .
The primary outcomes were the objectively assessed Number of Awakenings per hour of Night Sleep and Sleep Efficiency (%), which is defined as the ratio of Total Sleep Time (at night) and Sleep Opportunity (time spent in bed), at 3 and 6 months . Sleep Efficiency may be viewed as a marker for overall nighttime sleep quality, as decreased values can correspond to problems falling asleep, increased sleep fragmentation, early waking up, or a combination thereof. These two variables represent a sleep composite that we have previously reported as Sleep Activity .
Secondary objective sleep measures included 24-hour Total Sleep and addressed the four remaining infant sleep composites, which we have previously identified as Sleep Night, Sleep Day, Sleep Timing, and Sleep Variability . Accordingly, we selected the variables Sleep Period (time between Sleep Onset and Sleep Offset at night in minutes, which represents Sleep Night), Longest Wake (longest continuous wake period during the day in minutes, which indirectly correlates with the accumulation of daytime sleep need and thus Sleep Day), Sleep Offset (morning wake time in hours, which represents Sleep Timing), and Variability of Sleep Period (standard deviation of Sleep Period across measurement days (in minutes), which represents Sleep Variability).
Secondary outcomes also encompassed subjective data according to the parent-completed questionnaires, including data on infant sleep (24-hour Total Sleep, Number of Nighttime Awakenings, Duration of Daytime Sleep, and Sleep Problems), the parents’ sleep quality, and the behavioral development at ages 3 and 6 months.
Sample size calculation
Determination of the sample size was based on a two-sample t-test power calculation with a two-sided significance level of 0.05 to give greater than or equal to 80% power to detect differences of ≥25% in the Number of Nighttime Awakenings in propranolol-treated infants versus controls at 6 months. This resulted in 44 required datasets per group.
Statistical analysis was done using R  and RStudio  and the packages ggplot2, MASS, mosaic, kableExtra, pander, lme4, lmerTest, cowplot, purr, qwraps2, psych, and dpylr [34,35,36,37,38,39]. We compared the demographics of the control and propranolol cohort by applying chi-square and t-tests. The efficacy of treatment (baseline, T1, and T2) was assessed using a multilevel model (lmer), with varying intercepts for each patient. Differences in objective and subjective sleep variables, as well as parental sleep quality and behavioral development, were assessed using linear models with group (propranolol/control) as predictors and exact age at measurement start, sex, and gestational age at birth as control variables. Each timepoint was analyzed separately, as we did not have complete datasets for either timepoint, and not enough timepoints to utilize a multilevel model. Alpha level was set to p < 0.05. Because we included many objective and subjective sleep variables, we controlled for multiple comparisons by applying a False Discovery Rate (FDR) correction (Benjamini-Hochberg).