Abstract
Purpose
The SafeBoosC-III trial compared treatment guided by cerebral oximetry monitoring for the first 72 h after birth with usual care in 1601 extremely preterm infants. Incidence of death or severe brain injury at 36 weeks of postmenstrual age did not differ between the cerebral oximetry and usual care group (relative risk with cerebral oximetry, 1.03; 95% CI 0.90 to 1.18). To assess the probability of clinically important benefit or harm, we conducted secondary analyses in a Bayesian framework.
Methods
Primary analyses used a weakly informative prior assuming a wide range of effects, to assess the probability that treatment guided by cerebral oximetry monitoring carried an a priori defined clinically important benefit or harm (a relative risk below 0.90 or above 1.10) or any benefit or harm (a relative risk differing from 1.0) for death or severe brain injury. Secondary analyses used an evidence-based prior derived from relevant prior trials.
Results
Posterior probabilities of clinically important benefit or harm were 1.5% and 19.2%, respectively. Posterior probabilities of any benefit or harm were 28.7% and 71.3%, respectively. Probabilities derived from secondary analyses using evidence-based priors were consistent with those from the primary analysis.
Conclusion
Treatment guided by cerebral oximetry monitoring during the first 72 h after birth in extremely preterm infants resulted in posterior probabilities of 1.5% for a clinically important benefit and 19.2% for a clinically important harm with respect to the incidence of death or severe brain injury at 36 weeks of postmenstrual age.
Trial registration
ClinicalTrials.gov NCT03770741. Registered on 12 July 2018.
Supplementary Information
The online version contains supplementary material available at 10.1186/s13063-026-09742-z.
Keywords: Randomised clinical trial, Bayesian analysis, Cerebral oximetry, Near-infrared spectroscopy, Extremely preterm, Neonatal intensive care
Introduction
Infants born extremely preterm carry a high risk of death or survival with substantial neurodevelopmental impairment [1, 2]. During the first days of life, extremely preterm infants are at risk of experiencing low systemic blood flow [3]. Combined with poorly functioning cerebral autoregulation [4, 5], perfusion to the brain can be compromised [6–8]. Cerebral oximetry monitoring may aid clinicians in detecting impaired cerebral and systemic blood flow, thereby enabling adjustment of the cardiorespiratory support accordingly, to potentially reduce the risk of organ damage [6, 8, 9]. The SafeBoosC-II trial initially tested the combination of cerebral oximetry monitoring and a treatment guideline in extremely preterm infants. A total of 166 infants were randomised to receive either treatment guided by cerebral oximetry monitoring for the first 72 h after birth or usual care. The trial demonstrated a more than 50% reduction in the accumulated burden of cerebral hypoxia and hyperoxia in the intervention period. While the cerebral oximetry group showed a trend towards reduced mortality and fewer events of severe brain injury, there was a higher incidence of bronchopulmonary dysplasia and retinopathy of prematurity, when compared to the usual care group [10]. However, the trial was insufficiently powered to detect smaller but still relevant differences in clinical outcomes. The SafeBoosC-III trial randomised 1601 extremely preterm infants to either treatment guided by cerebral oximetry monitoring for the first 72 h after birth or usual care and did not find similar differences in clinical outcomes. The primary outcome of death or survival with severe brain injury at 36 weeks’ postmenstrual age showed no difference between the cerebral oximetry and the usual care group (35.2% versus 34.0%, p = 0.64) [11, 12]. Since the SafeBoosC-III trial used a frequentist statistical design, powered to detect a relative risk difference of 22% for the primary outcome, smaller and still clinically important effects could not be excluded (relative risk with cerebral oximetry, 1.03; 95% confidence interval, 0.90 to 1.18) [13]. Arguably, the minimal clinically important difference is lower than 22% (e.g., a 10% relative risk reduction or even much less). Designing a trial similar to SafeBoosC-III, with sufficient statistical power to detect a 10% relative risk reduction would require enrolment of more than 6000 infants (https://www.openepi.com/SampleSize/SSCohort.htm). Given that it took 30 months to enrol and randomise 1601 infants across 70 neonatal intensive care units in 17 countries, conducting a trial capable of detecting a relative risk difference of 10% or less would be challenging [11]. To assess the probability of smaller effect sizes than proposed by the alternative hypothesis in SafeBoosC-III, we conducted a Bayesian analysis of SafeBoosC-III.
Methods
This Bayesian analysis was a pre-specified analysis of the SafeBoosC-III trial [11]. Reporting adheres to the Reporting of Bayes Used in Clinical Studies guideline [14]. A detailed statistical analysis plan has been published elsewhere, outlining the entire methodology for this Bayesian analysis [15]. The full statistical report is available in the Supplemental material. Before initiation of the SafeBoosC-III trial, the protocol was registered at ClinicalTrials.gov (number, NCT030770741) and the protocol was approved by the participating sites’ relevant ethics committees. The trial protocol, the frequentist statistical analysis plan and the results of the primary analysis using frequentist statistics have been published elsewhere [11–13].
The SafeBoosC-III trial
SafeBoosC-III was a pragmatic, investigator-initiated, multinational randomised clinical trial, evaluating the effects of cerebral oximetry monitoring combined with an evidence-based treatment guideline during the first 72 h after birth compared with usual care, in extremely preterm infants [12]. A total of 1601 infants were randomised across 70 neonatal intensive care units from 17 countries [11]. Eligibility criteria included gestational age below 28 weeks, decision to provide full life-support, and the possibility to initiate cerebral oximetry monitoring within 6 h from birth. Infants were randomly assigned in a 1:1 ratio, using a computer-generated, concealed allocation method with varying block size, and stratified by gestational age (26 weeks or older compared to less than 26 weeks) and site [12].
Outcomes
The primary outcome of SafeBoosC-III was a composite of death or severe brain injury diagnosed by cranial ultrasound at 36 weeks of postmenstrual age [11]. Severe brain injury was defined as one or more of the following brain injuries: intraventricular hemorrhage grade 3 or 4 [16], cystic periventricular leukomalacia [17], cerebellar hemorrhage, post-hemorrhagic ventricular dilatation, and cerebral atrophy. Data for the primary outcome analysis was available for 1579 infants (772 in the cerebral oximetry group and 807 in the usual care group). Exploratory outcomes included bronchopulmonary dysplasia, retinopathy of prematurity, necrotising enterocolitis, late-onset sepsis, and a major neonatal morbidity count (comprising up to three of the following diagnoses: bronchopulmonary dysplasia, retinopathy of prematurity or severe brain injury) [18]. Serious adverse events included both components of the primary outcome (death and severe brain injury) as well as the exploratory outcomes.
Bayesian analysis priors
For the primary analyses, we used weakly informative priors with a uniform probability distribution, assuming equal probability of a wide range of effects (mean = 0 and standard deviation = 10) (Fig. 1) [19]. Using weakly informative priors enables estimation of posterior probabilities of benefit or harm, without prior assumptions on the effect estimate influencing the posterior probabilities. For the secondary Bayesian analyses, we generated evidence-based, informative priors, incorporating data from previous randomised clinical trials on cerebral oximetry monitoring combined with a treatment guideline in extremely preterm infants (Fig. 1) [10, 20, 21].
Fig. 1.
Prior probability distributions for the priors used for the Bayesian primary outcome analyses of SafeBoosC-III. Prior probability distributions for the weakly informative (solid line) and evidence-based (dashed line) priors used for the Bayesian primary outcome analyses (death or severe brain injury at 36 weeks’ postmenstrual age) in the SafeBoosC-III trial. The evidence-based prior is based on trial data from two previous neonatal intensive care trials on cerebral oximetry-guided treatment in preterm infants, the SafeBoosC-II [10] and COSGOD-II trials [20]
Evidence-based priors were constructed using a Bayesian meta-analytic approach. For each outcome, event counts and sample sizes from all relevant trials were extracted and formatted for analysis. All trials contributed to the posterior, with larger trials naturally receiving more weight through the likelihood. Trial-level heterogeneity was accounted for using fixed-strata modeling. Posterior means and standard deviations from these models were used as priors in subsequent analyses.
The SafeBoosC-II trial randomised 166 extremely preterm infants to receive either cerebral oximetry monitoring combined with the SafeBoosC treatment guideline for the first 72 h after birth or usual care [22]. The COSGOD-II and COSGOD-III trials evaluated the effects of cerebral oximetry monitoring to guide resuscitation in the delivery room in infants with a gestational age below 34 weeks and 32 weeks, respectively [20, 21]. None of the three trials had published the incidence of the composite outcome of death or severe brain injury. Therefore, to inform the evidence-based prior, we derived data on the incidence of death or severe brain injury from the SafeBoosC-II trial through our own databases. For the COSGOD trials we contacted the corresponding author who delivered data on the composite outcome of death or severe brain injury from the COSGOD-II trial [10, 20]. Data on death or severe brain injury from COSGOD-III was not accessible [21].
Statistical analysis
Statistical analyses were performed independently by two statisticians (MHO and TH) and carried out in R version 4.4.2 (R Core Team, Vienna, Austria). Consistent with the primary, frequentist analysis, we used the intention-to-treat population for the Bayesian analyses of all outcomes [15]. The dichotomous outcomes (death or severe brain injury, bronchopulmonary dysplasia, retinopathy of prematurity, late-onset sepsis, and necrotising enterocolitis) were analysed using Bayesian logistic regression models fitted with a Bernoulli likelihood, with the stratification parameter gestational age as a covariate. The analyses with site as a covariate did not converge using weakly informative priors with a uniform probability distribution. Thus, site was excluded as a covariate. To enhance transparency, the results from the non-convergent analysis including site as a covariate can be found in the Supplemental appendix. Treatment effects are reported with posterior-derived relative risks and risk differences based on predicted probabilities with 95% credible intervals.
For the major neonatal morbidity count, a Bayesian linear regression model was used, with the same covariate as used for the dichotomous outcome analyses. Effect estimates derived from the count outcome analyses are presented as ratio of means with 95% credible intervals. The probability of clinically important benefit or harm was estimated for all outcomes. For the primary analyses of dichotomous outcomes, clinically important benefit was a priori defined as a relative risk below 0.9 and clinically important harm as a relative risk above 1.10. For the major neonatal morbidity count, clinically important benefit was defined as a ratio of means below 0.9 and clinically important harm as a ratio of means above 1.1. We also calculated the posterior probability of any benefit or harm. For dichotomous outcomes, any benefit was defined as the probability of a relative risk below 1.0, and the probability of any harm as a relative risk above 1.0. For the major neonatal morbidity count, any benefit was defined as the probability of a ratio of means below 1.0 and the probability of any harm as a ratio of means above 1.0. Sensitivity analyses of the primary outcome assessed the probability of various effect sizes, ranging from a relative risk of 0.70 to 1.30 (Table 1).
Table 1.
Sensitivity analyses of the primary outcome of death or severe brain injury using a weakly informative prior
| Relative difference (relative risk of benefit and harm) | Probability of benefit (%) | Probability of harm (%) | Probability of a relative difference smaller than the evaluated risk of benefit or harm1 |
|---|---|---|---|
| 0.03 (0.97 and 1.03) | 15.5% | 54.9% | 29.6% |
| 0.05 (0.95 and 1.05) | 9.0% | 43.4% | 47.6% |
| 0.15 (0.85 and 1.15) | 0.1% | 5.9% | 94.0% |
| 0.20 (0.8 and 1.20) | 0.0% | 1.4% | 98.6% |
| 0.25 (0.75 and 1.25) | 0.0% | 0.2% | 99.8% |
| 0.30 (0.70 and 1.30) | 0.0% | 0.0% | 100% |
Sensitivity analyses determining the posterior probabilities of benefit or harm for a range of effect sizes presented as relative risks for death or severe brain injury using a weakly informed prior. As an example, the posterior probability of a relative risk of 0.97 (benefit) or lower with cerebral oximetry is 15.5%; the posterior probability of a relative risk of 1.03 (harm) or higher with cerebral oximetry is 54.9%. and the posterior probability of a relative risk within the range of 0.97 to 1.03 is 29.6%
1As exemplified above, the’Probability of a relative difference smaller than the evaluated risk of benefit or harm’ refers to the posterior probability of a relative risk within the relative risk range in the adjacent table row in Column 1
Conclusions were made based on the Bayesian analyses of the primary outcome. As stated in the published statistical analysis plan, a probability of benefit or harm above 99% was interpreted as a ‘high probability’ [15]. Results from ‘best-worst’ and ‘worst-best’ case analyses and multiple imputation analyses in the original, frequentist statistical framework of SafeBoosC-III showed that missingness did not have the potential to substantively affect results [11]. Thus, we used complete-case analyses for all Bayesian analyses.
Results
Characteristics of the intention-to-treat population, including baseline, clinical and treatment, and intervention characteristics are available in the publication of the primary frequentist analysis [11].
Primary analysis
As in the frequentist analysis, a total of 1579 infants were included in the Bayesian analysis of the primary outcome. In the cerebral oximetry group, 272 of 772 (35.2%) infants died or survived with a severe brain injury as compared to 274 of 807 (34.0%) infants in the usual care group. The frequentist analysis resulted in a relative risk of 1.03 (95% confidence interval, 0.90 to 1.18; p = 0.64). The primary Bayesian analysis of death or severe brain injury using the weakly informative prior resulted in a relative risk of 1.04 (95% credible interval, 0.91 to 1.18). The posterior probability of clinically important benefit (relative risk below 0.9) was 1.5% and the posterior probability of clinically important harm (relative risk above 1.10) was 19.2%. The posterior probability of any benefit was 28.7%, and the posterior probability of any harm was 71.3% (Fig. 2 illustrating posterior derived relative risks with 95% credible intervals and posterior probabilities for any benefit or harm, clinically important benefit or harm and no clinically important benefit, for primary and exploratory outcomes).
Fig. 2.
Primary Bayesian analysis of all outcomes using weakly informative priors. Primary outcome: death or severe brain injury up until follow-up. Severe brain injury is defined below. Severe brain injury: one or more of the following diagnoses: intraventricular hemorrhage grade 3 or 4, cystic periventricular leukomalacia, post-hemorrhagic ventricular dilatation, cerebellar hemorrhage, or cerebral atrophy on any of the routine cerebral ultrasound scans conducted until follow-up. Major neonatal morbidity count: a count from 0 to 3 of bronchopulmonary dysplasia, retinopathy of prematurity, and severe brain injury. Effect estimate is reported as ratio of means (RoM). Bronchopulmonary dysplasia: any respiratory support and/or supplementary oxygen requirements at 36 weeks’ postmenstrual age. Missing data was mainly due to death before follow-up. Retinopathy of prematurity: stage 3 or above at any time point until follow-up. Late-onset sepsis: initiation of antibiotics later than 72 h from birth for at least five days. Necrotising enterocolitis: Stage 2 or higher based on the modified Bell’s staging system and/or focal intestinal perforation at any time point until follow-up. Clinically important benefit was defined as a RR (RoM) < 0.9 and clinically important harm as a RR (RoM) > 1.1. CrI = Credible interval
For the exploratory outcomes, mean major neonatal morbidity count was 0.85 (standard deviation 0.79) in the cerebral oximetry group and 0.88 (standard deviation 0.79) in the usual care group. A total of 322 of 583 (55.2%) and 367 of 625 (58.7%) suffered from bronchopulmonary dysplasia in the cerebral oximetry and usual care group, respectively. For retinopathy of prematurity, the incidences were 72 of 772 (9.3%) and 69 of 807 (8.6%); for late-onset sepsis, the incidences were 476 of 772 (61.7%) and 523 of 807 (64.8%); and for necrotising enterocolitis, the incidences were 94 of 772 (12.2%) and 88 of 807 (10.9%). Additional information on exploratory outcome data can be found in the publication of the primary frequentist analysis [11]. Effect sizes for the exploratory outcomes derived from the primary Bayesian analysis using the weakly informative prior were similar to those derived from the primary frequentist analysis [11]. The posterior probabilities of clinically important benefit or harm were 5.9% and 0.5% for major neonatal morbidity; 18.0% and 0.0% for bronchopulmonary dysplasia; 10.9% and 48.0% for retinopathy of prematurity; 7.6% and 0.0% for late-onset sepsis; and 6.2% and 54.1% for necrotising enterocolitis (Fig. 2). Posterior probabilities of any benefit or harm were 72.1% and 27.9% for major neonatal morbidity; 90.0% and 10.0% for bronchopulmonary dysplasia; 28.7% and 71.3% for retinopathy of prematurity; 90.2% and 9.8% for late-onset sepsis; and 21.4% and 78.6% for necrotising enterocolitis (Fig. 2).
Evidence-based priors
Secondary analysis of death or severe brain injury using the evidence-based prior resulted in a relative risk of 1.04 (95% credible interval 0.91 to 1.18), thus identical to the effect size and credible intervals obtained from the primary Bayesian analysis using the weakly informative prior. Posterior probabilities of clinically important benefit or harm were 1.7% and 18.8%, and posterior probabilities of any benefit or harm were 29.3% and 70.7%, respectively, similar to those derived from the primary analyses using a weakly informative prior (Fig. 3, illustrating posterior probabilities derived from primary analyses using the evidence-based prior and secondary analyses using the weakly informed prior). Relative risks and posterior probabilities for the exploratory outcomes were also similar to those derived from the weakly informative prior analysis (Supplemental material, Evidence-basedpriors).
Fig. 3.
Posterior probability distributions from Bayesian primary outcome analyses of SafeBoosC-III. Bayesian analyses of the primary outcome death or severe brain injury at 36 weeks’ postmenstrual age in the SafeBoosC-III trial. A The posterior probability distribution using the weakly informative prior. B The posterior probability distribution using the evidence-based prior. The dark-grey shading to the right of the right-sided dashed vertical line indicates a 19.2% (weakly informative prior) or 18.8% (evidence-based prior) probability of clinically important harm (RR > 1.1). The light-grey shading to the right of the solid vertical line indicates a 71.3% (weakly informative prior) or 70.7% (evidence-based prior) probability of any harm (RR > 1). The light-blue shading to the left of the solid vertical line indicates a 28.7% (weakly informative prior) or 29.3% (evidence-based prior) probability of any benefit (RR < 1.0). The dark-blue shading to the left of the left-sided dashed line indicates a 1.5% (weakly informative prior) and 1.7% (evidence-based prior) probability of clinically important benefit (RR < 0.9)
Sensitivity analyses
Posterior probabilities of benefit or harm for a range of effect sizes for the primary outcome are available in Table 1. As expected, probabilities for benefit or harm increased with smaller effect sizes. For relative risk differences of 0.15 or greater, the probabilities for benefit or harm were very low (Supplemental material, Sensitivity analyses). The posterior probability of benefit or harm for a relative risk difference of 0.05 was 9.0% and 43.4%, respectively, and for a relative risk difference of 0.03, the probability of benefit was 15.5% and the probability of harm 54.9%. Posterior probabilities derived from sensitivity analyses of the exploratory outcomes are available in the Supplemental material, Sensitivity analyses.
Discussion
In this Bayesian analysis of the SafeBoosC-III trial, the analyses broadly support the main conclusion from the frequentist analysis that cerebral-oximetry guided treatment in extremely preterm infants is unlikely to provide clinically important benefit or harm with respect to death or severe brain injury [11]. However, the present results yielded a posterior probability of clinically important harm exceeding that of clinically important benefit (19.2% versus 1.5%). The posterior probability of approximately 19% for clinically important harm should be interpreted with appropriate caution. Although we consider 19% to be low within the pre-specified framework, where a posterior probability greater than 99% was defined as high, the analyses do not rely on universally accepted cut-offs for categorizing probabilities as high, moderate or low. From a parental perspective and for some clinicians, a 19% probability of clinically important harm may be interpreted as a high probability—especially against a 1.5% probability of clinically important benefit.
Based on SafeBoosC-II trial data, the SafeBoosC-III trial was designed to detect a relative risk difference of 22% or greater for death or severe brain injury between the cerebral oximetry and usual care group by frequentist analysis. For this Bayesian analysis, we predefined a clinically important difference for death or severe brain injury as a relative risk difference of 10%, equivalent to an absolute risk difference of 3.5% [15]. The trial results did not support such differences. For the most severe outcomes, such as death or severe brain injury, which are associated with substantial neurodevelopmental impairment [23–25], smaller effects may be of clinical relevance [26].
The findings from SafeBoosC-III are in line with the recently published results of the COSGOD-III trial [21]. In COSGOD-III, 665 preterm infants below 32 weeks of gestational age were randomised to resuscitation in the delivery room guided by cerebral oximetry monitoring or usual care. Incidence of the primary outcome, survival without severe brain injury, did not differ between the two groups, nor did cerebral tissue oxygenation values measured during the intervention period [21]. The predecessors of SafeBoosC-III and COSGOD-III—SafeBoosC-II and COSGOD-II—showed more ‘promising’ results in terms of the potential benefit of cerebral oximetry-guided treatment [10, 20]. In COSGOD-II, 60 preterm infants below 34 weeks of gestational age were randomised to receive either resuscitation in the delivery room guided by cerebral oximetry monitoring or usual care [20]. As in SafeBoosC-II, the COSGOD-II trial showed a reduction in the burden of cerebral hypoxia and hyperoxia during the intervention period and a trend towards benefit in death or brain injury in the cerebral oximetry group. However, both trials were designed to detect a relevant difference in the burden of cerebral hypoxia and hyperoxia, and not clinical outcomes such as death or brain injury [20, 22]. It is well-documented that smaller trials tend to show larger intervention effects when compared to larger trials [27, 28]. Thus, the difference in death or brain injury between the cerebral oximetry and usual groups may be due to chance. A recently published individual patient-data meta-analysis of the COSGOD trials used a Bayesian statistical framework to evaluate whether survival without cerebral injury was more probable when cerebral oximetry monitoring guided resuscitation in the delivery room, compared with usual care [29]. Using a Bayesian hierarchical model to describe the posterior probability, the meta-analysis found a 93% probability of treatment superiority of the intervention when compared to usual care for survival without severe brain injury (a 4.5% increase in the rate of survival without severe brain injury with 95% credible interval from −0.014 to 0.105). However, treatment superiority reflected the posterior probability of any benefit, e.g., a relative risk below 1.0. If the meta-analysis had applied a threshold for superiority, similar to the clinically important difference definition used in this study, the probability of benefit would likely have been lower, given that a smaller portion of the posterior distribution would surpass such a threshold [29]. Regardless, the results of the COSGOD meta-analysis revealed a much higher posterior probability of any benefit when compared to the 28.7% posterior probability of benefit found in this study.
Conducting a trial of sufficient size to detect smaller, clinically relevant differences than the 22% relative risk reduction hypothesized in SafeBoosC-III would be challenging. Detecting the 10% relative risk reduction defined as clinically important in this Bayesian analysis, with adequate statistical power, would require approximately a fourfold increase in the sample size. To conduct such a large trial without substantially increasing costs, recruitment would need to be optimised. ‘Opt-out’ as a method of participant enrolment has previously been proposed as a means to increase recruitment rates in neonatal randomised clinical trials [30]. When the ‘opt-out’ method is used, infants within trial sites that meet the inclusion criteria will be enrolled and randomised by default unless the parents actively decide to ‘opt-out’ of the trial, which they are allowed to do at any given time point during the trial. Parents must be informed about the trial and default enrolment as well as their possibility to opt-out, before randomisation. Experiences from previous multicentre neonatal trial showed that both health professionals and parents found ‘opt-out’ as an enrolment method acceptable [31, 32]. In the SafeBoosC-III trial, participating sites were allowed to use the ‘opt-out’ method, if it was acceptable by their local research ethics board. However, only 12 of the 70 sites applied to use the ‘opt-out’ method and only six of the local research ethics boards allowed its use [33]. Thus, implementing the ‘opt-out’ method may be challenging.
A different approach to meeting the challenges of conducting large-scale trials in the extremely preterm population could be to incorporate automated data extraction from large neonatal databases in which sites already report clinical data on admitted newborns, such as the Vermont Oxford Network Global Neonatal Database [34]. Such an approach could ease the workload for site investigators and thereby potentially enable more sites to participate in a trial, making a larger sample size more obtainable. The COLLABORATIVE study is a planned, multicentre platform trial from the United Kingdom, designed to test nutritional interventions to prevent and treat necrotising enterocolitis in very preterm infants. In the COLLABORATIVE study, the investigators have planned to acquire data directly from the National Neonatal Research Database, data collection bypasses the usual process where investigators spend time reviewing clinical records and other databases to enter data in a study-specific electronic patient record form [35]. Given the potential of such a data collection setup, it will be interesting to follow the conduct of this trial.
Limitations of the SafeBoosC-III trial design [11–13] also apply to this secondary Bayesian analysis study. An important limitation is the lack of blinding of parents, clinical staff and assessment of mortality status, which may introduce a risk of bias [36]. However, the assessment of the subjective component of the primary outcome, i.e., severe brain injury, and statistical analysis, as well as interpretation of the results and abstract writing, were conducted blinded to intervention. Additionally, a previous meta-analysis including intensive care trials did not find evidence that unblinded assessment of mortality influences effect estimates [37]. Another important limitation is the increased uncertainty of the exploratory outcome results due to insufficient statistical power and multiple testing. Thus, the results from these analyses should merely be seen as hypothesis-generating, and the reader should avoid overinterpretation. Further details on the limitations of the SafeBoosC-III trial have been described elsewhere [10, 11].
Although the first version of the statistical analysis plan for this Bayesian analysis was written before the SafeBoosC-III trial results were unblinded, the final version was not submitted for publication until after unblinding [11, 15]. This may introduce a risk of selective reporting bias [38].
The SafeBoosC-III trial also has multiple strengths. These include detailed and published manuscripts on the trial design and statistical analysis plan to minimize the risk of selective reporting bias, extensive local and central data monitoring, less than 2% missing data for the primary outcome, and high external validity due to the multinational participation and pragmatic design. For this secondary analysis, a major strength is the publication of a detailed statistical analysis plan, despite being published after the primary outcome results [11, 15]. Additional details on the strength of the SafeBoosC-III trial and this secondary analysis have been described elsewhere [11, 12, 15].
Conclusion
Treatment guided by cerebral oximetry monitoring during the first 72 h after birth in extremely preterm infants resulted in posterior probabilities of 1.5% for a clinically important benefit and 19.2% for a clinically important harm with respect to the incidence of death or severe brain injury at 36 weeks of postmenstrual age.
Supplementary Information
Additional file 1: Supplementary material.
Acknowledgements
The authors would like to thank the families who let their newborn(s) participate in the trial; the clinical staff from all participating hospitals who treated and cared for the newborns in the trial; and the data monitoring and safety committee members Theis Lange, Andrew Whitelaw and James Boardman.
Collaborators
The SafeBoosC-III Trial Group consists of (listed in alphabetical order by forename) and should be defined as seen in the ‘Collaborator Names’ section in Authorship in MEDLINE (nih.gov):
Adelina Pellicer (La Paz University Hospital, Madrid, Spain);
Afif El-Kuffash (The Rotunda Hospital, Dublin, Ireland);
Agata Bargiel (Warsaw University of Medical Sciences, Warszawa, Poland);
Ana Alarcon (Hospital Sant Joan de Deu, Barcelona, Spain);
Andrew Hopper (Loma Linda University Children’s Hospital, Los Angeles, United States);
Anita Truttmann (University Hospital Center of Lausanne, Lausanne, Switzerland);
Anja Hergenhan (Children’s Hospital Lucerne, Lucerne, Switzerland);
Anja Klamer (Odense University Hospital, Odense, Denmark);
Anna Curley (National Maternity Hospital, Holles Street, Dublin, Ireland);
Anne Marie Heuchan (Royal Hospital for Children, Glasgow, United Kingdom);
Anne Smits (University Hospitals Leuven, Leuven, Belgium);
Asli Cinar Memisoglu (Marmara University Research and Educational Hospital, Istanbul, Turkey);
Barbara Krolak-Olejnik (Wroclaw Medical University, Wrocław, Poland);
Beata Rzepecka (Centrum Medycne “Ujastek” Sp. z.o.o, Kraków, Poland);
Begona Loureiro Gonzales (University Hospital Cruces, Barakaldo, Spain);
Beril Yasa (Basaksehir Cam and Sakura City Hospital, Istanbul, Turkey);
Berndt Urlesberger (University Hospital Graz, Graz, Austria);
Catalina Morales-Betancourt (12 de Octubre University Hospital, Madrid, Spain);
Chantal Lecart (Grand Hopital de Charleroi, Belgium);
Claudia Knöepfli (University Hospital Zurich, Switzerland);
Cornelia Hagmann (University Hospital Zürich, Switzerland);
David Healy (University College Cork, Cork, Ireland);
Ebru Ergenekon (Gazi University Hospital, Turkey);
Eleftheria Hatzidaki (University Hospital of Heraklion, Greece);
Elena Bergon-Sendin (12 de Octubre University Hospital, Madrid, Spain);
Eleni Skylogianni (University of Patras General Hospital, Patras, Greece);
Elzbieta Rafinska-Wazny (Centrum Medyczne “Ujastek” Sp. z o.o., Krakow, Poland);
Emmanuele Mastretta (Ospedale S.Anna – Cita della Salute e della Scienza di Torino, Turin, Italy);
Eugene Dempsey (University College Cork, Cork, Ireland);
Eva Valverde (La Paz University Hospital, Madrid, Spain);
Evangelina Papathoma (“Alexandra” University and State Maternity Hospital, Athens, Greece);
Fabio Mosca (Fondazione IRCCS Ca’ Granda Ospedale Maggiore Policlinico, Milan, Italy);
Gabriel Dimitriou (University of Patras General Hospital, Patras, Greece);
Gerhard Pichler (University Hospital Graz, Graz, Austria);
Giovanni Vento (Fondazione Policlinico Universitario A. Gemelli IRCCS, Rome, Italy);
Gitte Holst Hahn (Copenhagen University Hospital, Copenhagen, Denmark);
Gunnar Naulaers (University Hospital Leuven, Leuven, Belgium);
Guoqiang Cheng (Children's Hospital of Fudan University, Shanghai, China);
Hans Fuchs (Medical Center, University of Freiburg, Freiburg, Germany);
Hilal Ozkan (Bursa Uludag University Hospital, Bursa, Turkey);
Maria Isabel De Las Cuevas (Donostia University Hospital, Biogipuzkoa Health Research Institute, Donostia/San Sebastian, Spain);
Itziar Serrano-Vinuales (Miguel Servet University Hospital, Zaragoza, Spain);
Iwona Sadowska-Krawczenko (Collegium Medicum in Bydgoszcz, Nicolaus Cupernicus University in Torun, Bydgoszcz, Poland);
Jachym Kucera (The Institute for the Care of Mother and Child, Prague, Czech Republic);
Jakub Tkaczyk (University Hospital Motol, Prague, Czech Republic);
Jan Miletin (Coombe Women and Infant University Hospital, Dublin, Ireland);
Jan Sirc (The Institute for the Care of Mother and Child, Prague, Czech Republic);
Jana Baumgartner (Medical Center, University of Freiburg, Freiburg, Germany);
Jonathan Mintzer (Mountainside Medical Center, Montclair, United States);
Julie De Buyst (Tivoli Hospital, La Louviere, Belgium);
Karen McCall (University Hospital Wishaw, Wishaw, United Kingdom);
Konstantina Tsoni (Ippokrateion General Hospital of Thessaloniki, Thessaloniki, Greece);
Kosmas Sarafidis (Ippokrateion General Hospital of Thessaloniki, Thessaloniki, Greece);
Lars Bender (Aalborg University Hospital, Aalborg, Denmark)
Laura Serrano Lopez (Virgen De Las Nieves University Hospital, Granada, Spain);
Le Wang (Children’s Hospital of Fudan University, Shanghai, China);
Liesbeth Thewissen (University Hospital Leuven, Leuven, Belgium);
Lin Huijia (Children’s Hospital of Zheijang, Hangzhou, China);
Lina Chalak (UT Southwestern Medical Center, Dallas, United States)
Ling Yang (Hainan Women and Children’s Medical Center, Hainan, China);
Luc Cornette (AZ St. Jan Bruges, Bruges, Belgium);
Luis Arruza (Hospital Clinico San Carlos, Madrid, Spain);
Maria Wilinska (Centre of Postgraduate Medical Education, Warsaw, Poland);
Mariana Baserga (University of Utah Hospital, Salt Lake City, United States);
Marie Isabel Skov Rasmussen (Copenhagen University Hospital, Copenhagen, Denmark);
Marta Mencia Ybarra (La Paz University Hospital, Madrid, Spain);
Marta Teresa Palacio (Hospital Clínic Barcelona, Barcelona, Spain);
Martin Stocker (Children’s Hospital Lucerne, Lucerne, Switzerland);
Massimo Agosti (“Fillipo del Ponte” Hospital, Varese, Italy);
Merih Cetinkaya (Kanuni Sultan Süleyman Training and Research Hospital, Istanbul, Turkey);
Miguel Alsina (Hospital Clinic Barcelona, Barcelona, Spain);
Monica Fumagalli (Fondazione IRCCS Cà Granda Ospedale Maggiore Policlinico, Milan, Italy);
Munaf M. Kadri (Loma Linda University Children’s Hospital, Los Angeles, United States);
Mustafa Senol Akin (Ankara City Hospital, Ankara, Turkey);
Münevver Baş (Gazi University Hospital, Ankara, Turkey);
Nilgun Koksal (Bursa Uludag University Hospital, Bursa, Turkey);
Olalla Otero Vaccarello (Hospital Universitario De Tarragona Juan XXIII, Tarragona, Spain);
Olivier Baud (Children’s University Hospital of Geneva, Geneva, Switzerland);
Pamela Zafra (Puerta Del Mar University Hospital, Cadiz, Spain);
Peter Agergaard (Aarhus University Hospital, Aarhus, Denmark);
Peter Korcek (The Institute for the Care of Mother and Child, Prague, Czech Republic);
Pierre Maton (Clinique CHC Montlegia, Liege, Belgium);
Rebeca Sanchez-Salmador (La Paz University Hospital, Madrid, Spain);
Ruth del Rio Florentino (Hospital De Sant Joan De Deu, Barcelona, Spain);
Ryszard Lauterbach (Jagiellonian University Hospital, Kraków, Poland);
Salvador Piris Borregas (12 De Octubre University Hospital, Madrid, Spain);
Saudamini Nesargi (St Johns Medical College Hospital, Karnataka, India);
Serife Suna Oguz (Ankara City Hospital, Ankara, Turkey);
Shashidhar Appaji Rao (St Johns Medical College Hospital, Karnataka, India);
Shujuan Zeng (Longgang District Central Hospital, Shenzhen, China);
Silvia Pisoni (Fondazione IRCCS Cà Granda Ospedale Maggiore Policlinico, Milan, Italy);
Simon Hyttel-Sørensen (Copenhagen University Hospital, Copenhagen, Denmark);
Sinem Gulcan Kersin (Marmara University Research and Education Hospital, Istanbul, Turkey);
Siv Fredly (Oslo University Hospital, Oslo, Norway);
Sylvia Marciniak (Specialist Hospital No.2, Bytom, Poland);
Tanja Karen (University Hospital Zürich, Zürich, Switzerland);
Tomasz Szczapa (Poznan University of Medical Sciences, Poznan, Poland);
Tone Nordvik (Oslo University Hospital, Oslo, Norway);
Veronika Karadyova (University Hospital Motol, Prague, Czech Republic);
Xiaoyan Gao (Maternal and Child Health Hospital of Zhuang Autonomous Region, Guangxi, China);
Xin Xu (Xiamen Children’s Hospital, Xiamen, China);
Zachary Vesoulis (St. Louis Children’s Hospital, St. Louis, United States);
Zhang Peng (Children’s Hospital of Fudan University, Shanghai, China);
and Zhaoqing Yin (The People’s Hospital of Dehong, Dehong, China).
Authors’ contributions
MHO, MLH, JCJ, GG, LT, TL and CG designed the study and data was collected by the collaborators (see below). Statistical analysis was performed by MHO, TH and RC. The first version of the manuscript was written by MLH and MHO. All authors revised the manuscript, and all authors and collaborators approved the final version of the manuscript.
Funding
Open access funding provided by Copenhagen University. The SafeBoosC-III trial are supported by The Elsass Foundation, Svend Andersen Foundation, and Aage and Johanne Louis-Hansen Foundation. Section for Biostatistics and Evidence-Based Research, the Parker Institute, Bispebjerg and Frederiksberg Hospital is supported by a core grant from the Oak Foundation (OFIL-24-074).
Data availability
The datasets used during the current study are available from the corresponding author on reasonable request, which will be decided by the SafeBoosC-III steering committee.
Declarations
Ethics approval and consent to participate
All sites that participated in the SafeBoosC-III trial obtained approval of the trial protocol by their local ethics committee, before initiation of the trial. No participants were enrolled before parental consent was obtained, unless sites had obtained approval by their local ethics committee to use deferred consent or prior assent (opt-out). When deferred consent was used, consent was obtained after a randomisation. When opt-out was used, parents were informed about the trial and that their infant would be randomised by default, unless they decided to ‘opt-out’.
Consent for publication
All authors and collaborators approved the manuscript for publication.
Competing interests
The authors declare that they have no competing interests. Disclosure forms are published online.
Footnotes
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Contributor Information
Mathias Lühr Hansen, Email: mathias.luehr.hansen@regionh.dk.
SafeBoosC-III Trial Group:
Adelina Pellicer, Afif El-Kuffash, Agata Bargiel, Ana Alarcon, Andrew Hopper, Anita Truttmann, Anja Hergenhan, Anja Klamer, Anna Curley, Anne Marie Heuchan, Anne Smits, Asli Cinar Memisoglu, Barbara Krolak-Olejnik, Beata Rzepecka, Begona Loureiro Gonzales, Beril Yasa, Berndt Urlesberger, Catalina Morales-Betancourt, Chantal Lecart, Claudia Knöepfli, Cornelia Hagmann, David Healy, Ebru Ergenekon, Eleftheria Hatzidaki, Elena Bergon-Sendin, Eleni Skylogianni, Elzbieta Rafinska-Wazny, Emmanuele Mastretta, Eugene Dempsey, Eva Valverde, Evangelina Papathoma, Fabio Mosca, Gabriel Dimitriou, Gerhard Pichler, Giovanni Vento, Gitte Holst Hahn, Gunnar Naulaers, Guoqiang Cheng, Hans Fuchs, Hilal Ozkan, Isabel De Las Cuevas, Itziar Serrano-Vinuales, Iwona Sadowska-Krawczenko, Jachym Kucera, Jakub Tkaczyk, Jan Miletin, Jan Sirc, Jana Baumgartner, Jonathan Mintzer, Julie De Buyst, Karen McCall, Konstantina Tsoni, Kosmas Sarafidis, Lars Bender, Laura Serrano Lopez, Le Wang, Liesbeth Thewissen, Lin Huijia, Lina Chalak, Ling Yang, Luc Cornette, Luis Arruza, Maria Wilinska, Mariana Baserga, Marie Isabel Skov Rasmussen, Marta Mencia Ybarra, Marta Teresa Palacio, Martin Stocker, Massimo Agosti, Merih Cetinkaya, Miguel Alsina, Monica Fumagalli, Munaf M. Kadri, Mustafa Senol Akin, Münevver Baş, Nilgun Koksal, Olalla Otero Vaccarello, Olivier Baud, Pamela Zafra, Peter Agergaard, Peter Korcek, Pierre Maton, Rebeca Sanchez-Salmador, Ruth del Rio Florentino, Ryszard Lauterbach, Salvador Piris Borregas, Saudamini Nesargi, Serife Suna Oguz, Shashidhar Appaji Rao, Shujuan Zeng, Silvia Pisoni, Simon Hyttel-Sørensen, Sinem Gulcan Kersin, Siv Fredly, Sylvia Marciniak, Tanja Karen, Tomasz Szczapa, Tone Nordvik, Veronika Karadyova, Xiaoyan Gao, Xin Xu, Zachary Vesoulis, Zhang Peng, and Zhaoqing Yin
References
- 1.Adams-Chapman I, Heyne RJ, DeMauro SB, Duncan AF, Hintz SR, Pappas A, et al. Neurodevelopmental impairment among extremely preterm infants in the Neonatal Research Network. Pediatrics. 2018;141:e20173091. 10.1542/peds.2017-3091. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Stoll BJ, Hansen NI, Bell EF, Shankaran S, Laptook AR, Walsh MC, et al. Neonatal outcomes of extremely preterm infants from the NICHD Neonatal Research Network. Pediatrics. 2010;126:443–56. 10.1542/peds.2009-2959. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Kluckow M. The pathophysiology of low systemic blood flow in the preterm infant. Front Pediatr. 2018. 10.3389/fped.2018.00029. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Greisen G. Autoregulation of cerebral blood flow in newborn babies. Early Hum Dev. 2005;81:423–8. 10.1016/j.earlhumdev.2005.03.005. [DOI] [PubMed] [Google Scholar]
- 5.Soul JS, Hammer PE, Tsuji M, Saul JP, Bassan H, Limperopoulos C, et al. Fluctuating pressure-passivity is common in the cerebral circulation of sick premature infants. Pediatr Res. 2007;61:467–73. 10.1203/pdr.0b013e31803237f6. [DOI] [PubMed] [Google Scholar]
- 6.Sortica da Costa C, Cardim D, Molnar Z, Kelsall W, Ng I, Czosnyka M, et al. Changes in hemodynamics, cerebral oxygenation and cerebrovascular reactivity during the early transitional circulation in preterm infants. Pediatr Res. 2019;86:247–53. 10.1038/s41390-019-0410-z. [DOI] [PubMed] [Google Scholar]
- 7.Munro MJ, Walker AM, Barfield CP. Hypotensive extremely low birth weight infants have reduced cerebral blood flow. Pediatrics. 2004;114:1591–6. 10.1542/peds.2004-1073. [DOI] [PubMed] [Google Scholar]
- 8.Martini S, Frabboni G, Rucci P, Czosnyka M, Smielewski P, Galletti S, et al. Cardiovascular and cerebrovascular responses to cardio-respiratory events in preterm infants during the transitional period. J Physiol. 2020;598:4107–19. 10.1113/JP279730. [DOI] [PubMed] [Google Scholar]
- 9.Pellicer A, de Boode W, Dempsey E, Greisen G, Mintzer J, Naulaers G, et al. Cerebral near-infrared spectroscopy guided neonatal intensive care management for the preterm infant. Pediatr Res. 2024. 10.1038/s41390-024-03649-8. [DOI] [PubMed]
- 10.Hyttel-Sorensen S, Pellicer A, Alderliesten T, Austin T, van Bel F, Benders M, et al. Cerebral near infrared spectroscopy oximetry in extremely preterm infants: phase II randomised clinical trial. BMJ. 2015;350:g7635–g7635. 10.1136/bmj.g7635. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Hansen ML, Pellicer A, Hyttel-Sørensen S, Ergenekon E, Szczapa T, Hagmann C, et al. Cerebral oximetry monitoring in extremely preterm infants. N Engl J Med. 2023;388:1501–11. 10.1056/NEJMoa2207554. [DOI] [PubMed] [Google Scholar]
- 12.Hansen ML, Pellicer A, Gluud C, Dempsey E, Mintzer J, Hyttel-Sørensen S, et al. Cerebral near-infrared spectroscopy monitoring versus treatment as usual for extremely preterm infants: a protocol for the SafeBoosC randomised clinical phase III trial. Trials. 2019;20:811. 10.1186/s13063-019-3955-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Hansen ML, Pellicer A, Gluud C, Dempsey E, Mintzer J, Hyttel-Sorensen S, et al. Detailed statistical analysis plan for the SafeBoosC III trial: a multinational randomised clinical trial assessing treatment guided by cerebral oxygenation monitoring versus treatment as usual in extremely preterm infants. Trials. 2019;20:746. 10.1186/s13063-019-3756-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Sung L, Hayden J, Greenberg ML, Koren G, Feldman BM, Tomlinson GA. Seven items were identified for inclusion when reporting a Bayesian analysis of a clinical study. J Clin Epidemiol. 2005;58:261–8. 10.1016/j.jclinepi.2004.08.010. [DOI] [PubMed] [Google Scholar]
- 15.Olsen MH, Hansen ML, Lange T, Gluud C, Thabane L, Greisen G, et al. Detailed statistical analysis plan for a secondary Bayesian analysis of the SafeBoosC-III trial: a multinational, randomised clinical trial assessing treatment guided by cerebral oximetry monitoring versus usual care in extremely preterm infants. Trials. 2023;24:737. 10.1186/s13063-023-07720-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Papile L-A, Burstein J, Burstein R, Koffler H. Incidence and evolution of subependymal and intraventricular hemorrhage: a study of infants with birth weights less than 1,500 gm. J Pediatr. 1978;92:529–34. 10.1016/S0022-3476(78)80282-0. [DOI] [PubMed] [Google Scholar]
- 17.Volpe JJ. Brain injury in the premature infant: neuropathology, clinical aspects, pathogenesis, and prevention. Clin Perinatol. 1997;24:567–87. 10.1016/S0095-5108(18)30159-3. [PubMed] [Google Scholar]
- 18.Holsti A, Serenius F, Farooqi A. Impact of major neonatal morbidities on adolescents born at 23–25 weeks of gestation. Acta Paediatr. 2018;107:1893–901. 10.1111/apa.14445. [DOI] [PubMed] [Google Scholar]
- 19.Goligher EC, Heath A, Harhay MO. Bayesian statistics for clinical research. Lancet. 2024. 10.1016/S0140-6736(24)01295-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Pichler G, Urlesberger B, Baik N, Schwaberger B, Binder-Heschl C, Avian A, et al. Cerebral oxygen saturation to guide oxygen delivery in preterm neonates for the immediate transition after birth: a 2-center randomized controlled pilot feasibility trial. The Journal of Pediatrics. 2016;170:73-78.e4. 10.1016/j.jpeds.2015.11.053. [DOI] [PubMed] [Google Scholar]
- 21.Pichler G, Goeral K, Hammerl M, Perme T, Dempsey EM, Springer L, et al. Cerebral regional tissue oxygen saturation to guide oxygen delivery in preterm neonates during immediate transition after birth (COSGOD III): multicentre randomised phase 3 clinical trial. BMJ. 2023;e072313. 10.1136/bmj-2022-072313. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Hyttel-Sorensen S, Austin T, van Bel F, Benders M, Claris O, Dempsey E, et al. A phase II randomized clinical trial on cerebral near-infrared spectroscopy plus a treatment guideline versus treatment as usual for extremely preterm infants during the first three days of life (SafeBoosC): study protocol for a randomized controlled trial. Trials. 2013;14:120. 10.1186/1745-6215-14-120. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Horsch S, Muentjes C, Franz A, Roll C. Ultrasound diagnosis of brain atrophy is related to neurodevelopmental outcome in preterm infants. Acta Paediatr. 2005;94:1815–21. 10.1111/j.1651-2227.2005.tb01858.x. [DOI] [PubMed] [Google Scholar]
- 24.O’Shea TM, Kuban KCK, Allred EN, Paneth N, Pagano M, Dammann O, et al. Neonatal cranial ultrasound lesions and developmental delays at 2 years of age among extremely low gestational age children. Pediatrics. 2008;122:e662-9. 10.1542/peds.2008-0594. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Nongena P, Ederies A, Azzopardi DV, Edwards AD. Confidence in the prediction of neurodevelopmental outcome by cranial ultrasound and MRI in preterm infants. Arch Dis Child Fetal Neonatal Ed. 2010;95:F388–90. 10.1136/adc.2009.168997. [DOI] [PubMed] [Google Scholar]
- 26.McGlothlin AE, Lewis RJ. Minimal clinically important difference: defining what really matters to patients. JAMA. 2014;312:1342. 10.1001/jama.2014.13128. [DOI] [PubMed] [Google Scholar]
- 27.Dechartres A, Trinquart L, Boutron I, Ravaud P. Influence of trial sample size on treatment effect estimates: meta-epidemiological study. BMJ. 2013;346:f2304–f2304. 10.1136/bmj.f2304. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Ioannidis JPA. Why most discovered true associations are inflated. Epidemiology. 2008;19:640–8. 10.1097/EDE.0b013e31818131e7. [DOI] [PubMed] [Google Scholar]
- 29.Bruckner M, Suppan T, Suppan E, Schwaberger B, Urlesberger B, Goeral K, et al. Brain oxygenation monitoring during neonatal stabilization and resuscitation and its potential for improving preterm infant outcomes: a systematic review and meta-analysis with Bayesian analysis. Eur J Pediatr. 2025;184:305. 10.1007/s00431-025-06138-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Gale C, Hyde MJ, Modi N. Research ethics committee decision-making in relation to an efficient neonatal trial. Arch Dis Child Fetal Neonatal Ed. 2017;102:F291–8. 10.1136/archdischild-2016-310935. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.McLeish J, Alderdice F, Robberts H, Cole C, Dorling J, Gale C. Challenges of a simplified opt-out consent process in a neonatal randomised controlled trial: qualitative study of parents’ and health professionals’ views and experiences. Arch Dis Child Fetal Neonatal Ed. 2021;106:244–50. 10.1136/archdischild-2020-319545. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Mitchell T, Andrzejewska I, Battersby C, Cole C, Daskalopoulou Z, Dorling J, et al. Parent and practitioner experiences of opt-out consent in neonatal intensive care: a mixed methods study within a trial. Arch Dis Child Fetal Neonatal Ed. 2025;fetalneonatal-2025–328693. 10.1136/archdischild-2025-328693. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Vestager ML, Hansen ML, Greisen G, Pellicer A, Chathasaigh CN, Lecart C, et al. Alternative consent methods used in the multinational, pragmatic, randomised clinical trial SafeBoosC-III. Trials. 2024;25:236. 10.1186/s13063-024-08074-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Edwards EM, Ehret DEY, Soll RF, Horbar JD. Vermont Oxford Network: a worldwide learning community. Transl Pediatr. 2019;8:182–92. 10.21037/tp.2019.07.01. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Modi N, Chehrazi M, Boardman J, Boyd A, Bradley P, Cirelli C, et al. Designing a precision-medicine platform trial to improve the nutritional care and intestinal health of very preterm babies: the COLLABORATE study. Efficacy and mechanism evaluation. 2025;1–18. 10.3310/LMHT3521. [PubMed]
- 36.Savović J, Turner RM, Mawdsley D, Jones HE, Beynon R, Higgins JPT, et al. Association between risk-of-bias assessments and results of randomized trials in Cochrane Reviews: the ROBES meta-epidemiologic study. Am J Epidemiol. 2018;187:1113–22. 10.1093/aje/kwx344. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Anthon CT, Granholm A, Perner A, Laake JH, Møller MH. No firm evidence that lack of blinding affects estimates of mortality in randomized clinical trials of intensive care interventions: a systematic review and meta-analysis. J Clin Epidemiol. 2018;100:71–81. 10.1016/j.jclinepi.2018.04.016. [DOI] [PubMed] [Google Scholar]
- 38.Sterne JAC, Savović J, Page MJ, Elbers RG, Blencowe NS, Boutron I, et al. RoB 2: a revised tool for assessing risk of bias in randomised trials. BMJ. 2019;l4898. 10.1136/bmj.l4898 [DOI] [PubMed]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Additional file 1: Supplementary material.
Data Availability Statement
The datasets used during the current study are available from the corresponding author on reasonable request, which will be decided by the SafeBoosC-III steering committee.