Oral morphine analgesia for preventing pain during invasive procedures in non-ventilated premature infants in hospital: the Poppi RCT

Screening

Infants were considered eligible for the trial if they met the following inclusion criteria:

• were inpatients on the neonatal unit at the John Radcliffe Hospital, Oxford

• were born at < 32 weeks’ gestation or at a birthweight of < 1501 g

• at the time of the study, were 34–42 weeks gestational age

• required a clinical heel lance and ROP screening on the same test occasion

• their parents/guardians had given written informed consent for inclusion in the trial

• a senior clinician considered inclusion in trial to be medically appropriate.

Infants were considered ineligible for the trial if they met any of the following exclusion criteria:

• had an intraventricular haemorrhage (IVH) > grade II

• had short-bowel syndrome

• were receiving nil by mouth because of documented gut pathology

• had received opiates in the last 72 hours

• had received other analgesics or sedatives in the last 24 hours

• had a previously documented episode of morphine sensitivity

• had congenital malformation or a genetic condition known to affect neurological development

• were born to mothers who regularly used opiates during pregnancy or while breastfeeding or while expressing breast milk.

Training in the protocol and eligibility criteria was provided at the site initiation visit, and the secure randomisation website included eligibility checks. This system also highlighted ineligible infants at the point of randomisation, ensuring that staff reviewed this information and could not continue with the randomisation process unless correct.

A screening log (see Report Supplementary Material 8) was maintained throughout the duration of the trial, recording all infants born at < 32 weeks’ gestation or with a birthweight of < 1501 g. The screening log was updated at least once a week by the research staff, documenting information about the infant’s medical record number, date of birth, initials, gestational age at birth, sex, expected date of first ROP examination, eligibility status for the trial, the date the parent information leaflet (PIL; see Report Supplementary Material 6) was given, consent provided, and eligibility status at the proposed time of randomisation. The sponsor reviewed the screening log (see Report Supplementary Material 8) and participant eligibility during monitoring visits; any concerns were stated in the monitoring report and issued to the research team for resolution.

Parental approach

Recruitment was carried out by the research team and health-care professionals who had received Poppi trial training and Good Clinical Practice (GCP) training (as documented on the training log; see Report Supplementary Material 11 and 12), and were signed off to take consent on the delegation log (see Delegation; see Report Supplementary Material 4). To ensure that potentially eligible infants were not missed, Neonatal Intensive Care Unit (NICU) staff were informed about the Poppi trial to enable them to help with identification and recruitment. Promotional materials, such as roll banners, stating the trial eligibility criteria were on display around the NICU (see Improving consent rate).

The research team tried to identify eligible infants as early as possible during their hospitalisation. Following discussion with the clinical team in charge of the infant’s care, parents were approached by the research team and provided with a copy of the PIL for reference (see Appendix 4 and Report Supplementary Material 6). The eligibility of the infant was also documented in the infant’s medical notes at this time by a senior clinician. The research team ensured that parents were given plenty of time to consider participation in the trial, and arranged follow-up meetings for further discussion and to answer any questions, to ensure that parents gained a thorough understanding of the trial prior to consent.

During these meetings, the researchers introduced themselves and described their role within paediatric research at the John Radcliffe Hospital, Oxford. The researchers explained that the research team investigates brain development and how infants respond to sensory stimuli, in particular pain, and discussed the importance of this work to enable the identification of drugs that can relieve pain caused by essential clinical procedures required in premature infants in neonatal intensive care. The researchers informed the parents of the eligibility of their infant to participate in the trial and enquired whether or not the parents would be interested in receiving further details. If parents declined, the researchers did not provide further information, and the screening log (see Report Supplementary Material 8) was updated to confirm that the parents had declined. However, if parents were interested the researchers clarified that participation was voluntary, without obligation, and withdrawal from the trial was possible at any time. The parents were informed that the trial had been reviewed by an ethics committee, was funded by the Wellcome Trust and the National Institute for Health Research (NIHR) Efficacy and Mechanism Evaluation (EME) programme, and sponsored by the University of Oxford.

The parents were given an overview of the eligibility criteria and an explanation of the rationale behind the trial. The researcher explained that the trial was investigating whether or not morphine could provide pain relief for premature infants. They discussed the widespread use of morphine for pain relief in adults and older children, and its intravenous use in ventilated preterm infants. Furthermore, the researcher explained that morphine is not currently used for short-term pain relief for procedures in infants. The researcher stated two aims of the trial: to identify whether or not morphine reduces infant pain scores following ROP screening, and whether or not morphine reduces pain-specific brain responses to heel lancing.

The researcher explained that the trial was a RCT and, therefore, half of the infants taking part would be randomly allocated to receive morphine and half would receive an inactive placebo of the same volume (containing the additive ingredients of the morphine solution only). The researcher explained that the infants would be allocated to a treatment using a computer system and that the trial was blinded, which meant that nobody would know whether the infant had received morphine or placebo, to prevent bias in the analysis. The researcher always clarified, however, that in an emergency the clinical staff could find out which treatment the infant had been given.

Parents were informed that the trial involved monitoring the infants for 24 hours before and 24 hours after their ROP screen, using the same type of leads and machine used clinically to monitor their infant’s heart rate, breathing rate and oxygen saturations. The infant would be given the study drug by mouth or via their feeding tube approximately 1 hour before their planned heel lance and ROP screen. After receiving the study drug, the research team would set up EEG monitoring, which measures brain activity, by placing 10 electrodes (small discs) on the infant’s head using a soluble paste. Sticker electrodes would also be placed on the infant’s thighs to measure their leg reflex away from the stimulus. The researcher explained that these measures would record the infant’s brain activity and leg activity during heel lancing and ROP screening. The researcher described the blood sampling procedure, which would be conducted using a heel lance as standard on the unit, around 5–10 minutes before the ROP screen. Parents were also told that before the heel lance for blood sampling, their infant would also have a control heel lance, which involved the lancet being held against the heel of the foot with the end pointing away from the foot so that when the device was depressed the blade did not pierce the skin. This provided a non-painful comparison with the actual heel lance.

The researcher told the parents that the team would video their infant during the procedures to score the infant’s pain using recordings of their facial expression, and changes in their heart rate and oxygen saturations. The infant would then have their routine ROP screen conducted as usual by one of the ophthalmologists. The EEG and EMG recording would carry on throughout this, and the infant’s face would be videoed again for approximately 2 minutes in order to score the pain after the ROP screen. At this point, the team would settle the infant if needed and take the EEG off. The researcher then explained that the infant would be monitored for a further 24 hours for any effects of the drug or ROP screen (such as an increase in desaturations or bradycardias). Parents were told that the research team would communicate with the clinical team if any problems were identified on the recordings. The researcher also explained that the infant would be randomised only once and studied on one occasion. Parents were told that the team aimed to recruit 156 participants over a 3-year period.

Advantages and disadvantages of participating were discussed. The disadvantages outlined included potential reactions to morphine. The researcher explained that the team would monitor the infant closely for reactions and would work with the clinical staff on the ward to ensure the infant’s safety. There were no direct advantages to taking part, as the team did not know whether or not morphine reduced the effects of the ROP screen and blood test, but the results would help improve understanding of how infants respond to pain and morphine, and demonstrate whether or not morphine is effective at reducing pain in infants. The researcher explained that the team would publish the results of the trial so that they could be used to inform clinical practice and further research. Parents were told that they would be sent a copy of the results once available.

Finally, the research team also explained to parents that the team would review their infant’s medical notes in order to record their birthweight, gestation, age and relevant medical history on a secure online system. Parents were informed that other members of the research team, trial sponsor, regulatory authorities and Oxford University Hospitals (OUH) NHS Foundation Trust may also review relevant sections of the medical notes. Each infant would be referred in documents using an allocated study number, to preserve anonymity. All relevant medical information and video footage would be kept securely, and personal identifiable data would be treated confidentially and according to UK legislation. Parents were also told that if they decided not to enter the trial, their infant would receive the same care that other infants on the neonatal unit receive.

Informed consent

Written consent was obtained before each infant was randomised. Only the mother, father, or person designated by legal process, could sign the consent form (see Appendix 4 and Report Supplementary Material 2). Fathers with parental responsibility could sign a consent form if married to the mother, named on the birth certificate, or if granted parental responsibility through a court order or parental responsibility agreement.

In the case of twins or triplets, each infant had a separate signed consent form. The researcher taking informed consent verified that the parents had read and understood the PIL (see Report Supplementary Material 6), and asked the parents to outline their understanding of what the study involved. The researcher clarified any missing or inaccurate details, before reiterating each item on the consent form (see Report Supplementary Material 2). Parents initialled each box of the consent form and the full names of both the parent and the health professional were clearly recorded with the date.

Following randomisation (see Randomisation), the consent form (see Report Supplementary Material 2) was annotated, and the screening log (see Report Supplementary Material 8) and enrolment log (see Report Supplementary Material 5) were updated. The four copies of the consent form were distributed as follows: the original was sent to the co-ordinating centre [i.e. the National Perinatal Epidemiology Unit Clinical Trials Unit (NPEU CTU), University of Oxford, Oxford, UK], one copy was given to a parent, one copy was filed in the investigator site file (ISF) and one copy was filed in the infant’s medical notes together with a copy of the PIL (see Report Supplementary Material 6) and randomisation details. The recruitment process is outlined in Figure 3.

FIGURE 3.

The recruitment process.

Training and reference documentation for the process of obtaining informed consent, and ensuring accurate formal patient identification, were provided for the site team. This was to ensure that consent forms (see Report Supplementary Material 2) were completed accurately and according to GCP. Consent forms were also checked by the trial manager before filing in the trial master file (TMF), to ensure that they were completed accurately by parents and health professionals. The trial manager also checked that all individuals taking consent had been delegated responsibility by the chief investigator to do so. Copies of delegation logs (see Report Supplementary Material 4) and training logs (see Report Supplementary Material 11 and 12) were supplied to the trial manager on a regular basis. Language barriers were not a concern because the neonatal unit had access to face-to-face interpreters and a telephone interpretation line if required.

Randomisation

Infants were randomised as close as possible to the start of the 48-hour monitoring period (24 hours prior to a clinical heel lance). Eligibility was re-assessed at this time and documented in the medical notes.

Randomisation of participants to morphine sulphate or placebo was managed via a secure web-based randomisation facility hosted by the co-ordinating centre. Participants had an equal chance of receiving morphine sulphate or placebo, and a minimisation algorithm was used to ensure that there was an approximate balance between the groups with respect to gestational age at the time of randomisation, gestational age at birth, number of days on a ventilator, presence of a gastric tube at the time of randomisation, number of days since morphine had been given, intrauterine growth restriction and previous surgery. The minimisation algorithm used a probabilistic mechanism to increase unpredictability and the users of the system had no insight into the next allocation. The balance between the groups for each of the minimisation variables was reviewed by the Data Monitoring Committee (DMC; see Data Monitoring Committee) and reported in the trial publication.

Following confirmation of eligibility and parental informed consent, a researcher randomised the infant by logging into the randomisation program with a username and password at https://rct.npeu.ox.ac.uk/poppi or via the Poppi trial website (www.npeu.ox.ac.uk/poppi) using the ‘randomisation’ link. The researcher verified the inclusion or exclusion criteria to confirm eligibility. The infant was recruited and randomised into the trial only following confirmation that the entered data were correct. The randomisation program provided a five-digit numeric study number and five-character alphanumeric pack identifier (ID) (relating to morphine or placebo). The researcher subsequently checked that the correct pack ID was available and unopened in the temperature-controlled cupboard in the NICU (see Stock management and storage). If it could not be located, or the package was open, the researcher could select ‘Allocate other pack’ from the main menu and the randomisation program allocated a new pack ID of the same type as the original allocation (i.e. active or placebo). The researcher printed two copies of the randomisation details, filing one in the infant’s medical notes and one in the ISF. The researcher added the study number to the consent form (see Report Supplementary Material 2) and the Investigational Medicinal Product (IMP) label; the study number and pack ID to the IMP accountability log (see Report Supplementary Material 1) and the IMP compliance assessment and accountability form (see Stock management and storage); and the pack ID to the infant’s drug chart or electronic prescribing system.

If there were problems accessing the online randomisation program during normal working hours, the research team contacted the co-ordinating centre by telephone (the co-ordinating centre could carry out the randomisation on behalf of the research team). The trial manager would report system failures to the programming team, and they would be investigated by the senior trials programmer. Following randomisation, the researcher placed a cot card (see Report Supplementary Material 3) on the infant’s cot and a notes sticker on the infant’s medical notes, and ensured that the clinical team were aware that the infant was enrolled in the Poppi trial. The trial’s randomisation process is outlined in Figure 4.

FIGURE 4.

Randomisation process.

Recruitment challenges

Throughout the trial, the research team encountered a number of challenges to recruitment. A summary of recruitment is presented in Chapter 4, Results, but overall the recruitment rate was approximately 50% of the target rate.

Stock management and storage

Investigational Medicinal Product prescription and administration

Only individuals signed off on the delegation log (see Report Supplementary Material 4) were prescribed the IMP, using an IMP compliance assessment and accountability form (a dedicated prescription chart; see Figure 5) and the drug monograph for the IMP. Doses were oral/nasogastric and prescribed at 100 µg/kg. Once the IMP compliance assessment and accountability form was completed, the same prescription was written on the infant’s drug chart as ‘Poppi IMP Morphine or Placebo Pack ID XXXX’. The working weight was used to calculate the IMP prescription; an example dose calculation is presented in Figure 6.

FIGURE 5.

Investigational Medicinal Product compliance assessment and accountability form.

FIGURE 6.

Investigational Medicinal Product prescription calculation.

The dose was also double checked by qualified clinical staff and training was provided to include discarding excess solution prior to administration to the infant. Members of staff prescribing, preparing and administering the IMP were listed on delegation log (see Report Supplementary Material 4) with their designated duties. Any qualified member of the neonatal nursing staff was able to check the IMP dose; staff did not need to be listed on the delegation log (see Report Supplementary Material 4) because checking doses forms part of standard neonatal care. The IMP was administered after the first set of ROP eye drops; this was approximately 1 hour before the heel lance and ROP screen, allowing sufficient time for the morphine to be absorbed. The IMP could be administered by orogastric tube, nasogastric tube or directly into the mouth. The IMP was never added to a bottle feed in case the infant did not complete the feed. Following administration, the second part of the IMP compliance assessment and accountability form and the infant’s drug chart were signed to indicate administration (Figure 7). The form was returned to the co-ordinating centre each time the IMP was administered and checked by the trial manager to ensure that accountability had been demonstrated. Any errors identified regarding drug administration would have been reported as incidents and assessed, before being reviewed by the PMG. The trial sponsor also reviewed drug accountability records during site monitoring.

FIGURE 7.

The IMP compliance assessment and accountability form.

Accountability and destruction

Once the IMP had been administered, any IMP remaining in the vial was disposed of as per trust procedure (excess controlled drugs were disposed of in a sharps bin along with the bottle, syringe and packaging).

Any quarantined packs, packs that could not be allocated (e.g. for participants who had withdrawn after randomisation and pack allocation, but before IMP administration) and expired packs were returned to the pharmacy for destruction. A stock transfer form was completed documenting the transfer of IMP packs from the neonatal unit to the pharmacy, and the accountability log (see Report Supplementary Material 1) was updated to reflect that these packs had been returned to pharmacy. Stock transfer forms and updated accountability logs (see Report Supplementary Material 1) were e-mailed to the co-ordinating centre for storage in the TMF. In the case of expired or quarantined packs, the head of trials programming updated the online stock management system to prevent these packs from being allocated on randomisation.

The pharmacy contacted the trial manager to request authorisation for the destruction of transferred IMP packs. Following authorisation and destruction, the pharmacy team e-mailed the co-ordinating centre a copy of the trust accountability and destruction log to confirm that all relevant IMP packs had been destroyed. The trial manager marked each IMP pack on the online stock management system as ‘destroyed’.

Recording techniques

Approximately 1 hour before the heel lance (and after randomisation), the IMP was administered. The research team then set up the EEG, EMG and video monitoring. When the drug reached maximum efficacy (i.e. approximately 1 hour post administration) a heel lance control stimulus and clinical heel lance were performed. During each procedure noxious-evoked brain activity (as measured using EEG), reflex withdrawal activity (as measured using EMG on the biceps femoris of each leg), physiological activity (including oxygen saturation and heart rate) and facial expression change (using video monitoring) were recorded.

After the heel lance was complete, routine ROP screening was carried out by an ophthalmologist or an appropriately trained ophthalmology trainee. The time taken for the screening to be performed was recorded. Pulse oximetry, respiratory monitoring and video monitoring were recorded during and immediately after the ROP screen.

Once the ROP screen was complete, pulse oximetry, oxygen requirement and blood pressure continued to be monitored and recorded by the data-logging equipment for 24 hours after the start of the clinical intervention. Clinical stability was assessed throughout the 48-hour trial period. These measures were calculated from pulse oximetry recordings and requirement for respiratory support. Pulse oximetry data were monitored and downloaded to the data-logging equipment for 24 hours before and 24 hours after the start of the clinical intervention. Throughout the 48-hour trial period, blood pressure was monitored every 6 hours and changes in respiratory support (including type of respiratory support modality and oxygen requirement) were recorded. Any identification of abnormal clinical findings would have been reported to the clinical team. The recording measures used in the trial are summarised in Table 1.

Outcome measures

Clinical procedures

Administrative database

For the collection and management of non-clinical data, a web-based trial administration database application (TADA) was used (developed in-house at the co-ordinating centre). TADA allowed the study team to manage all non-clinical data relating to the trial, such as participant contact details and site staff details, and helped co-ordinate the collection of clinical data collection forms (including chasing missing forms, helping co-ordinate the management and resolution of clinical data queries, and generating reports for PMG meetings and other trial overview metrics).

OpenClinica

Power calculation

The primary objective was to determine whether or not morphine administration reduced clinical pain scores (as measured using the PIPP-R) during the 30-second period after ROP screening, in comparison with placebo. The co-primary objective was to determine whether or not morphine administration reduced noxious-evoked brain activity in response to a heel lance, in comparison to placebo. Although the trial closed early (see Trial closure), the trial was designed to be powered to show meaningful differences in the co-primary outcomes.

A two-point reduction in PIPP scores is considered to be a clinically meaningful reduction in pain,31 and scores below seven are considered to reflect minimal pain. 14 Three studies have used PIPP scores to assess analgesic efficacy following ROP screening. 32–34 Using the most conservative data, where the mean PIPP score post ROP screening was 8.3 [standard deviation (SD) 3.5], it was calculated that the trial would require 66 infants per group to observe a two-point reduction in PIPP-R scores, with a power of 90% at a two-sided significance level of 0.05.

The research team also considered the sample size required to observe a significant difference between the groups in the other co-primary outcome measure (noxious-evoked brain activity evoked by a heel lance). In previous analyses,35 the noxious-evoked brain activity evoked by a heel lance had a mean magnitude of 0.2 (SD 0.14). In the Poppi trial, the magnitude of the noxious-evoked brain activity was calculated and compared between the two groups. Studies in adults with chronic pain indicate that morphine treatment attenuates the amplitude of laser-evoked potentials by 33.1%. 36 Furthermore, adults administered tramadol (an alternative opioid treatment) showed up to a 50% reduction in the amplitude of laser-evoked potentials. 2 This reduction in nociceptive brain activity is coupled with the verbal report that tramadol is providing effective pain relief. Tramadol is widely used as an analgesic in adults and trials have shown that it effectively reduces neuropathic pain. 37,38 In the Poppi trial, the investigators, therefore, assumed that a 40% reduction in the magnitude of the noxious-evoked brain activity represented a clinically meaningful reduction in brain activity. A 40% reduction would lead to the noxious components having a post-treatment magnitude of approximately 0.12 (SD 0.14), down from a mean of 0.20. This would also require 66 infants per group for a power of 90%, at a two-sided significance level of 5%.

Allowing for a loss to follow-up rate of approximately 10% (e.g. as a result of technical difficulties during physiological monitoring), a total of 148 participants were required. Moreover, the research team anticipated that the proportion of twins eligible for the trial would be approximately 25% based on past studies [e.g. the National Perinatal Epidemiology Unit (NPEU)-run trials BOOST-II UK (a study investigating which oxygen saturation range should be maintained in very premature infants), I2S2 (a RCT of iodine supplementation in extremely preterm infants with follow-up at 2 years), PiPS (early administration to preterm infants of the probiotic Bifidobacterium breve strain BBG to prevent infection and necrotising enterocolitis) and SIFT (a multicentre RCT of two speeds of daily increment of milk feeding in very preterm or very low-weight infants)]. There is evidence that the correlation in pain outcomes between twins is 0.5. 39 Hence, the effect of clustering was calculated to be 1.06. This inflated the total sample size required (90% power, 5% two-sided significance level, 10% loss to follow-up and accounting for multiple births) from 148 to 156 participants (i.e. 78 per group). This was the recruitment target for the Poppi trial.

Significance levels

In the original statistical analysis plan (SAP; prepared for the trial to continue until reaching the target sample size), a p-value of 0.05 (two-sided 5% significance level) was planned to indicate statistical significance for the analysis of the co-primary outcome measures. The multiplicity issue would be addressed using Hochberg’s procedures for multiple testing to control the family-wise error rate. 40 Therefore, the research team planned to reject the null hypothesis for both outcomes if the p-value was < 0.05 for both outcomes. If the p-value was > 0.05 for one of the outcomes, then the other outcome would be tested at the 2.5% significance level. This method was less stringent than the Bonferroni adjustment, while preserving the original power of the study.

Comparisons of all other outcomes would be reported in full for completeness and transparency. For all other analyses a p-value of 0.01 (two-sided 1% significance level) would be used to indicate statistical significance, in order to take account of the number of comparisons. Two-sided statistical tests and corresponding p-values would be presented throughout; however, for the purposes of interpretation of results, confidence intervals (CIs) would dominate, rather than p-values.

Missing data

There were minimal numbers of missing data in this trial. Where this was the case, in some instances it may have been a result of equipment failure or artefacts within the EEG recording. Missing data occurred at random and so the collected data were representative of the population. Prospectively during trial design, the sample size was inflated by 10% to account for missing data.

Revisions to the statistical analysis plan

Prior to database lock and analysis of the results, the study team further developed the SAP and discussed each analysis in detail to ensure that the data were presented accurately and appropriately. Although the SAP would usually be completed before recruitment closure, as a result of the early suspension of the trial this document was still in draft form. Furthermore, the draft SAP required substantial revisions to accommodate the lower sample size (i.e. 31 participants from a target of 156). The following deviations from the original planned analyses described in the protocol were considered necessary:

• Significance levels – 95% CIs should be reported for all comparisons of outcomes, instead of reporting 99% CIs for secondary outcomes as specified in the protocol.

• Models should not adjust for minimisation factors because of the small sample size.

• Clinical stability analysis should not use Poisson or linear regression for count outcomes (i.e. the number of episodes of bradycardia, tachycardia, desaturation and apnoea). Instead, the Wilcoxon rank-sum test was performed on the standardised difference in the number of episodes of bradycardia, tachycardia and desaturation before and after the clinical intervention. The median differences between the groups with CIs were calculated for these standardised differences.

• Increased respiratory support was to be analysed using crude risk ratios, instead of logistic regression.

• Episodes of apnoea was treated as a binary composite of new-onset apnoeas or an increase in the number of apnoeas, analysed using crude risk ratios.

• The safety analysis did not use logistic regression, but instead calculated risk differences with CIs.

• Additional exploratory analyses were added (see Exploratory analyses).

The trial statisticians checked the data for inconsistencies and updated the dummy tables according to the revised SAP. It was ensured that the SAP provided a comprehensive report of the planned analyses and the document was reviewed thoroughly by the investigators and oversight committees. A joint TSC and DMC meeting was held on 6 April 2018 for the oversight committees to review the updated SAP (version 1.0; 6 April 2018). Minor edits to the document were suggested, but overall the attendees approved the document, which was signed off by the chief investigator, principal investigator, senior statistician and TSC chairperson.

However, following sign-off of SAP version 1.0 four errors/omissions were identified:

1. Stopping boundaries had been stated incorrectly (this matched an error identified in the protocol – the stopping boundaries are stated correctly in two other places in the protocol).

2. A typographical error occurred in the criteria for participants randomised in error – ‘born ≤ 32 weeks’ gestation or birthweight ≥ 1501 g’ should be stated instead of ‘born ≥ 32 weeks’ gestation and birthweight ≥ 1501 g’.

3. Postnatal age should have been included in the characteristics at the time of study. This was discussed during trial design and agreed that it could be calculated from the eCRF data. The PMG agreed that this should be incorporated into the new version of the SAP.

The PMG agreed that errors 1 and 2 were typographical errors and error 3 was an oversight, all of which should be corrected in SAP version 2.0. The co-ordinating centre quality assurance (QA) associate agreed that this was an appropriate course of action. The errors corrected were clearly explained in the ‘document history’, and stated to have occurred after unblinding of the investigators. The changes were incorporated into SAP version 2.0 (6 June 2018), which was re-signed by the chief investigator, principal investigator, senior statistician and TSC chairperson.

Primary and co-primary outcome measures analysis

Secondary outcome measures analysis

Planned interim analyses

An interim statistical analysis of the primary and secondary outcomes was planned when 50% of trial participants (78 infants) had completed data collection. In the light of the interim data analysis, and other evidence from relevant studies (including updated overviews of relevant RCTs), the role of the DMC was to inform the TSC if in its view there was proof beyond reasonable doubt that the data indicate that any part of the protocol under investigation was either clearly indicated or contraindicated. Guidelines for early cessation owing to safety concerns were agreed with the DMC and documented. Unless modification or cessation of the trial was recommended by the DMC, the TSC, investigators, collaborators and administrative staff (except those who supply the confidential information) would remain ignorant of the results of the safety analyses.

Trial stopping rules

Drug safety was assessed by measuring the number of occurrences of apnoea that required intervention using Neopuff or bag-and-mask ventilation, or hypotension that required treatment with inotropes, in the 24-hour period after drug administration. The DMC planned to review trial safety outcomes after every 25 participants had been randomised and safety data had been collected (i.e. n = 25, 50, 75, 100 and 125). In addition, the chief investigator (or suitably trained delegate) was notified of every such occurrence.

A formal sequential safety procedure was applied and presented to the DMC for occurrences of apnoea that required intervention using Neopuff or bag-and-mask ventilation. A stopping boundary using group sequential methods with a boundary agreed by the DMC and specified in the DMC charter was employed.

Specifically, a one-sided gamma spending function, with gamma equal to 4.5, was used. This is a type of α-spending function, that is a stopping boundary where the type I error rate is spread across a number of analyses. The boundary is a function of the amount of information collected (in this case, the number of infants with safety outcome data available). A type I error rate of 0.2 was used, which was estimated to result in power of 0.79. This boundary was based on the assumption of a control group event rate of 7% and a difference between the group event rates of 12%.

At the first DMC safety review of 25 infants and the subsequent requested review of 31 infants, a graph showing evidence of the relative safety of the treatments was provided as a guide, in addition to trial data summarised by trial arm, clinical summaries for infants who had experienced safety events and the completed CRFs for all infants. The DMC charter stated that, if the prespecified stopping boundary was crossed, then the DMC should consider terminating the trial, taking into account other considerations (e.g. implications on future infants/clinical practice and follow-up, how convincing the evidence was). It is important to note that the stopping boundary was intended as an aid to the decision to stop the trial, and not as a definitive stopping rule. There was no stopping rule or plan to stop early as a result of benefits from the treatment.

Adherence to the intervention

Adherence to the intervention was reported in a process outcomes table. The median, IQR and range of the volume of IMP given, and the number and percentage of infants receiving more than the correct dose for their weight (i.e. 100 µg/kg), plus a 5% tolerance to account for measurement error, were reported.

Timing of the study treatment in relation to the clinical intervention was also reported in this table. The time between study treatment administration and the clinical intervention was summarised by trial arm using median and IQR, and ranges.

Analysis populations

Descriptive analyses

Additional analyses

Additional objectives not prespecified in the protocol were added, in the light of the early suspension of the trial recruitment, as detailed in the following sections.

Study documentation

The TMF was created at the beginning of the trial by the trial manager. This was organised to allow efficient maintenance and review of trial-related documentation using a template contents document from the co-ordinating centre. The trial manager, trial sponsor, co-ordinating centre, QA team, MHRA (should an inspection have taken place) and the archivist had access to the TMF; maintenance of the TMF was the responsibility of the trial manager. All documentation stored separately from the TMF had a signed and dated file note within the relevant section of the TMF indicating its location. All essential documents, trial-related documents and important correspondence were stored in the TMF, which was updated regularly throughout the trial. Previous versions of documents were retained in the TMF, but clearly labelled as superseded.

The trial manager provided the site team with an ISF, and the pharmacy contact with a pharmacy folder, containing all relevant documentation for the trial. Throughout the trial, updated documentation was sent to the research team and the pharmacy contact to keep the ISF and pharmacy folder up to date. Details of the protocol, PIL, consent form and other parent-facing materials are presented in Appendix 4. Final versions of the documents approved by the regulatory bodies are listed in Table 2.

Details of regulatory reporting and communication are presented in Appendix 5.

Training and delegation

Monitoring and oversight

Information about the site initiation and close-out visits from the trial manager is presented in Appendix 6. Details about the sponsor-led monitoring of the trial are provided in Appendix 7 and PMG-led monitoring in Appendix 8.

Oversight of the trial was maintained by the DMC and TSC, the members of which were approved by the trial funder.

Safety reporting and unblinding

Throughout the trial any staff could report safety events at any time. The safety reporting process is detailed in Appendix 10. There were two reported SAEs in the trial (see Chapter 4, Serious adverse events) and no suspected unexpected serious adverse reactions (SUSARs).

Unblinding

Treatment allocation for any participant could be unblinded by the clinical team using the randomisation website. The requirements were that the treating clinician felt that it was a genuine emergency and that knowledge of the treatment allocation (i.e. morphine or placebo) was necessary to guide the appropriate clinical management of the infant. The document box, provided at the site initiation visit (see Appendix 6, Site initiation), included a sealed envelope containing the randomisation website address, login details and password, and a single-use access code. These details were entered onto the randomisation website, alongside the study number, details about who was carrying out unblinding and information regarding why unblinding was necessary. The system would then reveal the arm allocation of the infant. The co-ordinating centre would be automatically informed of any cases of unblinding. The unblinding process is detailed in Figure 8.

FIGURE 8.

Unblinding process.

Withdrawals

Any staff member could withdraw a participant from the trial if parents wished to withdraw or if it was deemed necessary for clinical care. Parents did not need to give a reason for withdrawing a participant. When withdrawing a participant, staff completed a paper withdrawal form to record the time, date, reason for withdrawal, and permissions from the parents to use data already collected. These data were then entered onto OpenClinica. The time and date of withdrawal were then documented in the infant’s medical notes, together with any other necessary information.

There was one withdrawal from the trial. At the time of study set-up, the clinical team informed the parents that the infant could be discharged on the day of the ROP screen. The parents, therefore, chose to withdraw their infant before monitoring began, in order to prevent delay to discharge (if they had participated, the infant would have been monitored for a further 24 hours after the ROP screen, delaying discharge by 1 day). No data were collected for this infant and, therefore, the infant was excluded from analyses.

Oversight

The DMC met on 4 December 2017 to review the safety outcomes of the first 25 participants. The safety analysis indicated that the stopping boundary had been crossed for the first 25 infants. In order to make a fully informed decision, the independent DMC arranged to meet again on 7 December 2017 to review the data for the co-primary outcomes and re-review the safety data with additional data from the six participants who had been recruited (one of whom was not studied). The stopping boundary for the safety outcome was still crossed with inclusion of the additional participant data, and the efficacy outcomes did not suggest that there was pain relief. Therefore, the DMC recommended that the TSC have access to the unblinded safety report and consider suspending the trial.

The independent TSC members were granted access to the unblinded safety report by the trial statistician and the recommendation of trial suspension was discussed at the TSC meeting on 13 December 2017. At this meeting, the TSC reviewed the unblinded safety report and agreed with the DMC’s view that the stopping boundary had been crossed with little evidence of analgesic efficacy. The TSC recommended suspension rather than cessation of the trial, as it considered that this would perpetuate a serious uncertainty in clinical practice and could lead to the continued infliction of a painful and destabilising procedure on preterm infants without evidenced-based analgesia. It considered that a revised trial design had the potential to resolve a major uncertainty in newborn care, which could provide robust evidence to guide pain management and improve clinical outcomes for infants. Although the TSC considered that the trial was conducted to an extremely high standard with strong methodological rigour, it could not recommend that the trial continue in its current form. The decision was primarily based on concerns about drug safety and a lack of evidence that morphine (at the current dose) provided adequate relief for procedural pain. The TSC recommended that significant changes be made to the trial design and that investigators review the unblinded data in order to guide the development of a revised protocol. These changes would provide insight into (1) the safety issues that resulted in the stopping boundary being crossed; (2) the lack of analgesic efficacy identified from the primary outcome data; and (3) the effect of both morphine and the ROP screening procedure on infant physiology, which could guide future clinical practice. As the investigators would be unblinded, data collected could not be included in the proposed future data set and, instead, should be reported as a standalone study. The trial sponsor and funder were in agreement with these recommendations.

The randomisation system was therefore disabled and recruitment was suspended on 13 December 2017. A substantial amendment for temporary suspension was submitted and approved by the REC and regulatory bodies. Key stakeholders were informed of the decision (including the trust, drug manufacturers and relevant research staff). The trial team met in January 2018 to develop a plan. In mid-February the team provided the trial funder with a detailed breakdown of the required actions before review of the data, protocol development and publication of results. A progress report was submitted to the trial funder that further explained the steps required before database lock and the complexities of the changes required for the SAP (see Revisions to the statistical analysis plan). The team planned to submit an options document to the funder in early May to outline the possible protocol changes following review of the data.

In March, the team received a letter from the funder stating that funding would be withdrawn, and following review of the TSC chairperson’s letter and documentation, submitted in February, it had decided that the suggested trial revisions required submission of a new trial proposal, rather than protocol amendments. The funder also reiterated that the pilot data should be used to guide developments for the new trial proposal. Therefore, the central monitoring team stopped the trial on 15 March 2018 and the data from the 31 participants were analysed.

Regulatory

Following the TSC’s decision on 13 December 2017, recruitment to the trial was suspended immediately and a substantial amendment submitted to the Health Research Authority (HRA), REC and MHRA informing them of the temporary halt. The investigators planned to review the data once the database was locked and, consequently, devise a plan for the development of the trial in the light of the TSC’s recommendations.

Publicity

Following recommendations from the DMC following the safety analysis meeting on 7 December 2017, and those from the TSC on 13 December 2017, the trial was no longer publicised. The Poppi trial website was updated to inform viewers that recruitment to the trial had been paused while the trial data were reviewed. Key stakeholders were also informed of the trial’s current status.

Completion of data entry

Following the decision to suspend the trial, the site team then entered any outstanding CRFs and missing data onto OpenClinica and resolved discrepancy notes in which queries in the data had been raised. The protocol stated that the inter-rater and intra-rater reliability would be assessed for the co-primary outcome (i.e. PIPP-R score), so duplicate CRFs were created for this outcome and validated by the programming team. The PIPP-R score was then recalculated and re-entered by the site team. The programming team also needed to create new CRFs (and validate these) for a number of data items that were recorded during the trial but not transposed onto CRFs (e.g. noxious brain activity at baseline). Given the sudden suspension of recruitment, the PMG agreed that these additional data should be reported in the publication of the results to provide context. The site team then provided the programming team with the additional data, and the programming team uploaded these to the CRFs and completed validation. A sample of the original data within OpenClinica was also cross-checked with the paper CRFs for validation purposes.

Database lock

The senior trials programmer provisionally locked the database in order for any checks to be carried out, by removing permissions to edit the database from all personnel. These checks were signed off by the trial manager, the head of trials programming, the senior statistician and the trial statistician, and included:

• A sample of 10% of all participants, which was generated randomly, ensuring that at least 10% of each CRF was included. All data on the OpenClinica eCRFs, for each of these participants, were compared with the data in the Stata data set to be used for analysis. In the case of the SAE eCRF, the data in the Stata data set were compared with the data on the paper CRFs. All data on the paper SAE chief investigator assessment CRFs were compared with the data in the Stata data set to be used for analysis.

• Checking that all forms received at the co-ordinating centre had been entered onto the corresponding eCRF and marked as complete in OpenClinica.

• Checking that all discrepancy notes in OpenClinica were closed.

• Checking that the reports generated from the administrative database for consent form issues, CRF signature and delegation log checks.

• Validation of the process for importing extra details from an Microsoft Excel^® (Microsoft Corporation, Redmond, WA, USA) spreadsheet into OpenClinica.

• Sign-off of the SAP.

Authorisation to lock the database was provided by the chief investigator, the principal investigator, the director of the co-ordinating centre, the senior statistician, the head of trials programming and the trial manager, and the senior trials programmer locked the database. The data set was exported to the trial statistician and the site research team for analysis (the principal investigator signed a memorandum of understanding for the transfer of the data). The trial statistician used this data set to create the final report detailing the results of the trial.

Two differences between the clinical stability source data and OpenClinica data set were discovered on 14 May 2018 by the site team. The error was identified when the raw data were analysed in MATLAB (MathWorks, Natick, MA, USA) rather than the exported data set from OpenClinica, and the outputs differed from the final report (provided by the trial statistician using the OpenClinica data set). Although correction of these errors would not significantly alter the results, the values in the final report at that time were inaccurate.

The PMG met on 17 May 2018 and agreed that the database should be unlocked, these values corrected, then the database relocked. The PMG was attended by the co-ordinating centre QA associate, who provided QA input into the decisions made by the group. The PMG decided that, beforehand, a complete comparison should be made of the clinical stability raw data from MATLAB and data entered onto OpenClinica.

A further meeting of the PMG was held on 22 May 2018. The PMG agreed that the following should be checked against their source data and were selected because of their importance in the final results:

• co-primary outcome

  • magnitude of noxious-evoked specific brain activity evoked by heel lance (source data from MATLAB – value for each infant)

• important secondary outcomes

  • increase in respiratory support before and after the clinical intervention (source data on cot-side log – yes/no for each infant)

• safety outcomes

  • at least one incident of apnoea that required an intervention using Neopuff or bag-and-mask ventilation in the 24-hour period following drug administration (source data on the cot-side log – yes/no for each infant)

  • at least one incident of hypotension that required treatment with inotropes in the 24-hour period following drug administration (source data on cot-side log – yes/no for each infant).

The PIPP-R scores (at ROP screening, heel lance and control heel lance) were already validated by the trial statistician and postdoctoral researcher on 27 March 2018, when the source data were checked. These data, therefore, did not need to be re-checked.

These further outcome data were entered by the postdoctoral researcher and research nurse onto a separate spreadsheet on 24 May 2018. The head of trials programming ran the comparison with the OpenClinica data set and confirmed that there were no further inconsistencies.

The database unlock documentation was then prepared by the co-ordinating centre, documenting any required changes to the database. Following approval from all individuals named on the unlock form (co-ordinating centre director, chief investigator and senior statistician), the database was unlocked on 4 June 2018 by the senior trials programmer, and the trial manager and postdoctoral researcher together entered the correct values. The co-ordinating centre programming team then checked the audit log and verified that only the agreed changes had been made to the data set. The relock was then authorised by the senior trials programmer, head of trials programming, senior statistician, chief investigator, principal investigator, co-ordinating centre director and trial manager. The database was relocked on 6 June 2018 by the senior trials programmer, and the new data set exported for the trial statistician to revise the relevant tables and update the final report. The new data set was transferred to the principal investigator and postdoctoral researcher from the site team to run the additional analyses.

Analysis

The final analysis was completed by the trial statistician according to the SAP, version 2.0 and reviewed by the senior statistician. One deviation was made from the SAP: where risk ratios were not estimable because of no events in one trial arm, risk differences were calculated. This applied to the outcome ‘required an increase in respiratory support’ at both 6 and 24 hours following the clinical intervention. Additional post-hoc exploratory analyses were performed by the on-site Poppi trial team (see Exploratory analyses for more detail).

The co-primary outcomes were validated by an independent statistician at the co-ordinating centre prior to the analysis being finalised.

Trial design

Patient and public involvement input was essential in designing the trial and was considered critical to its success. To ensure that the parents’ views and opinions were considered from the beginning of the trial, the SSNAP parent help team leader was an applicant on the grant application. In collaboration with SSNAP, the research team organised a workshop in which parents of premature infants discussed the proposed clinical trial with the research team. This workshop provided the parents with an opportunity to discuss their ideas about the trial and explore the ethical, practical and emotional issues associated with the research. Furthermore, parents were actively involved in the application for ethical and regulatory approvals. The parents suggested amendments to the research design based on their experiences and edited the proposed PIL (see Report Supplementary Material 6). Parents advised on timing and manner of approach to potential families, and suggested ideas regarding dissemination of information about the trial. Importantly, parents were able to advise on how their expectations about pain management differed from actual clinical practice, and made suggestions about how the team could sensitively approach this with families.

Trial conduct

Parents of prematurely born infants are often unaware and shocked when they realise their infants will experience painful procedures without the provision of pain relief. Raising awareness of this and explaining why so few pain-relieving drugs are available for infants highlighted the importance of the trial and encouraged parent participation. This is important in neonatal research, as families can be in the neonatal unit for extended periods and provide a support network for each other. The SSNAP parent help team leader was the primary support contact for parents and clinical staff, and provided key information to parents about the trial. This endorsement by SSNAP was critical to ensure that parents felt that the trial had been designed and was conducted with the care of their infant, and their own concerns, in mind.

The parent help team leader identified a volunteer from SSNAP to act as a PPI representative on the trial TSC. All SSNAP volunteers are individuals who have had a child or grandchild cared for in the neonatal unit of the John Radcliffe Hospital and, therefore, have the experience to both support and promote the charity. The PPI representative provided invaluable insight into the parents’ perspective and considerations for parents’ views at each stage of the trial, by attending and inputting into every TSC meeting that was held.

The parent help team leader also identified volunteers from SSNAP to raise awareness of the trial on the neonatal unit. The parent help team leader co-ordinated an information session where the research team met with the volunteers and gave a presentation about pain in premature infants and the trial. This provided the volunteers with the opportunity to ask questions about the trial and consider how they might be able to promote the trial. These volunteers acted as a further source of information for parents who were considering participating. The parent help team leader and the SSNAP volunteers provided information about the trial at SSNAP coffee mornings, where parents of premature infants meet for support. The research team attended these meetings, where possible, in order to build further relationships with the parents and answer any questions they may have had about their infant and pain management.

Partway through the trial, the Poppi trial team adapted the existing trust ROP screening PIL to improve and clarify the information provided to parents about the procedure (see Chapter 1, Retinopathy of prematurity screening). The SSNAP team reviewed the document to ensure that it was parent-friendly and provided all relevant information in a suitable format. SSNAP approved the wording and endorsed the document.

Trial dissemination

Both the SSNAP parent help team leader and the TSC PPI representative attended the TSC/DMC/collaborator meeting to discuss the draft results manuscript for The Lancet [see Results (The Lancet)]. They also attended the family event held on 11 July 2018 to thank the families who participated, and to feedback the results and explain the early trial closure (see Participants). Their involvement in this event was critical to its success, as they had established relationships with many of the families. They were able to provide further support and information from SSNAP, and encouraged families to complete a feedback survey about their experiences of the trial (see Family feedback).

Summary protocol publication (The Lancet)

The summary protocol for the trial was published in The Lancet. 43 This was accepted for publication on 22 January 2016.

Protocol publication

The protocol was published with Wellcome Open Research on 15 November 2016 and the revised protocol was published on 26 January 2017. 28

Results (The Lancet)

A joint TSC, DMC and collaborator meeting was held on 11 June 2018 to review the draft The Lancet manuscript detailing the results of the trial. All invited attendees had the opportunity to review and comment on the draft manuscript electronically prior to the meeting. The attendees largely approved the manuscript and the underlying messages, and provided suggested minor edits throughout the text. The revised manuscript was published in The Lancet in November 2018. 42

Commentary (The Lancet Child & Adolescent Health)

A commentary discussing the ethical considerations within the trial and the regulatory and oversight procedures in place was co-published in The Lancet Child & Adolescent Health in November 2018. 44

Other relevant publications

Members of the Poppi trial team published a paper in Science Translational Medicine22 that described and validated the methods that were used to quantify the primary outcome measure (i.e. noxious-evoked brain activity) in the trial.

Participants

The parents/guardians of the participants of the trial were invited by telephone and by post to a family event on 11 July 2018. The families of 10 participants attended, including parents, grandparents, siblings and aunts. The research team used the event to thank the participants and their families for their contribution and support for the trial. An animation explaining the challenges of assessing pain in infants, and exploring the purpose of the research, was shown to the families. The principal investigator thanked the families, disclosed the results of the analysis and explained the sequence of events leading up to trial closure. The research team were keen that families clearly understood the decision to stop the trial and had the opportunity to discuss any queries or requests for further information with the team in person. Parents of five participants also requested (in writing) to know the treatment allocation of their infant. The event was extremely successful and greatly supported by representatives from SSNAP. A letter was also sent to the parents/guardians of all participants in the post in November 2018, summarising the trial results and explaining that these would soon be published.

Staff and stakeholders

An end-of-trial event was held on 12 September 2018 to thank all the Poppi trial staff and stakeholders for their efforts and input into the trial.

Website

A trial website was maintained on the co-ordinating centre website (www.npeu.ox.ac.uk/poppi; accessed 15 June 2019). This resource provided a link to the randomisation and stock management website, and also provided a source of up-to-date information about the trial. The home page provided a short summary of the trial, a recruitment count, videos showing animations about infant pain and the group’s research, and general information about the trial (including eligibility criteria and trial registration details). The home page also included acknowledgement of the funding bodies (i.e. the NIHR EME programme and the Wellcome Trust), and endorsement from Bliss (a charity for babies born premature or sick) and SSNAP. The Poppi trial website provided contact details for core members of the team, information and links about the team’s involvement with the media, information for parents (including a flow diagram of study processes and the PIL), the protocol, relevant publications and a recruitment summary by month. News items were also added regularly, which included updates on recruitment, media coverage and publications.

Media

The publication of the Science Translational Medicine paper (see Other relevant publications) generated a large amount of interest from the media. Articles about the trial, and the impact of this work on the measurement of pain in premature infants were widely published and included articles in the New Scientist (www.newscientist.com/article/2129744-electrode-can-tell-you-if-a-baby-is-really-experiencing-pain/; accessed 15 June 2019), The Guardian and IEEE Spectrum (http://spectrum.ieee.org/the-human-os/biomedical/diagnostics/brain-activity-detector-helps-study-pain-relief-for-babies; accessed 15 June 2019), as well as many other news outlets. Professor Slater took part in an interview about the work and this is available on the International Business Times UK website (www.ibtimes.co.uk/your--pain-brain-scan-can-detect-physical-suffering-infants-1619859; accessed 15 June 2019). Professor Slater and Dr Moultrie also featured in a BBC Radio 4 programme entitled ‘From Agony to Analgesia’, which was broadcast on 16 August 2017. This programme focused on specific research areas in the field of pain, during which Professor Slater and Dr Moultrie discussed novel methods to measure pain in premature infants and the objectives of the trial.

In November 2017, the Poppi trial team joined the #startedinoxford initiative with a video introducing its work to explore how infants feel pain. The #startedinoxford initiative has a website (http://innovation.ox.ac.uk/innovation-news/startedinoxford/; accessed 15 June 2019 ) and YouTube (YouTube, LLC, San Bruno, CA, USA) channel (https://www.youtube.com/results?searchquery=%23StartedinOxford; accessed 15 June 2019) publicising the innovative research being conducted at the University of Oxford.

The clinical doctoral fellow from the Poppi trial team (Dr Fiona Moultrie) collaborated with a professional animator (Miss Charlotte Moultrie) to create an animation explaining the importance of understanding pain in infants, the difficulties of assessment, and new imaging and recording techniques that are being used to understand infant brain development. This animation also discussed the aims of the trial, and the video appealed to parents of infants on the neonatal unit, adults and school children, as well as scientists, doctors and nurses. The video was available on the Poppi trial website, shown to parents at the family event, and exhibited at various public engagement events and public venues including the Oxford Natural History Museum, the Ashmolean Museum and the Cheltenham Science Festival. The work was also recently recognised with an honourable mention at the OxTalent 2018 awards in the digital media category.

Public engagement events

The Poppi trial team presented information about the trial alongside interactive games and displays about neonatal brain development at the ‘FrightFriday’ public engagement event on 25 November 2016 at the Ashmolean Museum in Oxford, which was supported by the Wellcome Trust and the national Festival Finale for the Being Human Festival. The event comprised a late-night opening of the museum to explore the art and science of hope and fear, with performances, workshops and stalls run by researchers across the University of Oxford. The Poppi trial team had a large display, which included Poppi trial posters, an interactive three-dimensional (3D) brain model to demonstrate the similarities between the experience of pain in adults and infants, 3D printed brains and a video animation (https://www.youtube.com/results?search_query=%23startedinoxford; accessed 25 July 2019). The event was extremely successful, and the team’s display was featured in the University of Oxford’s Medical Sciences newsletter (December 2016). The SSNAP parent help team leader (also a grant applicant) was also involved with the Poppi trial and paediatrics department activities at this event.

Conferences

Dr Moultrie gave a presentation entitled ‘Neonatal Pain: Improving our Understanding of the Assessment and Treatment of Pain in Infants’ at the Congenital Cardiac Anaesthesia Network Meeting on 13 May 2016 in Southampton, UK. During the presentation, Dr Moultrie described the definition of pain, methods to measure pain in the newborn, how to investigate the effects of early pain and current research to test analgesics in this population, specifically the Poppi trial.

Dr Hartley also gave a presentation entitled ‘Using Noxious-Evoked Brain Activity to Measure Analgesic Efficacy in Infants’ at the London Pain Consortium Research Update Meeting on 12 May 2016. Dr Hartley provided information about infant pain, surrogate measures of infant pain, the balance of reflex withdrawal and brain activity in early development in response to noxious stimuli, the potential use of brain activity to explore analgesic efficacy and the plans for the Poppi trial.

Dr Moultrie presented a talk entitled ‘Determining Infant Pain Indicators to Support New Approaches to Pain Assessment and Effective Pain Management Strategies’ in a workshop at the International Symposium for Paediatric Pain, in Malaysia, in July 2017. Dr Moultrie explained how quantitative neurophysiological techniques can be used to evaluate nociceptive responses in infants and how these measures can be used as sensitive objective end points in clinical trials to assess analgesic efficacy. She outlined the Poppi trial protocol, which generated substantial interest and discussion. Following the early cessation of the trial and final analysis, Dr Moultrie also described the trial results at the Neonatal Society Autumn Meeting in November 2018 in London, UK, where she won the prize for best presentation.

Literature searches

The research team searched PubMed (National Centre for Biotechnology Information, Bethesda, MD, USA) and Google Scholar (Google Inc., Mountain View, CA, USA) at least fortnightly for any relevant literature using a range of search terms such as ‘infant’, ‘pain’, ‘analgesia’ and ‘morphine’. The research team are based within the Paediatric Neuroimaging Group at the University of Oxford; when any member identified relevant literature, this was circulated to the full research team.

Emergencies

Trial documentation was provided to the neonatal unit, including contact numbers of trial staff who should be contacted in an emergency. A clinical telephone cascade included four members of staff, who gave permission for their contact details to be distributed, and the trial enquiries cascade included three members of staff.

Complaints

All parents provided written informed consent for their infant taking part in the trial. If parents were anxious, members of the clinical and research team answered any questions and discussed concerns with them. If the parents had any complaints during the Poppi trial these would have been documented, any action taken registered in their medical notes, and an incident form completed. In the PIL (see Report Supplementary Material 6), parents were given the contact details of the trial manager, chief investigator and principal investigator if they wanted more information, alongside the contact information for the trial sponsor and the Patient Advice and Liaison Service should they have wished to report a complaint or concern. No complaints were received about the Poppi trial.

Network support

The trial was supported by the Thames Valley and South Midlands Clinical Research Network (CRN). This provided funding to support the full-time salary for a dedicated neonatal research nurse, working exclusively at the newborn care unit at the John Radcliffe Hospital, Oxford. The research team liaised with the neonatal research nurse throughout the set-up of the trial to ensure that clinical staff had up-to-date GCP training. The team continued to collaborate with the neonatal research nurse about establishing effective methods for training neonatal staff members about trial processes during set-up. Once recruitment started, the research team continued to work with the neonatal research nurse to ensure that there was staff training and protocol compliance.

Trial funder

The trial funder was regularly provided with trial updates using the NIHR Evaluation, Trials and Studies Coordinating Centre (NETSCC)’s online management information system. Recruitment updates were submitted on a monthly basis and biannual progress reports were prepared providing information on all aspects of research activity in the trial. Minutes from oversight committee meetings (with actions for the research team) were uploaded periodically, as well as outputs for all public engagement activities, publicity resources and publications. The trial funder was made aware of any trial amendments, and trial documents (both draft and approved versions) were similarly uploaded to the online management system. Partway through recruitment, the Poppi team also met with the EME programme director to discuss the progress and the future of the trial. The programme director enquired about the practicalities of the trial, including information about the live birth rate at the site, standard practice for ROP screening, assessment of trial outcomes, appropriateness of morphine and the proposed plans to increase recruitment. The principal investigator provided detailed information about the trial, trial processes and future research plans. The meeting concluded with all attendees in agreement regarding the importance of the research area and the critical need for continued research into the prevention of pain in infants.

The first serious adverse event

The infant experienced respiratory problems at the time of a bottle feed and profound apnoea approximately 6 hours after IMP administration. The participant became apnoeic and required bag-and-mask ventilation. It was unclear at the time if this was associated with the IMP. The decision was made to unblind this participant because of intermittently poor respiratory effort. The participant had received the active IMP (i.e. morphine sulphate). The infant was transferred to the HDU and started on high-flow oxygen treatment. The infant was given naloxone at 8 hours and at 9 hours post IMP. The SAE was considered ‘expected’ and the initial report was that it was ‘probably’ related to the IMP.

The infant’s oxygen saturation recording was reviewed after the incident. These data highlighted that in the 24 hours prior to IMP administration, the infant had frequent episodes of physiological instability at the time of feeds. Following the ROP screen, the infant did not have any episodes of desaturation (< 80%) for a period of almost 4 hours, until the bottle feed (which was started at approximately 5 hours 15 minutes post IMP administration). This suggested that the SAE was related to the bottle feed. However, on review of the impedance pneumogram and CRFs it was identified that the infant had four apnoeas between the time of IMP administration and the bottle feed. This suggested that the original assessment, that the SAE was ‘probably’ related to the IMP, was correct.

The second serious adverse event

During a feed (≈4 hours post IMP), the infant had profound desaturations and bradycardias, requiring brief application of facial oxygen and only half the feed volume was given. The infant continued to have recurrent episodes of desaturation and low-flow oxygen treatment was started 5.5 hours post IMP. The infant was transferred from the LDU to the HDU for a partial septic screen and intravenous antibiotics. The infant displayed signs of respiratory distress (grunting and increased work of breathing) and was started on high-flow oxygen at a rate of 4 l/minute. The infant stabilised and high-flow oxygen treatment was discontinued the following day. Chest radiography revealed right middle lobe pneumonia, potentially consistent with aspiration, and the C-reactive protein level rose to a maximum of 13.9 mg/l. The infant was treated with antibiotics for 7 days. The infant remained on the HDU and received a red cell transfusion for anaemia 2 days later. The infant continued to have feed-related desaturations/bradycardias and was commenced on omeprazole 1 week after the study.

The vital signs data were reviewed retrospectively. In the 24 hours pre IMP administration the infant had four apnoeic episodes (as identified on the impedance pneumogram), 12 episodes of desaturation and five bradycardias. In the 24 hours post IMP, the infant had four apnoeic episodes, 14 episodes of desaturation and eight bradycardias. The SAE was considered ‘possibly’ related to the IMP, as morphine could increase the likelihood of aspiration.

Temperature excursion

For the first batch of IMP, a lower limit of 15 °C was set on the TempTale for monitoring the IMP temperature in transit. On receipt at the pharmacy, the team observed that the data logger had alarmed during transit. On review of the data, the excursion was < 2 °C below the lower limit for a time period of less than 3 hours. The quality control team at Stockport Pharmaceuticals confirmed that this minor temperature excursion placed insignificant risk of product deterioration. Chemical deterioration (such as hydrolysis or oxidation) is a temperature-dependent process, which slows (rather than accelerates) with a lower temperature, so this was not a cause for concern in this instance. There was a theoretical possibility that some components may have become insoluble at lower temperatures, but all the components were freely soluble at the concentrations found in the product and would endure far greater temperature reductions before precipitating. It was identified that a lower temperature limit was not necessary for this IMP, as per standard practice. No lower temperature limit was applied to future batches of IMP.

Overlabelling

After the manufacture of IMP batch 1, further stability data were made available from Stockport Pharmaceuticals permitting a shelf life of 6 months. The QP at Stockport Pharmaceuticals confirmed that the stability data supported the extension of the expiry date and that another QP release certificate was not required. These stability data were filed in the product specification file held at Stockport Pharmaceuticals.

The extension conformed to the MHRA approval conditions, which state that ‘the initial shelf life of the drug product will not exceed twice the time period for which there are satisfactory stability data at 25 degrees/60% room humidity’. The remaining 29 IMP packs from batch 1 were therefore over-labelled to extend the IMP expiry date from 29 November 2016 to 3 March 2017. The expiry date was also updated on the co-ordinating centre pack management system (see Chapter 3, Electronic stock management).

Change in SyrSpend supplier

The active IMP comprised morphine sulphate in a cherry-flavoured oral suspending base (SyrSpend). The placebo was SyrSpend in isolation. Stockport Pharmaceuticals sourced SyrSpend from Fagron BV. In December 2016, Stockport Pharmaceuticals was alerted that Fagron BV had changed its supplier of SyrSpend to Tiofarma BV (Oud Beijerland, the Netherlands). In January 2016, Stockport Pharmaceuticals was made aware that the new SyrSpend supplier did not have an IMP licence and, therefore, considered that its product could not be used for future IMP batches. The IMP batch at the time was due to expire 6 March 2017 (IMP batch 1).

Stockport Pharmaceuticals had some remaining stock of SyrSpend from the previous licensed supplier, Terra Pharma (London, UK), which expired on 31 May 2017. A request for a new IMP batch was placed using this stock as a short-term solution (batch 2), but another IMP batch with a different SyrSpend source was required to be manufactured before batch 2 expired. The PMG discussed the following potential long-term solutions:

• Further stability data were expected at the end of March 2017 for the original IMP. This might have provided stability data for extending the expiry date of the recently requested IMP batch and allow a greater time period to identify another long-term solution.

• Contact Terra Pharma directly to request a personal batch of SyrSpend. Terra Pharma had not yet confirmed if this was feasible, but had suggested that a minimum order of 1000 bottles would be required. Potential costs of this were explored by Stockport Pharmaceuticals.

• Stockport Pharmaceuticals had a large supply of unflavoured SyrSpend which could be used (expiry March 2018). This would require an update to the IMP dossier and a substantial amendment to the MHRA. Additionally, as the recruitment period was intended to be 33 months, another longer-term solution would be required for the remaining recruitment period after this SyrSpend supply expired.

• Use a completely different suspending agent. This would require an update to the IMP dossier and a substantial amendment to the MHRA. However, many other suspending agents were not suitable for neonatal use because of the sugar content or other ingredients.

However, the MHRA confirmed that as the SyrSpend is classed as an ‘excipient’ there was no requirement for the new supplier to hold an IMP licence. A substantial amendment was therefore submitted in March 2017 to change the SyrSpend supplier to Tiofarma BV in the IMP dossier. This amendment was approved by the MHRA, REC and HRA in March/April 2017. IMP batches 3 and 4 were manufactured using SyrSpend from the new supplier.

Version 1

This was the original version of the protocol submitted for approvals. This version of the protocol was approved by the REC on 23 July 2015 and MHRA on 21 August 2015.

Version 2

Version 2 of the protocol incorporated the following updates:

• Change of measure used to assess clinical pain scores: change from PIPP to PIPP-R.

• Change to timing of primary outcome measure: change from PIPP score 1 minute after ROP screening to PIPP-R score 30 seconds after ROP screening.

• Definition of clinical intervention: where applicable, the timing of objectives and outcomes were clarified to state that the measures were assessed ‘following the start of the clinical intervention’. The clinical intervention was defined as the ‘heel lance followed by ROP screening’.

• Change to eligibility criteria: requirement for a senior clinician to assess suitability of the infant prior to enrolment in the trial changed from an exclusion to an inclusion criterion.

• Unblinding procedure: changed to allow medical staff to unblind participants immediately without input from the chief investigator or trial team.

• Changes to procedure for recording AEs: only AEs identified as serious would be recorded.

• Reporting procedures for foreseeable SAEs: increased respiratory support with a significant increase in oxygen requirement or an increase in support modality of grades 1, 2 and 3 added to the list of foreseeable SAEs.

• Reporting procedures for SAEs: addition of electronic SAE reporting.

• Analysis section: simplified, and the implied emphasis switched from significance testing to estimating the magnitude and direction of treatment effects with corresponding CIs.

• PIPP-R analysis: separated into the 30 seconds post ROP screening (co-primary outcome) and post heel lance (secondary outcome). The analysis method for both was changed from repeated measures analysis of variance (ANOVA) to linear regression, to explicitly mention calculating the mean difference and corresponding CI at a single time point and to facilitate adjustment for minimisation factors (only if required and technically possible). For skewed data, the median difference and corresponding CI would be calculated; this was changed from using Friedman’s test.

• Independent two-sample t-tests: changed to linear regression (to explicitly mention calculating the mean difference and corresponding CI).

• Mann–Whitney rank-based test: the test was replaced with estimating the median difference between groups with corresponding CI.

• Clinical stability and safety analyses: independent two-sample t-tests were changed to either Poisson or linear regression, dependent on the distribution of data or logistic regression for very infrequent occurrences (to explicitly mention calculating the mean difference and corresponding CI). Adjusting for the 24-hour baseline period, if appropriate, was also added for the clinical stability analysis.

• Analysis: the CI used for reporting secondary outcome measures was changed from 95% to 99%.

• Interim analysis details: was added for primary and secondary outcome measures.

• Additional information: was added to provide further justification for the trial, trial procedures, and elaborated definitions.

This version of the protocol was approved by the REC on 6 November 2015; however, the MHRA provided grounds for non-acceptance (see Appendix 5, Research Ethics Committee, Medicines and Healthcare products Regulatory Agency and Health Research Authority). As a result, this version of the protocol was not issued full approvals.

Version 3

Following feedback from the MHRA, it was decided that changes in respiratory support should be removed from the list of foreseeable SAEs. The research team would therefore report any change in respiratory support as a SAE if it met the definition. As more detailed data regarding the potential risks of morphine in the neonatal population would be valuable to clinical practice, following MHRA advice, the research team updated the protocol to indicate that AEs (with the exception of those already listed in the protocol) would be recorded. The protocol was also updated to confirm that the number of infants requiring an increase in respiratory support would be tested using logistic regression as a result of the binary nature of the data, and typographical errors in reference numbers were also amended. This version of the protocol was approved by the REC on 4 January 2016 and MHRA on 2 February 2016.

Version 4

Updates to version 4 of the protocol involved changes of IMP details to reflect the change in IMP supplier. The sequential safety procedure presented to the DMC was also added to the protocol. This would no longer be using the continuous stopping methods, as described by Bolland and Whitehead. 51 There had been recent developments in sequential stopping methods that were supported by commercially available and validated software. The details of the stopping boundaries using group sequential methods would therefore be agreed by the DMC and documented within the DMC charter. This version of the protocol was approved by the REC on 23 August 2016, MHRA on 15 August 2016 and HRA on 31 August 2016.

Version 5

Version 5 of the protocol incorporated the following updates:

• typographical and grammatical corrections

• removal of the specification that the syringe used for drug administration would be a 3-ml syringe

• specific details relating to the two-tailed t-test in the statistical analysis section were removed.

This version of the protocol was approved by the HRA on 25 October 2016. REC and MHRA approval were not required for these non-substantial amendments.

Version 6

Version 6 of the protocol incorporated the following updates:

• Clarification that the primary outcome (i.e. PIPP-R score) would be recorded during the 30-second period after ROP screening.

• Typographical and grammatical corrections.

• Clarification that should exclusion criteria manifest in the first 24 hours of clinical stability monitoring, the IMP or placebo would not be administered at this time. If subsequent ROP screening was required and the infant met all eligibility criteria, the IMP or placebo would be administered at the required time point for this subsequent ROP screen.

• Clarification that once data analyses were complete, the randomisation code would be released to the primary researchers only (as the statistician would be unblinded to provide analyses to the DMC).

• Clarification of the approach to summarising and reporting measures of clinical stability, which would include a requirement of an increase in respiratory support.

• Clarification of the relationship between brain activity and nociceptive activity.

• Clarification of current practice in the neonatal unit where the study was conducted.

• Clarification that further detail for the planned analyses for this trial would be presented in a separate SAP, and the adjustments planned for comparative analyses.

• Clarification that p-values would be obtained alongside CIs for the outcomes of interest.

This version of the protocol was approved by the REC on 15 February 2017, MHRA on 2 March 2017 and HRA on 16 March 2017.

Version 1

Version 1 was original submitted to the REC. The recommendations from the REC were incorporated into version 2.

Version 2

Amendments were made following REC review. The changes include:

• Typographical and grammatical errors corrected.

• Under the section entitled ‘Who has reviewed the study?’, the Northampton REC was added.

• ‘What are the benefits?’ was updated to begin ‘We cannot guarantee any direct benefits . . .’

• The REC reference was inserted.

• ‘What is the purpose of the study’ was updated to provide details regarding current comfort techniques used in practice.

• The acronym ‘NPEU’ was spelt out in full at least once.

Version 2 of the PIL was approved by the REC on 23 July 2015 and MHRA on 21 August 2015.

Version 3

The wording in section ‘What is the purpose of the study’ was clarified to reassure parents that anaesthetic eye drops were used as standard routine practice. The word ‘novel’ was removed. This version of the PIL was approved by the REC on 4 January 2016 and MHRA on 2 February 2016.

Version 4

Version 4 was corrected for typographical errors. This version of the PIL was approved by the REC on 23 August 2016, MHRA on 15 August 2016, and HRA on 31 August 2016.

Version 1

The original was submitted to the REC. The recommendations from the REC were incorporated into version 2.

Version 2

Version 2 included changes requested by the REC. The REC requested typographical updates to ensure that there was consistency throughout the document. This version of the consent form was approved by the REC on 23 July 2015 and MHRA on 21 August 2015.

Version 3

The infant’s name was added to version 3 to identify the infant during recruitment (this was originally omitted in error). The colour coding specified in the footer was also removed in case copies were printed on different-coloured paper. This version of the consent form was approved by the REC on 4 January 2016 and MHRA on 2 February 2016.

Version 4

Version 4 was updated to refer to the revised PIL. This version of the consent form was approved by the REC on 23 August 2016, by the MHRA on 15 August 2016 and by the HRA on 31 August 2016.

Annual reports

The first annual REC report was submitted in July 2016. The second annual REC report was submitted in August 2017.

Development Safety Update Reports

The development safety update report (DSUR) for the IMP morphine sulphate (reporting period from 21 August 2015 to 20 August 2016) was submitted to the MHRA on 19 October 2016. A copy was also forwarded to the trial funder. The DSUR for the IMP morphine sulphate (reporting period from 21 August 2016 to 20 August 2017) was submitted on 19 October 2017.

Green-light release

The co-ordinating centre used a green-light release form to authorise the site to begin recruitment. This form documented the following information: confirmation of trial sponsor approval, REC favourable opinion, MHRA approval, batch certificate and QP release for IMP, completion of site initiation/training, completed site agreement, trust R&D approval, REC approval for site and receipt of the ISF/document box/pharmacy folder. Once these were confirmed, the trial manager provided, and documented, authorisation of release of the IMP to site. The trial manager confirmed that the required checks had been completed and IMP received at site, and the senior trials programmer activated the site on the online randomisation system and documented this action. The completed green-light release form was issued to site on 16 September 2016 to confirm approval to start recruitment.

Meeting 1: initial meeting

The first meeting of the DMC was held on 23 November 2015, at which the trial was introduced and the charter reviewed. The first meeting was face to face to facilitate full discussion and allow members to get to know each other.

Meeting 2: stopping boundaries meeting

A DMC meeting was held on 1 November 2016 to discuss options for group sequential stopping boundaries for safety. Of three possible options, the DMC agreed that a gamma spending function should be used, which specified a probability of incorrectly stopping for harm (type I error rate) equal to 0.20, resulting in a probability of correctly detecting a safety issue (power) of 0.790. The DMC confirmed that it would like to be alerted to safety outcomes approaching or crossing the stopping boundary and would like further clinical information to review each case individually. The DMC agreed that the next meeting would be scheduled once 25 infants had been recruited to the trial.

Meetings 3 and 4

Meetings 3 and 4 of the DMC were held on 4 and 7 December 2017, respectively, in which the committee recommended that the trial be stopped.

Meeting 5

Meeting 5 of the DMC was a joint committee meeting with the TSC to review the SAP.

Meeting 6

Meeting 6 of the DMC was a joint committee meeting with the TSC and collaborators to review the draft manuscript reporting the trial findings.

Meeting 1: initial meeting

The first meeting of the TSC was held on 14 December 2015, at which the trial was introduced and the charter reviewed. The TSC advised that the PMG should carry out pilot work to obtain baseline data on clinical stability, carry out a power calculation on the secondary clinical stability outcome measure, investigate use of the Medical Dictionary for Regulatory Activities (the team investigated this, but chose not to pilot this coding for the trial), and contact the trial ophthalmologist about the use of RetCam^® (a wide-field digital imaging system; Natus Medical Inc., Pleasanton, CA, USA) and timing of eye drops to ensure that there was consistency across participants (this was discussed at the collaborators meeting in March 2016). The TSC also requested a factual and objective summary outlining reasons for recruitment delays, the PIL and consent form, and the NPEU data-sharing policy to the TSC.

Meeting 2

A second TSC meeting took place on 16 November 2016, in which the trial progress update report and stopping guideline agreed by the DMC were discussed.

Meeting 3

Meeting 3 of the TSC was held on 13 December 2017, in which the decision to temporarily suspend the trial was made.

Meeting 4

Meeting 4 of the TSC was a joint committee meeting with the DMC to review the SAP.

Meeting 5

Meeting 5 of the TSC was a joint committee meeting with the DMC and collaborators to review the draft manuscript reporting the trial findings.