Article Text

Download PDFPDF
Blood pressure intervention levels in preterm infants: pilot randomised trial
  1. Sujith Stanley Pereira1,2,
  2. Ajay Kumar Sinha1,2,
  3. Joan Katherine Morris3,
  4. David F Wertheim4,
  5. Divyen K Shah1,2,
  6. Stephen Terence Kempley1,2
  1. 1 Neonatal Unit, Royal London Hospital, Barts Health NHS Trust, London, UK
  2. 2 Centre for Genomics and Child Health, Blizard Institute, Barts and the London School of Medicine and Dentistry, London, UK
  3. 3 Wolfson Institute of Preventive Medicine, Queen Mary University of London, London, UK
  4. 4 Faculty of Science, Engineering and Computing, Kingston University, Kingston, UK
  1. Correspondence to Dr Stephen Terence Kempley, Centre for Genomics and Child Health, Blizard Institute, Barts and The London School of Medicine and Dentistry, London E1 2AT, UK; s.t.kempley{at}


Objective To examine the feasibility of a trial allocating different blood pressure (BP) intervention levels for treatment in extremely preterm infants.

Design Three-arm open randomised controlled trial performed between February 2013 and April 2015.

Setting Single tertiary level neonatal intensive care unit.

Patients Infants born <29 weeks’ gestation were eligible to participate, if parents consented and they did not have a major congenital malformation.

Interventions Infants were randomised to different levels of mean arterial BP at which they received cardiovascular support: active (<30 mm Hg), moderate (<gestational age mm Hg) or permissive (signs of poor perfusion or <19 mm Hg). Once this threshold was breached, all were managed using the same treatment guideline. BP profiles were downloaded continuously; cardiac output and carotid blood flow were measured at 1 day and 3 days, and amplitude integrated EEG was recorded during the first week. Cranial ultrasound scans were reviewed blind to study allocation.

Main outcome measure Inotrope usage and achieved BP.

Results Of 134 cases screened, 60 were enrolled, with mean gestation 25.8 weeks (SD 1.5) and birth weight 817 g (SD 190). Invasively measured BP on the first day and inotrope usage were highest in the active and lowest in the permissive arms. There were no differences in haemodynamic or EEG variables or in clinical complications. Predefined cranial ultrasound findings did not differ significantly; no infants in the active arm had parenchymal brain lesions.

Conclusion The BP threshold used to trigger treatment affects the achieved BP and inotrope usage, and it was possible to explore these effects using this study design.

Trial registration number ISRCTN83507686.

  • blood pressure
  • extremely preterm infant
  • randomised trial

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

What is already known on this topic?

  • The criterion for supporting low blood pressure (BP) in extremely preterm infants remains controversial.

  • A gestation based approach to supporting BP remains the most common method.

  • The relation between low BP and adverse outcomes and the safety of antihypotensive treatment remains unknown.

What this study adds?

  • It is feasible to carry out a three-arm randomised controlled trial investigating BP intervention levels in extremely preterm infants.

  • The majority of infants below 25 weeks’ gestation received inotropes irrespective of the arm they were randomised to.

  • The BP threshold used to trigger treatment affects achieved BP and inotrope usage.


The management of cardiovascular disturbances in extremely preterm infants differs considerably between neonatal units.1–6 Blood pressure (BP) is routinely measured; some studies show low BP on the first day is associated with adverse outcomes,7 but observational studies have not shown links with neurodevelopmental sequelae.8 9 The criteria for supporting low BP has been identified as a research priority.10

With challenges in estimating cardiac output and systemic vascular resistance, BP is frequently used as a proxy.11–13 The ‘normal’ range of BP for extremely preterm infants remains elusive, and whether such ranges ensure optimal organ perfusion remains unknown.2 14

Over two decades ago, a pragmatic approach was proposed to maintain mean arterial blood pressure (MABP; in mm Hg) above the gestation of the infant in weeks,15 and this remains widely practised.4 5 16 Some units practise permissive hypotension, whereby BP is not supported without clinical or biochemical evidence of poor perfusion. A minority of units maintain MABP above 30 mm Hg16, with some studies showing cerebral white matter changes17 or impaired autoregulation18 with BP below this level.

Observational studies have suggested an association between treatment for hypotension and adverse outcomes.19 Antihypotensive treatment may simply identify cardiovascular instability but could itself be harmful. Randomised trials need to examine whether low BP in extremely preterm infants should be supported, to assess whether inotropic agents are beneficial or harmful and to determine BP levels at which clinicians should intervene.

The aim of this study was to compare in extremely preterm infants (≤28 weeks) the different BP intervention levels at which clinicians should commence circulatory support in the neonatal unit (these intervention levels were MABP <30 mm Hg, MABP <gestation, or treating impaired perfusion and/or MABP <19 mm Hg). We hypothesised that a higher BP intervention level would represent a lower threshold for treatment, resulting in higher achieved BP during the first week of life and more inotropic agent usage during the infant’s neonatal unit admission.


Study design

This single-centre, three-arm, parallel, pilot randomised clinical trial investigated intervention levels for BP support in extremely premature neonates. It was conducted in a tertiary medical and surgical neonatal intensive care unit (NICU) with 36 cots and over 600 annual admissions.


Infants were eligible if born <29 weeks’ gestation, recruited and randomised within 12 postnatal hours. The only exclusions were major congenital malformation, lack of parental consent or lack of assent from clinicians. Written information was provided to parents when delivery was anticipated, and parents offered assent before delivery. Written consent was obtained within 12 hours following delivery. For unexpected deliveries or postnatal transfers, this took place after admission. Postnatal consent was required by the ethics committee, following a parent consultation in which 67% stated a preference for formal postnatal consent before recruitment.

Randomisation and masking

Randomisation was performed using stratification in individual weeks of gestation, with permuted blocks of three patients. A random number generator allocated blocks, with allocation concealed in sequentially numbered opaque sealed envelopes (by STK). Clinicians were unaware of block size; all allocations were concealed and administered in correct order with no deviations. Following recruitment, a different team member (SSP) opened the envelope to reveal the BP intervention level.


Infants were randomised into one of three arms. They received BP support if they met these criteria:

Active arm

MABP was supported if it fell below 30 mm Hg for more than 15 consecutive minutes.

Moderate arm

MABP was supported if it fell below the infant’s gestational age in mm Hg for more than 15 consecutive minutes.

Permissive arm

Cardiovascular support was given for clinical evidence of impaired tissue perfusion assessed by the clinician (poor skin perfusion with capillary refill time >3 s, urine output <1 mL/kg/hour from weighing of urine, worsening base excess (<-8mmol/L) or rising lactate (>2 mmol/L) on blood gas analysis) and/or if MABP fell below a lower safety net value of 19 mm Hg for more than 15 consecutive minutes. Infants also qualified for cardiovascular support if MABP remained ≥19 mm Hg with evidence of impaired tissue perfusion.

The threshold of 19 mm Hg in the permissive arm was arbitrarily chosen, as the lowest value that clinicians accepted, giving separation from the lowest moderate arm value.

Infants received the same BP management across the three arms of the study. The therapeutic goal of cardiovascular support was to achieve a MABP (or measure of tissue perfusion), which was greater than the threshold at which the infant became eligible for the intervention.

Circulatory management

BP management prior to randomisation followed the unit’s established guidelines, the same as in the active arm. Perinatal management included antenatal steroids, a modest delay in umbilical cord clamping (30–60 s), with no routine intravascular volume before NICU.

A written protocol tailored for each gestation and study arm was at the cotside to guide staff, with BP intervention levels. Apart from this target, the guideline (online supplementary appendix A) was similar to standard NICU policy. Key elements included triggers for evaluation (BP, tachycardia and impaired perfusion), basic stabilisation and a maximum of 10–20 mL/kg intravascular volume. First-line inotropes were dopamine for low BP and dobutamine for poor perfusion. If inotropes were escalated, clinical re-evaluation was performed, usually involving functional echocardiography to target treatment (options included intravascular volume, second inotropes, pressors, hydrocortisone, pulmonary vasodilators or PDA treatment). Clinician discretion in starting inotropes was allowed, assessing overall clinical status, so starting therapy before breaching BP threshold was not a protocol deviation. Clinicians were not blinded to the various intervention groups.

All infants had amplitude-integrated electroencephalography (aEEG), right common carotid artery blood flow (RCCAF) volume, left ventricular output (LVO) and superior mesenteric artery (SMA) blood flow velocity measurements on days 1 and 3.

BP monitoring

Infants with both invasive and non-invasive BP monitoring were included to reflect clinical practice, where invasive BP monitoring is not universally instituted. The decision to monitor BP invasively was made by the responsible clinician.

Invasive BP measurements were obtained from transducers attached to umbilical arterial catheters in a high position (thoracic vertebral body level T6-10) calibrated daily at the midaxillary line. Precautions were taken to ensure good quality waveforms, displayed using GE Healthcare systems (Carescape Monitor B850) and downloaded onto a laptop every 10 s for the first week.

Non-invasive BP was performed every 15 min if below threshold, otherwise hourly for the first 12 hours and 4-hourly thereafter, using the GE oscillometric method with appropriate-sized cuffs.

Physiological measurements

Blood flow measurements were performed using Doppler ultrasound and diameter measurement, with a Phillips iE33 system (Bothwell, USA). A 4–12 MHz transducer was used for LVO using an apical view,20 which correlates well with phase contrast MRI.21 A 7–15 MHz probe was used to measure RCCAF volume using an established method22 shown to have good reproducibility. A 5–8 MHz transducer was used for SMA velocity.23

Electroencephalography recording (aEEG) was recorded for 72 hours using a two–channel BRM3 monitor (BrainZ Instruments, Canada). Hydrogel electrodes were placed in the centroparietal regions (C3–P3 and C4–P4) bilaterally according to the 10–20 system.24 A 2-hour artefact and seizure-free electroencephalogram trace, recorded before and after RCCAF measurement, was analysed. Cross-cerebral aEEG was assessed for median, minimum and maximum amplitude. The discontinuity of the trace was calculated from the raw electroencephalogram (threshold 20 microvolts) using software25 in MATLAB (The MathWorks, USA).

Classification of cerebral pathology on cranial ultrasound scan (CrUSS)

Unit policy was for CrUSS on days 1, 3, 7, 14 and 28 and at 36 weeks’ corrected age or before discharge. All first week CrUSS, and the CrUSS closest to 36 weeks, were reviewed independently by two neonatal consultants (STK and AKS) blind to allocation. Periventricular haemorrhage (PVH) was classified on the Papile scale26; later findings included porencephalic cyst, periventricular leukomalacia and ventricular dilatation.

Clinical and physiological outcomes

Predefined primary outcomes were MABP during the first week and the use/duration of inotropes. Predefined secondary clinical outcomes included death or parenchymal brain abnormality on cerebral ultrasound, death before discharge home from hospital, periventricular leukomalacia, parenchymal or other PVH, necrotising enterocolitis using Bell’s staging criteria and localised intestinal perforation, treatment for patent ductus arteriosus, maximum serum creatinine and potassium in the first 2 weeks, duration of respiratory support, oxygen dependency or respiratory support at 36 weeks’ postconceptual age and postnatal steroids including hydrocortisone.

Physiological outcomes were RCCAF, SMA blood flow velocity, LVO and EEG measurements as detailed above.

Statistical analysis

Sample size calculations were performed for 80% power at a 5% significance level, with two-sided tests in a three-armed study design, without correction for multiple comparisons. For the primary outcome of inotrope usage, previous data showed rates of 15% where permissive hypotension was practised27 and 68% in our unit using active management (unpublished data). Sample size was calculated from a 15% versus 65% difference in proportions, with continuity correction, requiring 18 patients in each group (STATA, V.14). Inotrope usage was used in these calculations as data were not available for duration of inotrope usage nor MABP during the first week. A planned study size of 20 patients in each study arm was used to allow for attrition or uneven recruitment.

To reduce multiple pairwise testing, data were examined for effects across three treatment arms, using tests for ordered effects (assuming active > moderate > permissive). Analysis of variance and predefined contrasts were used for normally distributed continuous variables, and the Jonckheere-Terpstra test for non-normal data. For categorical outcomes, χ2 test with linear-by-linear association was used, or Fisher’s exact test for predicted cell numbers <5.

After excluding artefacts, invasive BP during the first 3 days was analysed in a mixed effects general linear model, with random effects included in the intercept of the model to account for between-baby variation, and gestation modelled as a fixed effect by completed week of gestation and time as a fixed effect by 4-hourly epochs.

As an exploratory pilot trial, tables give statistical significance without correction for multiple testing and indicate whether p values remain significant using the Benjamini-Hochberg correction,28 with a false discovery rate of 10% for each analysis (clinical outcomes and physiological outcomes).


Patient characteristics

The trial ended after planned recruitment of 60 infants between February 2013 and April 2015. This represented 45% of 134 NICU admissions within eligible gestation and postnatal age ranges (figure 1). In two patients, clinicians requested that the patient should not be recruited because of a clinical and echocardiographic diagnosis of persistent pulmonary hypertension of the newborn. There were no differences between those recruited and those who were not in mean gestation (25.8 vs 26.0 weeks), inborn proportion (70% vs 64%) or mortality (15% vs 19%).

Figure 1

Participant flow through the study. NICU, neonatal intensive care unit; PPHN, persistent pulmonary hypertension of the newborn.

No parents gave antenatal assent. All those recruited were followed up and included in the analysis with no protocol deviations. Clinician discretion in starting inotropes was allowed, assessing overall clinical status, so starting therapy before breaching BP threshold was not a protocol deviation. Median randomisation age was 8 hours (range 1–12). Randomised groups were comparable (table 1).

Table 1

Perinatal characteristics at the time of recruitment and randomisation

BP and cardiovascular support

Initial analysis of intermittent, staff-recorded BP, including both invasive and non-invasive measurements, did not show MABP differences between groups. An MABP <19 mm Hg was found in four infants in the permissive, none in the moderate and one in the active arm. Non-invasive MABP measurement was on average 11 mm Hg higher than the preceding invasive value; the 95% limits of agreement between staff-recorded and continuously downloaded invasive BP was ±6mm Hg, with no systematic bias. Detailed analysis of BP was therefore restricted to 51 infants (85%) with continuous invasive BP monitoring between 12 hours and 72 hours.

On the first postnatal day, invasive BP was highest in the active arm and was most stable in this group, with no significant effects in any time epoch during the first 3 days (figure 2). In the moderate arm, BP was significantly reduced at 12–15 hours of age (−3.1 mm Hg, 95% CI −5.2 to −1.0) and overshot to become significantly elevated at 32–39 hours of age (+3.5 mm Hg at 36–39 hours, 95% CI 1.3 to 5.6). In the permissive arm, BP was significantly reduced at 8–19 hours of age (−5.2 mm Hg at 12–15 hours, 95% CI −6.3 to −4.1), with a short period of elevation at 24–27 hours (+1.7 mm Hg, 95% CI 0.6 to 2.7). After 72 hours, there were no major differences in BP between the groups (online supplementary table 1).

Figure 2

Mean arterial BP in each study arm, from continuously downloaded BP data, averaged in 1 hour periods. Statistical analysis was performed on data averaged in 4-hour epochs, by mixed effects general linear model, with Benjamini-Hochberg correction for multiple comparisons. Periods when there were significant effects of time epoch and study arm are indicated by blue/dark bars. (Values are hourly mean±SD, for patients with invasively measured BP only (n=51; active=17, moderate=18, permissive=16). BP, blood pressure.

Inotropic support was given most frequently, and for longest duration, in the active arm (79% inotropes, mainly dopamine) and was used least in the permissive arm (48%, p=0.05). Although 12/35 (34%) of the patients receiving inotropes had their treatment started before randomisation, this did not differ between groups, and after randomisation, inotropes were started significantly more often in the active arm. Intravascular volume use prior to inotropes was similar in different study arms (active 7/15, moderate 8/10 and permissive 8/10). In the permissive arm, three infants had an MABP <19 mm Hg recorded in the hour that inotropes were started. There were no differences in hydrocortisone use, intravascular volume expansion or in PDA treatment (table 2).

Table 2

Cardiovascular treatments, circulatory measurements and electroencephalographic variables

Haemodynamic and electroencephalographic variables

Detailed echocardiography and aEEG were performed simultaneously on days 1 and 3 (median 18 and 77 hours). Cardiac output, ductal shunt, RCCAF and aEEG variables did not differ significantly between study arms (table 2).

Clinical outcomes

There were no significant differences in clinical outcomes, including mortality, duration of care, respiratory or gastrointestinal complications, retinopathy or renal variables (supplementary table 2).

Cranial ultrasound findings

The review of CrUSS findings reclassified clinical team reports in 17% cases. There was blinded agreement in 92% and full agreement with minor adjustments after discussion.

Parenchymal PVH occurred in only 2/60 infants. A normal CrUSS was found most often in the active arm, and combined rates of grades 2–4 PVH were highest in the moderate arm (active 0%, moderate 30%, permissive 5%; Fisher’s exact: p=0.008). The predefined outcome of death or parenchymal brain abnormality did not differ between groups (table 3).

Table 3

Cranial ultrasound findings, assessed blinded to study allocation


This is the first prospective randomised trial of BP intervention levels in preterm neonates to report its findings, demonstrating feasibility of such a study. This design achieved separation between study arms in inotropic therapy and BP levels. However, differences were less than might have been supposed, with major differences in invasively measured BP found mainly in the first 2 days. Differences in inotrope usage mainly occurred at 25–26 weeks’ gestation; most infants below this received inotropes, and most infants at higher gestations did not. The differences between intermittent staff BP recordings and continuous invasive measurements suggest that future studies should concentrate on infants with invasive monitoring, using automatically downloaded data.

Differences in BP between groups were not reflected in cardiac output or EEG variables measured at fixed time points, but these measurements may have missed the period when BP differed most. This pilot study was not powered to detect differences in major clinical outcomes, and reassuringly there were no effects on mortality.

There were no statistically significant differences in prespecified cranial ultrasound findings between study arms. However, it was notable that very few infants in the active arm had significant intracranial pathology, only three had subependymal PVH and none had parenchymal lesions. The highest rates of PVH were found in the moderate arm. Fluctuating pressure passive cerebral blood flow in sick preterm infants is associated with PVH,29 30 so it is biologically plausible that the instability of BP in the moderate arm could provide a mechanism for more intracranial pathology in this group.

The design of randomising patients to different BP intervention levels, but instituting the same management in all study arms, encouraged parents and clinicians to let infants participate in the study, with reasonable recruitment. Added safety by measuring cardiac output may have reassured clinicians. This contrasts with difficult recruitment in a trial using fixed intervention levels, with patients randomised to treatment or placebo.31 We did not prevent clinical staff from starting inotropes before recruitment or when they considered it clinically indicated, even if the BP criteria had not been breached for 15 min. As a significant proportion of infants had inotropes started before randomisation, we probably underestimated the differences in inotrope use that these policies would have produced if instituted from birth, even if the policy subsequently modified inotrope duration and dose. The reasons for starting treatment appear to have been based on a number of factors including BP and perfusion markers. Future studies could explore these reasons further and may find larger differences using cluster randomised trials, or with a waiver of consent, if considered acceptable to parents.

As an exploratory pilot, this study had limited power for clinical outcome effects. By analysing cardiovascular and EEG measurements at fixed time-points, we may have missed perfusion disturbances occurring at different times in each patient. Future studies of physiological variables should expect the greatest differences at 12–36 hours of age. Infants did not have pretrial entry cranial/cardiac ultrasound scans, but the numbers of infants with normal CrUSS on day 1 did not differ significantly between groups. Given the benefits of delayed cord clamping,32 33 a modest delay in cord clamping (30–60 s) was the standard of care, but the exact time of delay was not recorded. Inotropic support pre-randomisation was relatively higher in the moderate arm but not statistically different between the arms. Though the effect of delayed cord clamping on inotropic support is well described, this does not account for the trend in inotropic support post-randomisation, where infants in the active arm received the most inotropic support. Volume expansion was given prior to inotropic support in the majority of infants, 23 out of 35 (66%), did not differ between study arms and was not related to delayed cord clamping. This study only examined a single aspect of cardiovascular management, BP intervention level. There are many other issues of equal importance, to be studied independently, including detection of altered perfusion, optimum therapies for hypotension or impaired perfusion, treatment of the ductus arteriosus and whether treatment should be individualised.

We recommend that future studies should concentrate on infants with invasive BP monitoring, with data preferably downloaded for analysis. RCTs may be clinically acceptable if there is a safety-net, using echocardiography or other non-invasive methods, to detect severe cardiovascular compromise. Important outcomes can include death or neurodevelopmental impairment but should also examine PVH using blinded classification of CrUSS. It may be worth examining approaches that are either more or less active than those in common use.


We would like to thank Dr N Aladangady and Dr A Groves for their input into the study data monitoring committee. We thank the parents and infants who participated in this study and also all the doctors and nursing staff of the neonatal intensive care unit at the Royal London Hospital. We are grateful to the parents who raised funds to support this study and to Barts Charity who administered this funding.



  • Contributors SSP, AKS, DKS and STK devised the study concept. All authors developed the study protocol and designed the study. SSP and STK supervised and gathered data. SSP and STK analysed and interpreted the data. DFW analysed the electroencephalographic data. JKM performed statistical analysis. SSP and STK drafted the report. All authors critically revised the report. SSP and STK had full access to the data and take responsibility for the integrity of the data and accuracy of the analysis.

  • Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.

  • Competing interests DFW is an inventor on a patent US5181520 ‘Method and apparatus for analysing an electro-encephalogram’.

  • Patient consent Parental/guardian consent obtained.

  • Ethics approval National Research Ethics Service, London Surrey Borders Research Ethics Committee (12/LO/1553) .

  • Provenance and peer review Not commissioned; externally peer reviewed.