The objective of the SafeBoosC-III follow-up study is to investigate the benefits and harms of treatment guided by cerebral oximetry monitoring in extremely preterm infants during the first 72 hours after birth, assessed at 2 years’ corrected age, as compared with usual care.
The hypothesis is that the intervention will decrease a composite of death or moderate or severe neurodevelopmental disability at 2 years’ corrected age, and/or increase cognitive function in survivors assessed by the Bayley-III/IV test, with insignificant harms.
Routine follow-up varies greatly from site to site and we aim to collect data on as many children as possible. The following outcome assessment tools may be used to collect data on children in the SafeBoosC-III FU. We defined three tiers of data: (1) routine clinical data from healthcare records, (2) parental questionnaire, and (3) informal assessments of neurodevelopment. The ‘3-tier model’ also represents a principle for prioritisation of the data source for classifying the co-primary dichotomous outcome death or moderate or severe NDD.
Tier 1: routine clinical data
Tier 2: parental questionnaire
The parental questionnaires will act as a common denominator in what may be a heterogeneous clinical data set. The parental questionnaire will be comprised of the Parent Report of Children’s Abilities – Revised (PARCA-R) non-verbal cognitive scale and a health and development questionnaire.
If no formal assessments of the child have been conducted between the ages of 18 and 30 months’ corrected age, the blinded assessor will provide an informal assessment of the presence or absence of moderate or severe NDD based on all available information on the child from at least 12 months’ corrected age. If no formal neurodevelopmental assessment (i.e. Bayley, Griffiths) of the child has been conducted, the local co-investigator will also provide an informal assessment of moderate or severe NDD.
For additional details on the 3-tier model, see Appendix
C.
Power estimations for the co-primary outcomes
The sample size calculation of 1600 participants for the SafeBoosC-III trial was based on mortality and the prevalence of survivors with severe brain injury in the SafeBoosC-II trial, i.e. a 22% relative risk reduction in the composite primary outcome from 34 to 26.5% (absolute risk reduction of 7.5%), at an alpha level of 5%, and a power of 90% [
15]. Of the 1601 participants randomised, 1276 were alive at 36 weeks’ post-menstrual age and will potentially be available for the SafeBoosC-III follow-up study. The following calculations were carried out using the software OpenEpi (Dean AG, Sullivan KM, Soe MM. OpenEpi: Open Source Epidemiologic Statistics for Public Health, Version.)
Death or moderate or severe neurodevelopmental disability
Based on answers to a questionnaire on systematic routine follow-up from participating sites, as well as implementing parental questionnaires and informal assessments to classify neurodevelopment, we believe it is reasonable to expect a minimal loss to follow-up and expect the total sample size for the outcome death or moderate or severe NDD to near 1600 participants.
Based on results from two randomised clinical trials investigating neuroprotection in preterm infants [
29,
30], it is estimated that the proportion of children with the outcome death or moderate or severe NDD will be 50% in the usual care group. An indicative power calculation shows that if we want to test an absolute risk difference of 8%, between the experimental and control group, at an alfa of 2.5% and a sample size of 800 participants in each group, i.e. a total of 1600, we will reach a power of 80% for this outcome.
Bayley-III/IV mean cognitive score
Based on answers to a questionnaire on systematic routine follow-up, it is expected that 65% of the sites will be able to provide data from a Bayley-III/IV assessment around 2 years’ corrected age for each child. Assuming that these sites enrol their proportion of the eligible participants to the SafeBoosC-III trial, approximately 850 participants will be available for Bayley-III/IV assessments at 2 years of age.
An indicative power estimation shows that if we want to test a mean difference of five points on the Bayley cognitive score, with a standard deviation of 20 (Cohens d’ 0.25) between the cerebral oximetry and the usual care group, at a 2.5% alfa-level, with a sample size of 425 participants from each group in the follow-up study, i.e. a total of 850, we will reach a power of 90%.
Primary analyses
The primary analyses of all outcomes will be based on the intention-to-treat population. Mixed-effect linear regression and mixed-effect logistic regression will be used to analyse the dichotomous and continuous co-primary outcomes, respectively. In the regression models, ‘site’ will be included as a random effect, while ‘gestational age below or above 26 weeks of postmenstrual age’ and ‘group allocation’ will be included as fixed effects. To correct for multiple testing, the threshold for statistical significance will undergo Bonferroni adjustment, and thus a p-value of 0.025 for each primary outcome is chosen. Superiority of the intervention will only be claimed if at least one of the two co-primary outcomes is statistically significant at this level. All other outcome results will be considered hypothesis-generating. In addition, we will perform several pre-defined sensitivity analyses to inform the interpretation of the results of the primary analysis. A full statistical analysis plan will be developed and submitted for publication before the analysis of the SafeBoosC-III FU data.
Data extraction and coding
All data extraction from health care records and data entry into the e-CRF will be conducted by an assessor who is blinded to group allocation. The co-primary outcome Bayley-III/IV mean cognitive score will be extracted from the health care records of the Bayley-III/IV test. The co-primary outcome death or moderate or severe NDD will primarily be based on an assessment of available health care records, from when the child was between 18 and 30 months’ corrected age. If more than one formal neurodevelopmental assessment has been conducted, the score of the latest should be used. If formal assessment of all four components is not available, the missing components will be substituted with answers from parental questionnaires (PARCA-R and health and development questionnaire, if available) as a step in the final data analysis, if these are available. The data collection and management for this study will utilise the OpenClinica open source software, version 3.1. Copyright OpenClinica LLC and collaborators, Waltham, MA, USA,
www.OpenClinica.com hosted by the Copenhagen Trial Unit. The answers from the questionnaires will be reported web-based into REDCap hosted at The Capital Region of Denmark by the parents or by the investigator who provides the questionnaire in the follow-up clinic or via phone.
Blinding
Due to the nature of the trial intervention, children and their parents were not blinded to treatment allocation, and therefore, completion of parental reported questionnaires cannot be blinded. However, a co-investigator who is blinded to group allocation will look through all relevant health care records, do classifications, and report data into the eCRF. To ensure that the co-investigator who will do the outcome assessment is blinded, the principal investigator and co-investigator from each site must develop a local blinding procedure describing the workflow. This process will be approved by the central trial unit before the study commencement. Thus, all outcome data except for the parental-reported questionnaires are assessed blindedly. Data managers, statisticians, and those writing the abstracts as well as drawing conclusions will also be blinded.
Ethical review
The need for supplementary approval of the follow-up study by a Research Ethics Committee will differ among countries. In some countries, no REC approval may be necessary. In some countries, a Research Ethics Committee approval may be an addition to the approval of the randomised clinical trial. As the SafeBoosC-III follow-up study does not include exposures to any additional interventions, there are no safety risks for children. Thus, an interim analysis is not necessary.
Consent
When infants were randomised in the SafeBoosC-III trial, the parents were made aware of the possibility of a potential follow-up study and permission to contact families as foreseen in the information sheet the parents were given at inclusion. The need for explicit consent for using clinical data from health care records for the SafeBoosC-III FU may differ between countries. In some countries, turning up for examination or answering questionnaires will be considered sufficient, implicit consent. Some centres will send out consent forms by mail while some will collect consent when the families come to the routine follow-up appointments. We have supplied the participating centres with a standard consent template, which may be altered according to local needs.
Data management plan
All the participants data are protected in accordance with the Danish Act on the processing of personal data and the Danish Health Act. Data will be stored in accordance with guidelines issued by the Danish Data Protection Agency, where the follow-up study will also seek approval from. Data are pseudonymised, and only site numbers and study numbers are used to identify children. Personal identifying information linking to study numbers will be kept in the trial master file at the local site. A central monitoring plan has been developed to maximise data quality. This will be done by identifying centres with missing data. This is possible since the date of birth of the participants of the randomised trial is known. Furthermore, centres with predefined quality deficiencies or noteworthy data deviations as identified by visual inspection of data distributions. Primary investigators of these centres will be contacted to remedy deficiencies, if possible. This procedure will be re-iterated every third month. The central data monitoring plan is available in the
Supplementary Appendix on page 13.
Six months after the acceptance of the main publication presenting the primary outcome of the study, the dataset will be transferred to the Danish data archive, after re-coding of variables (birth weight, gestational age, sex, site, study number) that may be used for reidentification. This data set can be made available for others as decided by the trial steering committee. Due to the residual risk of re-identification, the dataset will not be placed in public space.
Publication plan
Once all data from the study has been analysed, the results will be published in a peer-reviewed international journal. Members of the steering group will be offered authorship. Furthermore, one investigator per participating site can obtain co-authorship. All authors must fulfil the ‘Vancouver Criteria’. The blinded assessor completing the data entries will obtain non-byline co-authorship. Ancillary studies with results potentially affecting the main study or compromise its publication shall not be published before the main results of the study have been published, as decided by the trial steering committee.
Timeline
The first randomised infant in SafeBoosC III reached 2 years of corrected age in September 2021, while the last follow-up assessment is expected to be completed around autumn of 2024, which is when the last infant randomised will reach 2 years of corrected age plus 6 months of the reserve.