534 www.thelancet.com/respiratory Vol 3 July 2015

Articles

Lancet Respir Med 2015; 3: 534–43

Published OnlineJune 16, 2015

http://dx.doi.org/10.1016/S2213-2600(15)00204-0

See Comment page 499

*Members listed in the appendix

Division of Neonatology, University Hospital for Children

and Adolescents, University of Leipzig, Leipzig, Germany

(Prof U H Thome MD, C Gebauer MD); Division of

Neonatology, Dr. von Hauner University Children’s Hospital, Ludwig Maximilian University

of Munich, Munich, Germany (Prof O Genzel-Boroviczeny MD);

Division of Pediatric Pneumology, Allergology and

Neonatology, Hannover Medical School, Hannover,

Germany (Prof B Bohnhorst MD, Prof G Hansen MD); Division of

Neonatology and Pediatric Critical Care, University

Hospital for Children and Adolescents, University of Ulm,

Ulm, Germany (M Schmid MD, M Zernickel MSc,

Prof H D Hummler MD); Division of Neonatology and Pediatric

Critical Care, University Hospital for Children and

Adolescents, Albert Ludwigs University Freiburg, Freiburg,

Germany (H Fuchs MD, Prof R Hentschel MD); Division of Neonatology and Pediatric

Critical Care, Elisabeth Children’s Hospital, Klinikum Oldenburg, Medical Campus, Carl von Ossietzky University

of Oldenburg, Oldenburg, Germany (O Rohde MD,

J Seidenberg MD); Hospital for General Pediatrics and

Neonatology, Otto von Guericke University

Magdeburg, Magdeburg, Germany (S Avenarius MD, R Böttger MD); Division of

Neonatology, University Hospital for Children and

Permissive hypercapnia in extremely low birthweight infants (PHELBI): a randomised controlled multicentre trialUlrich H Thome, Orsolya Genzel-Boroviczeny, Bettina Bohnhorst, Manuel Schmid, Hans Fuchs, Oliver Rohde, Stefan Avenarius, Hans-Georg Topf, Andrea Zimmermann, Dirk Faas, Katharina Timme, Barbara Kleinlein, Horst Buxmann, Wilfried Schenk, Prof Hugo Segerer, Norbert Teig, Corinna Gebauer, Roland Hentschel, Matthias Heckmann, Rolf Schlösser, Jochen Peters, Rainer Rossi, Wolfgang Rascher, Ralf Böttger, Jürgen Seidenberg, Gesine Hansen, Maria Zernickel, Gerhard Alzen, Jens Dreyhaupt, Rainer Muche, Helmut D Hummler, for the PHELBI Study Group*

SummaryBackground Tolerating higher partial pressure of carbon dioxide (pCO₂) in mechanically ventilated, extremely low birthweight infants might reduce ventilator-induced lung injury and bronchopulmonary dysplasia. We aimed to test the hypothesis that higher target ranges for pCO₂ decrease the rate of bronchopulmonary dysplasia or death.

Methods In this randomised multicentre trial, we recruited infants from 16 tertiary care perinatal centres in Germany with birthweight between 400 g and 1000 g and gestational age 23–28 weeks plus 6 days, who needed endotracheal intubation and mechanical ventilation within 24 h of birth. Infants were randomly assigned to either a high target or control group. The high target group aimed at pCO₂ values of 55–65 mm Hg on postnatal days 1–3, 60–70 mm Hg on days 4–6, and 65–75 mm Hg on days 7–14, and the control target at pCO₂ 40–50 mmHg on days 1–3, 45–55 mm Hg on days 4–6, and 50–60 mm Hg on days 7–14. The primary outcome was death or moderate to severe bronchopulmonary dysplasia, defi ned as need for mechanical pressure support or supplemental oxygen at 36 weeks postmenstrual age. Cranial ultrasonogr ams were assessed centrally by a masked paediatric radiologist. This trial is registered with the ISRCTN registry, number ISRCTN56143743.

Results Between March 1, 2008, and July 31, 2012, we recruited 362 patients of whom three dropped out, leaving 179 patients in the high target and 180 in the control group. The trial was stopped after an interim analysis (n=359). The rate of bronchopulmonary dysplasia or death in the high target group (65/179 [36%]) did not diff er signifi cantly from the control group (54/180 [30%]; p=0·18). Mortality was 25 (14%) in the high target group and 19 (11%; p=0·32) in the control group, grade 3–4 intraventricular haemorrhage was 26 (15%) and 21 (12%; p=0·30), and the rate of severe retinopathy recorded was 20 (11%) and 26 (14%; p=0·36).

Interpretation Targeting a higher pCO₂ did not decrease the rate of bronchopulmonary dysplasia or death in ventilated preterm infants. The rates of mortality, intraventricular haemorrhage, and retinopathy did not diff er between groups. These results suggest that higher pCO₂ targets than in the slightly hypercapnic control group do not confer increased benefi ts such as lung protection.

Funding Deutsche Forschungsgemeinschaft.

IntroductionExtremely preterm infants who survive often develop bronchopulmonary dysplasia, which is characterised by severely impaired alveolarisation, which in turn results in a markedly reduced area for gas exchange.1 Infants with bronchopulmonary dysplasia often need long-term oxygen supplementation and frequent hospital re-admissions,2 with consequent high morbidity and health-care costs. Ventilator-induced lung injury is deemed one of the main pathogenic factors for the development of bronchopulmonary dysplasia and is mainly related to the magnitude of tidal volume.3,4 Reduction of tidal volumes might result in alveolar hypoventilation with increased blood partial pressure of carbon dioxide (pCO₂), which might be benefi cial.5 The intentional reduction of the intensity of mechanical ventilation and allowing pCO₂ values above 45 mmHg is referred to as permissive hypercapnia.

Increased respiratory drive through higher pCO₂ has been used for decades to wean infants off mechanical ventilation. Early use of permissive hypercapnia in newborn preterm infants from the fi rst day of life has been controversial, because it might increase cerebral perfusion, which can enhance oxygen delivery to the brain, but also increases the risk of intracranial haemorrhage.6 Furthermore, young rats have developed retinopathy when exposed to very high pCO₂ values (roughly 100 mm Hg).7 Instead, because the susceptibility to ventilator-induced lung injury might be highest soon after birth,4 the reduction of tidal volumes with resultant permissive hypercapnia could be most benefi cial when applied early.

Data from some retrospective analyses suggest an association between higher arterial pCO₂ (PaCO₂) values during the fi rst days of life in preterm infants and a reduced incidence of bronchopulmonary dysplasia,8,9

Articles

www.thelancet.com/respiratory Vol 3 July 2015 535

whereas others do not.10,11 Furthermore, fi ndings of a randomised trial12 comparing two diff erent tidal volume settings in adults with acute respiratory distress syndrome showed increased survival and decreased morbidity in patients randomly assigned to the lower tidal volume, who also had higher PaCO₂.

Several trials have assessed how strategies to avoid mechanical ventilation can improve outcomes for preterm infants.13 For this approach to be successful, investigators need to accept higher than normal pCO₂ values, but it is not clear whether lung protection is due more to the use of non-invasive support than to increased pCO₂. In preterm infants already on mech-anical ventilation, results of four previous randomised trials14–17 of permissive hypercapnia and a meta-analysis18 did not show reduced incidences of bronchopulmonary dysplasia. However, two of these trials had small sample sizes,14,15 and in the third with 220 infants, the PaCO₂ diff erence between the groups was only 4 mm Hg.16 In the fourth trial, management diff ered between treatment groups in several aspects other than pCO₂ target ranges.17 However, no signifi cant increases in adverse events, especially the rate of intracranial haemorrhage, associated with permissive hypercapnia were reported, and some secondary analyses of the

prematurely terminated SAVE trial17 suggest some benefi cial eff ects.

In their eff orts to improve outcome, many neonatologists have accepted permissive hypercapnia as their standard of care,19 despite the absence of clearly proven benefi cial eff ects. Consideration has been given to even higher pCO₂ targets as being of even greater benefi t. Hence permissive hypercapnia has spread in today’s neonatal intensive care without suffi cient supporting evidence,20 with the optimum PaCO₂ target range for ventilated preterm infants still to be established. This situation led us to do a large multicentre trial comparing two markedly diff erent PaCO₂ target ranges in extremely low birthweight infants. We aimed to study whether a higher pCO₂ target range would reduce the rate of moderate to severe bronchopulmonary dysplasia 21 or death in extremely low birthweight infants needing mechanical ventilation. Furthermore, we aimed to fi nd out whether hypercapnia would be most benefi cial to the infants requiring the most ventilatory support.

MethodsStudy design and patientsIn this randomised multicentre trial, infants were recruited from 16 tertiary care perinatal centres in

Adolescents, Friedrich-Alexander University Erlangen, Erlangen, Germany (H-G Topf MD, W Rascher MD); Mutter-Kind-Zentrum, Klinikum rechts der Isar, Technical University of Munich, Munich, Germany (A Zimmermann MD); University Hospital for General Pediatrics and Neonatology, Justus Liebig University Giessen, Giessen, Germany (D Faas MD); Division of Neonatology, Hospital for Children and Adolescents, Vivantes-Hospital Neukölln, Berlin, Berlin, Germany (K Timme MD, Prof R Rossi MD); Hospital for Children and Adolescents, Children’s Hospital of the Third Order, Munich, Germany (B Kleinlein MD, Prof J Peters MD); Division of Neonatology, University Hospital for Children and Adolescents of the J.W. Goethe University Frankfurt am Main, Frankfurt am Main, Germany (H Buxmann MD, Prof R Schlösser MD); Hospital for Children and Adolescents, Central Hospital Augsburg, Augsburg, Germany (W Schenk MD); St. Hedwig Hospital, University of Regensburg, Regensburg, Germany (Prof H Segerer MD); Department of Neonatology and Pediatric Intensive Care, Katholisches Klinikum, Ruhr University Bochum, Bochum, Germany (N Teig MD); Division of Neonatology and Pediatric Critical Care, University Hospital for Children and Adolescents, Ernst Moritz Arndt University Greifswald, Greifswald, Germany (Prof M Heckmann MD); Division of Pediatric Radiology, University Hospital of the Justus Liebig University Giessen, Giessen, Germany (Prof G Alzen MD); and Institute of Epidemiology and Medical Biometry, University of Ulm, Ulm, Germany (J Dreyhaupt PhD, R Muche PhD)

Correspondence to:Prof Ulrich Thome, Division of Neonatology, Department of Women’s and Children’s Medicine, University Hospital of Leipzig, 04103 Leipzig, Germanyuhthome@web.de

Research in context

Evidence before this studyData from animal experiments suggested that permissive hypercapnia could be benefi cial for subjects requiring mechanical ventilation because of lower tidal volumes and additional biochemical eff ects. We hypothesised that such benefi ts might also apply to preterm infants. We searchedPubMed from 1960 until 2014 for English language articles with the following search terms: “preterm infant”, “permissive hypercapnia”, and “minimal ventilation”. We also searched the reference lists of previous review articles. Several retrospective analyses were inconclusive. Three randomised trials were identifi ed in which ventilator-dependent extremely low birthweight infants were allocated to diff erent partial pressure of carbon dioxide (pCO₂) targets. Two were small and monocentric and the third, the SAVE trial, was prematurely terminated for reasons unrelated to the mechanical ventilation. All three used diff erent PCO₂ targets and the pCO₂ diff erences between the randomly allocated groups were small. None of the trials showed reduced incidence of bronchopulmonary dysplasia associated with permissive hypercapnia. One of the monocentric trials reported faster weaning off mechanical ventilation, the other, however, a worse neurodevelopmental outcome. Investigators of the SAVE trial reported a smaller number of infants requiring long-term mechanical ventilation beyond a postmenstrual age of 36 weeks. Results of a meta-analysis encompassing all three trials with 334 infants were calculated which showed trends but no signifi cant diff erences

between the results of diff erent pCO₂ targets. In another very large randomised trial, management diff ered between treatment groups in several aspects other than pCO₂ target ranges.

Retrospective analyses also suggested an increased risk of intracranial haemorrhage that was not substantiated by the randomised trials. Animal experiments suggested an increased risk of retinopathy of prematurity associated with severe hypercapnia. Whether this risk also applied to human beings was unknown.

Added value of this studyWith a sample size of 359, this multicentre trial did not show increased lung protection and improved outcomes associated with higher pCO₂ targets, despite lower ventilator pressures. On the contrary, higher pCO₂ targets were associated with an increased incidence of bronchopulmonary dysplasia or death in the subgroup of infants with the worst lung disease. Furthermore, there was an unexpected increased incidence of necrotising enterocolitis, but no increase in the incidence of intracranial haemorrhage or retinopathy of prematurity.

Implications of all available evidenceMildly hypercapnic pCO₂ targets as used in the control group of this trial and the hypercapnia group of the SAVE trial seem to be safe and are likely to be associated with small health benefi ts. Higher pCO₂ targets as used in the high target group of this trial do not increase these benefi ts, and might cause harm.

See Online for appendix

Articles

536 www.thelancet.com/respiratory Vol 3 July 2015

Germany (appendix p 11). All infants who weighed 1000 g or less at birth were screened. Inborn infants with a gestational age of 23–28 weeks plus 6 days, weighing 400–1000 g and receiving endotracheal intubation and mechanical ventilation within 24 h of birth were eligible. Exclusion criteria were birth outside the prenatal centre’s delivery ward, chromosomal anomalies, congenital malformations requiring early surgery or otherwise

compromising respiratory care or outcome, hydrops fetalis, air leaks before randomisation, severe birth asphyxia, or a decision to provide compassionate care only. The trial was approved by each centre’s institutional review board and written informed consent was obtained from the parents or legal guardians of all infants. The trial was approved by the institutional review boards of all participating centres. On-site monitoring was done by the Interdisciplinary Centre for Clinical Trials (IZKS), University Medical Centre, Mainz, Germany. An independent data safety monitoring board (DSMB) consisting of four experienced neonatologists and one biostatistician evaluated critical safety issues as well as the protocol amendment to enable the interim analysis (appendix p 11). A database was programmed and all data were entered in duplicate.

Patient allocation and maskingInfants were randomly assigned (1:1) with a secure web-based randomisation system (e-randomixer, IZKS), and randomisation results were applied immediately. Ran-domisation was done by a block randomisation scheme with variable block sizes (2–6) stratifi ed by site and birthweight (three strata: 400–499 g, 500–749 g, 750–1000 g). It was not feasible to mask caregivers and parents because of the many clinically indicated blood gas determinations and ventilator adjustments required in neonatal intensive care.

ProceduresEndotracheally intubated and mechanically ventilated infants were randomly allocated to two diff erent target ranges of pCO₂. In both groups, an age-dependent pCO₂ increase was permitted to make weaning easier. In the high target group (experimental intervention) the PaCO₂ target range was 55–65 mm Hg from 1–3 days of life (0–72 h postnatal age), 60–70 mm Hg from days 4–6 (73–144 h), and 65–75 mm Hg from days 7–14 (145–336 h). In the control target group the PaCO₂ target range was 40–50 mm Hg from days 1–3 (0–72 h), 45–55 mm Hg from days 4–6 (73–144h), and 50–60 mm Hg from days 7–14 (145–336 h).

Postnatal age was counted from birth, and infants were entered into the schedule when they were randomly assigned. Randomisation and assignment to the randomised target range had to be completed within 12 h of endotracheal intubation, and was applied as long as the infants remained intubated or until the end of day 14. Extubation could be attempted if the PaCO₂ was maintained within or below the target range assigned with a rate of less than 30 breaths per min and FiO₂ was less than 0·5. After extubation, no pCO₂ targets were defi ned by the study protocol. In the case of re-intubation before day 14, the target range according to the randomised group assignment and actual postnatal age was resumed. Blood pCO₂ was to be measured in at least 12-h intervals, or more frequently if clinically indicated or

High target group (n=179)

Control group (n=180)

Gestational age (weeks) 25·6 ± 1·4 25·7 ± 1·3

Birthweight (g) 713 ± 156 709 ± 153

Boys 105 (59%) 99 (55%)

Antenatal steroids (any) 162 (91%) 157 (87%)

PPROM >24 h 45 (25%) 35 (20%)

Apgar score at 5 min 7 (1–9) 8 (1–9)

Age at intubation (h) 0 (0–22; 2·1) 0 (0–21; 1·3)

Intubation age ≥1 h 55 (31%) 57 (32%)

Surfactant replacement 172 (96%) 175 (97%)

Total surfactant (mg/kg) 188 (65–774) 183 (55–829)

Methylxanthine treatment 168 (94%) 168 (94%)

Data are mean ± SD, median (min–max); mean, or n (%). PPROM=preterm premature rupture of the membranes.

Table 1: Demographic data for study infants

Figure 1: Trial profi lePMA=postmenstrual age. Infants not receiving invasive assisted ventilation within 24 h of birth were not eligible for the study. Randomisation had to be done within 12 h of intubation. The main reason for non-enrolment for some eligible infants was that parents declined consent. Further reasons for this were language barriers with some parents, or unavailability of parents able to give consent within 12 h of birth.

430 not on assistedventilation

362 randomlyassigned3 dropouts

411 no consent4 chromosomal

anomaly24 malformation

1 hydrops14 asphyxia40 palliative care52 outborn

196 other

179 high target group 180 control target group

25 died before36 week PMA

40 on O2 or mechanicalsupport at 36 week PMA

35 on O2 or mechanicalsupport at 36 week PMA

19 died before36 week PMA

1534 patients screened

Articles

www.thelancet.com/respiratory Vol 3 July 2015 537

when measurement results outside the target range occurred. Both arterial and capillary pCO₂ measurements were accepted, because routine care in several of the study centres did not include arterial line placement in all infants, and there was consensus that arterial lines should not routinely be left in place for 14 days. The study protocol allowed waiving of all management restrictions in the case of severe complications.

To minimise volutrauma, a high ventilation rate (60–80 per min) was favoured over high tidal volumes in both groups. Initial ventilator settings comprised a rate of 60–80 per min or greater, inspiratory time of 0·25–0·35 s, positive end-expiratory pressure 3–6 mbar, and a peak inspiratory pressure resulting in minimal to moderate chest rise. The rate was allowed to be decreased only if the peak inspiratory pressure was 14 mbar or lower. Synchronised ventilation or forms of volume control were allowed to be used at the discretion of the clinicians in charge of patient care. Because the administration of sodium bicarbonate has been linked to increased lung damage,22,23 its use was discouraged. Furthermore, we attempted to prevent inconsistent use between the two study groups to avoid it becoming a confounder. Therefore, bicarbonate administration to correct a low pH in combined acidosis was linked to the base defi cit rather than the pH or pCO₂ and allowed only if the base defi cit exceeded an arbitrary level of –8 mmol/L, independent of pH and pCO₂. We used high frequency ventilation only as a rescue method. The fi rst dose of natural surfactant was generally given immediately after intubation as a standard treatment in all participating centres. However, this was not governed by the trial protocol since intubation occurred before enrolment. Further doses were to be given to all enrolled infants suff ering from respiratory distress syndrome and requiring at least a fraction of inspired oxygen (FiO₂) of 0·3. Up to two additional doses were given within 24 h if the FiO₂ requirement exceeded 0·3, unless there were contraindications. Caff eine was recommended for all participating infants.

OutcomesThe primary outcome of the trial was death or bronchopulmonary dysplasia before 36 weeks post-menstrual age according to the physiological defi nition of bronchopulmonary dysplasia—ie, requiring mech-anical pressure support or supplemental oxygen at 36 weeks postmenstrual age within ±2 days, including an oxygen reduction test for infants requiring less than 0·3 FiO₂ (bronchopulmonary dysplasia or death).24 The bronchopulmonary dysplasia part of this defi nition also represents moderate to severe bronchopulmonary dysplasia according to the National Institute of Child Health and Development (NICHD) consensus def-inition.21 Bronchopulmonary dysplasia status was estab-lished independent of caregiver assessments by a computer algorithm, which applied the above defi nition to collected clinical data as part of the statistical analysis.

Major secondary outcomes included the severity of bronchopulmonary dysplasia according to the consensus defi nition,21 and the incidence and severity of intracranial haemorrhage according to Papile and colleagues.25

High target group (n=179)

Control group (n=180)

p value

Moderate or severe bronchopulmonary dysplasia or death at 36 weeks postmenstrual age*

65 (36%) 54 (30%) 0·18

Mortality to 36 weeks postmenstrual age 25 (14%) 19 (11%)† 0·32

Moderate or severe bronchopulmonary dysplasia at 36 weeks postmenstrual age

40 (22%) 35 (19%) 0·44

Death before day 28 22 (12%) 16 (9%) 0·29

O₂ or positive pressure for ≥28 days 147 (82%) 150 (83%) 0·44

Mild bronchopulmonary dysplasia (consensus defi nition)

104 (62%) 112 (63%) 0·26

Moderate bronchopulmonary dysplasia (consensus defi nition)

18 (10%) 23 (13%)

Severe bronchopulmonary dysplasia (consensus defi nition)

22 (12%) 12 (7%)

Intraventricular haemorrhage all grades‡§ 50 (28%) 55 (31%) 0·46

Intraventricular haemorrhage present on day 1 without progress‡

9 (5%) 8 (4%) 0·63

Severe intraventricular haemorrhage (grade 3–4)‡ 26 (15%) 21 (12%) 0·30

Severe intraventricular haemorrhage (grade 3–4) present on day 1 without progress‡

7 (4%) 3 (2%) 0·48

Combined death or moderate to severe bronchopulmonary dysplasia or intraventricular haemorrhage (grade 3–4)

82 (46%) 67 (37%) 0·09

Periventricular leukomalacia‡ 16 (9%) 11 (6%) 0·31

Hydrocephalus internus‡ 24 (16%) 27 (17%) 0·76

Hydrocephalus internus with shunt 8 (4%) 7 (4%) 0·78

Infants who received bicarbonate treatment 93 (52%) 79 (44%) 0·13

Bicarbonate dose cumulative over the entire hospital stay, in infants who received bicarbonate (mmol/kg)

5·8 (0·5–46·2) 6·3 (0·7–38·0) 0·29

Extubated within the fi rst 14 days of life 137 (77%) 137 (77%) 1·0

Re-intubated within the fi rst 14 days of life 51 (29%) 51 (29%) 1·0

Days of sedative use per infant 2 (0–14) 2·5 (0–14) 0·59

Pulmonary interstitial emphysema 25 (14%) 32 (18%) 0·32

Pneumothorax 8 (5%) 13 (7%) 0·27

pCO₂ targets waived in response to a severe complication

32 (18%) 26 (14%) 0·38

Days pCO₂ targets waived 4 (1–10) 3 (1–13) 0·45

Postnatal dexamethasone treatment 30 (17%) 29 (16%) 0·81

Cumulative dexamethasone dose (mg) 0·65 (0·06–4·8) 0·66 (0·10–3·00) 0·9

Postnatal hydrocortisone treatment 38 (22%) 50 (28%) 0·17

Cumulative hydrocortisone dose (mg) 8·8 (0·6–219) 5·8 (0·5–90) 0·31

Retinopathy of prematurity 78 (47%) 78 (47%) 0·92

Severe retinopathy ≥ grade 3 20 (11%) 26 (14%) 0·36

Necrotising enterocolitis ≥ grade 2 20 (12%) 8 (5%) 0·02

Weight at 36 weeks postmenstrual age (g) 1943 (1200–3120) 2000 (1000–2830) 0·76

Weight gain to 36 weeks postmenstrual age (g) 1244 (325–2384) 1250 (200–2030) 0·59

Data are n (%) with χ2 test, median with Mann-Whiney U test (min–max), unless stated otherwise. *Primary outcome of trial, p value by group sequential analysis. †One additional patient with severe bronchopulmonary dysplasia died 1 day after completing 36 postmenstrual weeks. ‡As diagnosed by the study radiologist; high target (n=178), one ultrasound exam missing. §One infant was transferred and no follow-up cranial ultrasound scans were available.

Table 2: Overall outcome indicators

Articles

538 www.thelancet.com/respiratory Vol 3 July 2015

Cranial ultrasound examinations were done on the fi rst day of life, at 12–14 days, and at 36 weeks postmenstrual age and were evaluated centrally by one masked paediatric radiologist (GA). Retinopathy of prematurity was routinely screened for and classifi ed according to the International Classifi cation.26 Necrotising enterocolitis was diagnosed when the clinical and radiological fi ndings corresponded to stage II or higher according to the staging criteria of Bell and colleagues.27

Statistical analysisThe sample size was calculated for the χ² test according to a group sequential design,28 allowing earlier termination of enrolment in case a signifi cant diff erence was detected during scheduled interim analyses. A pre-trial primary

outcome rate of death or bronchopulmonary dysplasia incidence of 47%, showing a 20% relative reduction (from 50% to 40%) with a power of 80% and a signifi cance level of 5%, using a two-sided group sequential test with two interim analyses, required a maximum sample size of 830 patients.

Predefi ned secondary analyses were done by χ² tests, Student’s t tests, Mann-Whitney U tests, survival analyses, and, for repeated measures, linear mixed eff ects regression models. Subgroup analyses were done to test the hypothesis that hypercapnia is of most benefi t to infants at the highest risk of poor outcomes, by analysing interactions29 with log-linear Poisson regression with robust estimation of error variance.30 The predefi ned subgroups were infants in the three birthweight strata, small for gestational age infants defi ned (according to published German reference data31) as birthweight less than 10th percentile for the gestational age, and infants with more severe lung disease (FiO₂ >0·4 or mean airway pressure >10 mbar for >4 h). Sex was added post hoc to the subgroup analyses because of its importance as a highly relevant risk factor.32

All analyses were done on an intention-to-treat basis. A p value of less than 0·05 was deemed signifi cant. Sample size calculation and interim and fi nal analyses of the primary outcome were done with the software East (Cytel Software Corporation, Cambridge, MA, USA) and Addplan (Aptiv Solutions, Reston, VA, USA). For all other analyses, SAS software (SAS Institute, Cary, NC, USA) was used.

Early in 2012, an amendment of the study protocol was developed to allow for the interim analysis requested by the review board of the funding agency. The sequential design’s boundaries had not been exceeded at that time (ie, there were no detectable signifi cant diff erences). The study design was changed from a three-stage group sequential design into a two-stage adaptive group sequential design with one interim analysis. The amendment was approved by the DSMB and the lead investigator’s responsible institutional review board. The interim analysis was based on 312 completed infants and carried out by an independent statistician (BM; appendix p 10). The results were presented to the DSMB, which recommended terminating enrolment. The funding agency’s review board concurred with this recommendation and it was implemented by the study coordinator.

This trial is registered with the ISRCTN registry, number ISRCTN56143743.

Role of the funding sourceThe funding agency’s review board requested the interim analysis early in 2012, when it became clear that the recruitment rate would remain below the original projections. Apart from this request, the funding agency had no role in data collection, analysis, and interpretation, nor in writing the report. Access to the raw data was limited to the data manager (MZ) and the statistician (JD). The corresponding author (UHT) had full access to

Figure 2: Daily mean values of the partial pressure of carbon dioxide (pCO2) in all patients who were intubated at the time of measurementError bars indicate standard deviations, lower bars indicate numbers of patients contributing data. Shaded areas indicate the target ranges of the high target and control groups. The pCO2 values were signifi cantly higher in patients randomly assigned to the high target group than in those in the control target group (linear mixed eff ects regression model; p<0·0001), although the high target range was frequently not achieved because of the patients’ own respiratory eff orts. The main reason for absent data was extubation.

0

100

200

300

400

500

600

020

30

40

50

60

70

80

n

mm

Hg

Day of life1 2 3 4 5 6 7 8 9 10 11 12 13 14

Control targetHigh target

Figure 3: Daily mean values of the pH in all patients who were intubated at the time of measurementError bars indicate standard deviations, lower bars indicate numbers of patients contributing data. The pH values were signifi cantly lower in patients randomly assigned to the high target group than in those in the control target group (linear mixed eff ects regression model; p<0·0001). The main reason for absent data was extubation.

Control targetHigh target

01 2 3 4 5 6 7 8 9 10 11 12 13 14

100

200

300

400

500

600

07·10

7·15

7·20

7·25

7·30

7·35

7·40

n

pH

Day of life

Articles

www.thelancet.com/respiratory Vol 3 July 2015 539

all of the data and ultimate responsibility for submission for publication.

ResultsBetween March 1, 2008, and July 31, 2012, we screened 1534 infants and recruited 362, of whom three had to be excluded: two because parental consent was withdrawn and one after being mistakenly randomly assigned despite meeting an exclusion criterion (malformation), leaving 359 patients (179 in the high target group, 180 in the control group) for the intention-to-treat analysis (fi gure 1). Table 1 shows baseline demographic and clinical characteristics of the two groups.

The primary endpoint of bronchopulmonary dysplasia or death was met in 36% of patients in the high target group and 30% in the control group, which was not a signifi cant diff erence according to the Addplan analysis (table 2). Looking separately at the components of the composite primary outcome (ie, requirement for mechanical support or supplemental oxygen and mortality, either at an age of 28 days or at 36 weeks postmenstrual age), there were no signifi cant diff erences. Comparisons of bronchopulmonary dysplasia severity according to the consensus defi nition21 were also inconclusive.

Day-by-day mean values of pCO₂ and pH diff ered signifi cantly between study groups (fi gures 2 and 3). The pCO₂ diff erence peaked at 7 mm Hg on day 4 and the mean diff erence between days 2 and 11 was 6·2 mm Hg. Furthermore, infants in the high target group had signifi cantly lower values for peak inspiratory pressure (appendix p 5), suggesting increased weaning eff orts in the high target group. Tidal volumes were not measured at all centres, and no signifi cant diff erences were detectable (appendix p 6).

The incidence of intraventricular haemorrhage was similar in both groups, as were the incidence of retinopathy of prematurity, periventricular leukomalacia, hydrocephalus, and air leaks (table 2). Infants in the high target group had a signifi cantly higher rate of necrotising enterocolitis. Weight gain was similar in both groups. Typical for modern neonatal care, we used diff erent ventilation methods across diff erent study sites (appendix p 7); the choice of ventilation modes did not diff er signifi cantly between study groups. However, fi nal independence from the ventilator and supplemental oxygen was not accelerated by the high target (fi gures 4 and 5).

We noted no signifi cant interactions of risk factors with the primary outcome of bronchopulmonary dysplasia or death. In the predefi ned analyses, we did not identify a subgroup that might have benefi ted from the high target (table 3). Furthermore, all interactions were non-signifi cant. However, in infants with severe lung disease, the high target was associated with a higher incidence of bronchopulmonary dysplasia or death.

The signifi cantly increased incidence of necrotising enterocolitis prompted another unplanned subgroup

analysis, which showed an association between the incidence of necrotising enterocolitis and the high target in infants with severe lung disease (p=0·06) and in infants with 500–749 g birthweight (p<0·01; table 4).

During the follow-up, there were three additional deaths. Of these, two of the infants (one in the high target group, one in the control group) had been classifi ed for the primary outcome as having bronchopulmonary dysplasia and died within the fi rst year. The third, in the high target group, had not met our bronchopulmonary dysplasia criteria and died at 2 years corrected age.

DiscussionThe high pCO₂ target did not reduce the primary outcome. Our results should be interpreted in view of the protocol specifi cations for this trial. Infants were only eligible

Figure 4: Kaplan-Meier analysis of weaning from invasive mechanical ventilationThere was no signifi cant diff erence between groups (p=0·18, log-rank test). Circles indicate censored datapoints. Censoring generally occurred in patients who were still intubated on day of life 99 or in patients who died.

00

25

50

75

100

20 40 60 80 100Pr

opor

tion

of in

tuba

ted

infa

nts (

%)

High targetControl targetCensored

Day of life

Figure 5: Kaplan-Meier analysis of weaning from supplemental oxygenThere was no signifi cant diff erence between groups (p=0·25, log-rank test). Circles indicate censored datapoints. Censoring generally occurred in patients who were still on positive pressure or supplemental oxygen on day of life 99 or in patients who died.

00

25

50

75

100

20 40 60 80 100

Prop

ortio

n of

infa

nts

on su

pple

men

tal o

xyge

n (%

)

High targetControl targetCensored

Day of life

Articles

540 www.thelancet.com/respiratory Vol 3 July 2015

when they needed endotracheal intubation and invasive mechanical ventilation because pCO₂ control is very limited in non-intubated infants. In combination with increasing clinical eff orts to avoid invasive mechanical ventilation altogether,17,33,34 eligible infants were only a minor fraction of the admitted extremely low birthweight infants, and this markedly slowed enrolment. However, including only intubated infants led to the selection of less stable infants who were more likely to have pronounced respiratory distress syndrome, and thus to a highly relevant trial population with regard to the questions under investigation. Enrolment, initially projected for 3 years, was terminated after the interim analysis done in the fourth year. Several considerations played a part in this decision. First, the interim analysis showed no benefi t, and a possible trend favouring the control rather than the high target group. This fi nding suggested that it was futile to continue recruitment because a benefi t to the high target group showing over the remainder of the trial had

become extremely unlikely. Second, proving the opposite, a worse outcome in the high target group, would not change current standard of care. Third, ethical concerns were raised about the need to randomly assign hundreds of additional patients to achieve the original sample size with limited further scientifi c gain, whereas other multicentre trials poised to test important hypotheses were held back. Fourth, too many changes to clinical standards can confound trial results if patient recruitment exceeds more than 3–5 years.

Although the study fi nished prematurely, the sample size was still higher than the sum of all three previously published trials, in which investigators had enrolled only intubated infants and randomly assigned them into diff erent pCO₂ target groups.14–16 Furthermore, we recorded higher pCO₂ diff erences between groups. Despite the preference for higher risk infants who needed mechanical ventilation in settings in which non-invasive support was the rule, the overall rate of bronchopulmonary dysplasia or death was lower in our trial, perhaps because of continuous improvements in care. However, the results of the three previous trials, an older meta-analysis of two of these trials,18 and this trial, were similar, showing no diff erences in the rates for primary outcome between the study groups.

Diff erent pCO₂ targets were also used in the very large SUPPORT trial,17 which enrolled 1316 infants and compared a strategy favouring non-invasive continuous positive airway pressure and an upper pCO₂ limit of 65 mm Hg for intubation and extubation with a strategy of using primary intubation, surfactant administration, and an upper pCO₂ limit of 50 mm Hg. However, there were no lower limits for the pCO₂ and actual pCO₂ values were not reported. Like previous trials of hypercapnia and our trial, the SUPPORT trial found no diff erence in the primary outcome. However, time spent on mechanical ventilation was shorter in the continuous positive airway pressure and hypercapnia group, which is probably the result of the study protocol favouring extubation to continuous positive airway pressure in this group. Available data suggest that the encouraging trend towards a better outcome in the continuous positive airway pressure group of the SUPPORT trial might be more related to favouring continuous positive airway pressure and avoiding mechanical ventilation13 than to favouring hypercapnia. Nasal continuous positive airway pressure could have intrinsic benefi ts, which might include full control of expiratory braking by the vocal cords and a large leak to discharge excessive air quickly if necessary.

Several considerations led to the selection of the pCO₂ targets for this trial. No optimum pCO₂ target range has ever been defi ned. Data from bench research and clinical trials in adults suggested the benefi cial eff ects of hypercapnia and hypercapnic acidosis.5,35 In preterm infants, secondary and subgroup analyses of the prematurely terminated SAVE multicentre trial suggested reduced need for prolonged mechanical ventilation was

High target group (n=179)

Control group (n=180)

Risk ratio (95% CI) pinteraction

Overall 65 (36%) 54 (30%) 1·21 (0·90–1·63)

Groups according to birthweight (g) 0·74

400–499 14/22 (64%) 10/18 (56%) 1·15 (0·68–1·93)

500 –749 39/83 (47%) 32/90 (36%) 1·32 (0·92–1·89)

750–1000 12/74 (16%) 12/72 (17%) 0·97 (0·47–2·02)

Infants small for gestational age 0·62

Small for gestational age (birthweight <10th percentile)

21/35 (60%) 18/41 (44%) 1·37 (0·88–2·12)

Appropriate for gestational age (birthweight ≥10th percentile)

44/144 (31%) 36/139 (26%) 1·18 (0·81–1·71)

Severe lung disease 0·30

Infants with severe lung disease 37/64 (58%) 29/72 (40%) 1·44 (1·01–2·04)*

Infants without severe lung disease 28/115 (24%) 25/108 (23%) 1·05 (0·66–1·68)

Sex 0·24

Boy 40/105 (38%) 36/99 (36%) 1·05 (0·73–1·50)

Girl 25/74 (34%) 18/81 (22%) 1·52 (0·91–2·55)

Data are n (%). Eff ect of intervention according to baseline characteristics (log-linear Poisson regression with robust estimation of error variance for bronchopulmonary dysplasia or death).*Signifi cantly increased risk associated with the high target (p=0·04).

Table 3: Subgroup analyses for eff ect on moderate or severe bronchopulmonary dysplasia or death at 36 weeks postmenstrual age

High target group

Control group

p value

More severe lung disease 10/61 (16%) 4/68 (6%) 0·06

Less severe lung disease 10/113 (9%) 4/106 (4%) 0·12

Birthweight (g)

400–499 2/21 (10%) 1/17 (6%) 0·39

500–749 14/79 (18%) 5/89 (6%) p<0·01

750–1000 4/74 (5%) 2/68 (3%) 0·68

Data are n (%) with χ² test.

Table 4: Subgroup analyses for necrotising enterocolitis ≥ stage 2

Articles

www.thelancet.com/respiratory Vol 3 July 2015 541

associated with mild hypercapnia16 as compared with normocapnia, without adverse eff ects. Mild hypercapnia has since been increasingly introduced into routine care, and discussion centred on how high to increase the pCO₂ goal to maximise benefi ts. Additionally, the risk of accidental hypocapnia fell.36 Consequently, a control group with a normocapnic pCO₂ target of 35–45 mm Hg seemed unethical and a mildly hypercapnic target range was chosen. Because previous trials had been criticised for their small pCO₂ diff erences, the high target group was planned with much higher pCO₂ targets to increase the diff erence in pCO₂ between the groups, although it was anticipated that these targets might not always be achieved because of increased respiratory eff orts by the patients. Furthermore, we assumed that lung protection might be increased with higher pCO₂ goals, because ventilator requirements and mechanical forces are decreased, and biochemical benefi ts of a high pCO₂ might depend on its concentration.37 Finally, stepwise increases of the target ranges of both groups were included because this was standard practice in the participating units to help with weaning off mechanical ventilation. We kept the planned diff erence between groups constant. Therefore, our trial is a comparison of two diff erent ranges of hypercapnia, one higher than the other, particularly after the fi rst 3 days of life.

In view of these considerations, our results are surprising. Signifi cantly lower ventilator pressures in the high target group did not translate into better outcomes. Furthermore, the most important outcomes tended to favour the control group rather than the high target group, arguing against the benefi ts of a high target that might have been missed because of the smaller-than-intended sample size or the smaller-than-intended pCO₂ diff erence.

To explain the negative results, we hypothesised that only infants who were extremely ill benefi ted from the high target. The predefi ned subgroup analyses had been designed to address this hypothesis, but had low power owing to limited sample sizes. However, our results suggested that there was no benefi cial eff ect associated with the high target for even the smallest infants, consistent with subgroup analyses of the SUPPORT trial.17 Furthermore, in infants with more severe lung disease, the high target might have been associated with detrimental eff ects, according to our data. Similarly, a reanalysis of the SUPPORT trial showed associations between high pCO₂ values and adverse outcomes.38

Lung protective ventilation is a complex issue, and the pCO₂ is just one factor. Low ventilator pressure might be lung protective, or lead to increased lung injury if atelectasis formation is not prevented. However, the most important determinant is a low tidal volume,3 and a higher pCO₂ might be a surrogate marker for lower tidal volumes when other ventilatory parameters are kept constant. As shown in experimental models,39,40 additional lung protection can arise from the biochemical eff ects of higher pCO₂ and lower pH. On the other hand, high

pCO₂ and corresponding low pH values might also have harmful side-eff ects that outweigh any benefi ts, and this could explain why the high pCO₂ target was not associated with an improved outcome in this trial.5,41,42 Indeed, there were no pH limits, fi rstly because lower pH limits for premature infants have not been defi ned, and, second, the only way to correct a low pH while maintaining hypercapnia is by bicarbonate infusion, which itself can have negative side-eff ects22,23 and thus be a possible confounder. To prevent such a confounding eff ect and promote equal bicarbonate use in both study groups, we used bicarbonate administration to correct the pH when the base defi cit rather than the pH exceeded a limit. Consequently, the varying use of bicarbonate was not signifi cantly diff erent between the two study groups, but resulted in lower pH values in the high target group. Therefore, important enzymes could have been too far outside their optimum pH to function as needed in the high target group, which could have, along with other eff ects, impaired wound repair in injured lungs43,44 and alveolar fl uid clearance45 under hypercapnic conditions. In this case, the most severely injured lungs would be most aff ected, as suggested by the respective subgroup analysis. Similar eff ects of hypercapnic acidosis could also have impaired the function of intestinal cells and might have caused the signifi cantly higher rate of necrotising enterocolitis in the high target group, because an association between acidosis and NEC has already been described.46 Alternatively, the diff ering NEC rate might be an incidental fi nding because we did several univariate analyses.

As in general neonatal care, we used diff erent ventilation modes. Most of the newer modes have not been proven to be more benefi cial in rigorous large-scale clinical trials. Additionally, the measurement of tidal volume or synchronisation of ventilation with an in-line fl ow sensor at the expense of a higher dead space, possibly increasing the need for mechanical support, has not been shown to reduce lung injury and lead to better outcomes in extremely low birthweight infants. Therefore, some centres in our study group avoided using fl ow sensors (appendix p 7). The use of fl ow sensors did not have a signifi cant eff ect on the outcomes of this trial (data not shown).

We noted no increases of intraventricular haemorrhage or retinopathy of prematurity attributable to the high target. The rate of IVH might seem high, but only the less stable infants with the highest risk for IVH were eligible for this trial. Severe IVH (grade 3 or 4) might seem slightly more frequent in the high target group, but there were already a few more cases of severe IVH present on day 1—ie, before the protocol specifi cations had been fully enforced. Likewise, fi ndings of previous trials did not show an increased incidence of IVH associated with permissive hypercapnia.14–17,47 Data from previous retrospective analyses had already suggested that fl uctuations of pCO2 were more strongly associated with IVH than sustained high values.48,49

Articles

542 www.thelancet.com/respiratory Vol 3 July 2015

The main limitation of this trial is probably its premature cessation, reducing statistical power to answer the pertinent questions. Furthermore, the pCO₂ values were lower than intended in the high target group, which could be attributed either to the patient’s own respiratory drive, or to insuffi cient adherence to the protocol by the clinicians. The signifi cantly lower peak inspiratory pressure values in the high target group probably refl ect the clinicians’ attempts to minimise ventilation and let the pCO₂ rise in accordance with the protocol. Clinical decisions on choosing ventilator settings were not only driven by pCO₂ targets but tended to be more complex. Looking at the collected data on ventilator settings, these decisions did not always follow the study protocol. However, our trial represents a pragmatic comparison of two management strategies in daily clinical care.

In any case, all fi ndings should be interpreted with appropriate caution. Previous trials also failed to fully achieve pCO₂ targets in the high target groups,14–16 which suggests that the target ranges in the high target group were impossible to achieve in most infants most of the time and probably should not be attempted either clinically or in future studies.

In summary, we recorded no signifi cant diff erence in the primary outcome of bronchopulmonary dysplasia or death between the two pCO₂ target groups in infants requiring intubation and mechanical ventilation. We conclude that managing extremely low birthweight infants with pCO₂ targets according to the high target group, as compared with the control group, is not associated with improved outcomes. Moreover, the high target might be associated with an increased incidence of necrotising enterocolitis, and in infants with more severe lung disease, with an increased incidence of bronchopulmonary dysplasia or death, and therefore cannot be recommended.ContributorsUHT was the study coordinator and lead investigator and wrote the grant application and institutional review board application, developed and drafted the study protocol, co-developed the statistical analysis plan, recruited patients, gathered data, and wrote the manuscript, and generated fi gures; OG-B developed the study protocol, recruited patients, gathered data, and edited the manuscript; BB developed the study protocol, recruited patients, gathered data, edited the manuscript; MS, HF, OR, SA, DF, AZ, BK, HB, WS, HS, NT, and RH developed the study protocol, recruited patients, gathered data, and edited the manuscript; HF developed the study protocol, recruited patients, gathered data, and edited the manuscript; H-GT, KT, and CG recruited patients, gathered data, and edited the manuscript; MH, RS, JP, RR, and JS developed the study protocol, recruited patients, and edited the manuscript; WR and GH developed the study protocol, and edited the manuscript; RB recruited patients, gathered data, and edited the manuscript; MZ programmed the study database, managed data and queries, entered data into the database, and edited the manuscript; GA developed the study protocol, analysed all cranial ultrasound exams, selected chest radiographs, and edited the manuscript; JD developed the fi nal statistical analysis plan and the protocol amendment, carried out the statistical analyses, generated fi gures, and edited the manuscript; RM developed the study protocol, initial statistical plan, sample size calculation, and edited the manuscript; and HDH had the initial idea to do this study, edited the grant application, developed the study protocol, recruited patients, gathered data, and edited the manuscript.

Declaration of interestsWe declare no competing interests.

AcknowledgmentsThis trial was funded by the Deutsche Forschungsgemeinschaft (German Research Foundation, DFG, project number Th626/5-1), which is taxpayer funded. The trial underwent extensive review by anonymous expert reviewers before funding was approved by the funding agency’s review board. We thank the parents of our infants for their consent and support of this project, which was given at a very diffi cult time; all physicians and nurses who worked in the participating units, many without being directly involved in this project, for their support; and Waldemar A Carlo, Namasivayam Ambalavanan, and Frank Pohlandt for their invaluable advice in developing the study protocol and writing this manuscript; Sabine Schmid for entering data, and Evelyn Killick for language editing.

References1 Baraldi E, Filippone M. Chronic lung disease after premature birth.

N Engl J Med 2007; 357: 1946–55.2 Greenough A, Alexander J, Burgess S, et al. Home oxygen status and

rehospitalisation and primary care requirements of infants with chronic lung disease. Arch Dis Child Fetal Neonatal Ed 2002; 86: 40–43.

3 Dreyfuss D, Saumon G. Ventilator-induced lung injury: lessons from experimental studies. Am J Respir Crit Care Med 1998; 157: 294–323.

4 Bjoerklund LJ, Ingimarsson J, Curstedt T, et al. Manual ventilation with a few large breaths at birth compromises the therapeutic eff ect of subsequent surfactant replacement in immature lambs. Pediatr Res 1997; 42: 348–55.

5 Curley G, Laff ey JG, Kavanagh BP. Bench-to-bedside review: carbon dioxide. Crit Care 2010; 14: 220.

6 Kaiser JR, Gauss CH, Pont MM, Williams DK. Hypercapnia during the fi rst 3 days of life is associated with severe intraventricular hemorrhage in very low birth weight infants. J Perinatol 2006; 26: 279–85.

7 Holmes JM, Zhang S, Leske DA, Lanier WL. Carbon dioxide-induced retinopathy in the neonatal rat. Curr Eye Res 1998; 17: 608–16.

8 Kraybill EN, Runyan DK, Bose CL, Khan JH. Risk factors for chronic lung disease in infants with birth weights of 751 to 1000 grams. J Pediatr 1989; 115: 115–20.

9 Garland JS, Buck RK, Allred EN, Leviton A. Hypocarbia before surfactant therapy appears to increase bronchopulmonary dysplasia risk in infants with respiratory distress syndrome. Arch Pediatr Adolesc Med 1995; 149: 617–22.

10 Van Marter LJ, Allred EN, Pagano M, et al. Do clinical markers of barotrauma and oxygen toxicity explain interhospital variation in rates of chronic lung disease? The Neonatology Committee for the Developmental Network. Pediatrics 2000; 105: 1194–201.

11 Subramanian S, El-Mohandes A, Dhanireddy R, Koch MA. Association of bronchopulmonary dysplasia and hypercarbia in ventilated infants with birth weights of 500-1,499 g. Matern Health J 2011; 15 (suppl 1): S17–26.

12 The Acute Respiratory Distress Syndrome Network. Ventilation with lower tidal volumes as compared with traditional tidal volumes for acute lung injury and the acute respiratory distress syndrome. N Engl J Med 2000; 342: 1301–08.

13 Fischer HS, Bührer C. Avoiding endotracheal ventilation to prevent bronchopulmonary dysplasia: a meta-analysis. Pediatrics 2013; 132: e1351–60.

14 Mariani G, Cifuentes J, Carlo WA. Randomized trial of permissive hypercapnia in preterm infants. Pediatrics 1999; 104: 1082–88.

15 Thome UH, Carroll WF, Wu T-J, et al. Outcome of extremely preterm infants randomised at birth to diff erent PaCO2 targets during the fi rst seven days of life. Biol Neonate 2006; 90: 218–25.

16 Carlo WA, Stark AR, Wright LL, et al. Minimal ventilation to prevent bronchopulmonary dysplasia in extremely low birthweight infants. J Pediatr 2002; 141: 370–74.

17 Finer NN, Carlo WA, Walsh MC, et al. Early CPAP versus surfactant in extremely preterm infants. N Engl J Med 2010; 362: 1970–9.

18 Woodgate PG, Davies MW. Permissive hypercapnia for the prevention of morbidity and mortality in mechanically ventilated newborn infants. Cochrane Database Syst Rev 2001; 2: CD002061.

Articles

www.thelancet.com/respiratory Vol 3 July 2015 543

19 Hermeto F, Bottino MN, Vaillancourt K, Sant’Anna GM. Implementation of a respiratory therapist-driven protocol for neonatal ventilation: impact on the premature population. Pediatrics 2009; 123: e907–16.

20 Van Kaam AH, De Jaegere AP, Rimensberger PC, on behalf of the Neovent Study Group. Incidence of hypo- and hyper-capnia in a cross-sectional European cohort of ventilated newborn infants. Arch Dis Child Fetal Neonatal Ed 2013; 98: F323–26.

21 Ehrenkranz RA, Walsh MC, Vohr BR, et al. Validation of the National Institutes of Health consensus defi nition of bronchopulmonary dysplasia. Pediatrics 2005; 116: 1353–60.

22 Laff ey JG, Engelberts D, Kavanagh BP. Buff ering hypercapnic acidosis worsens acute lung injury. Am J Respir Crit Care Med 2000; 161: 141–46.

23 Higgins BD, Costello J, Contreras M, Hassett P, O’Toole D, Laff ey JG. Diff erential eff ects of buff ered hypercapnia versus hypercapnic acidosis on shock and lung injury induced by systemic sepsis. Anesthesiology 2009; 111: 1317–26.

24 Walsh MC, Wilson-Costello D, Zadell A, Newman N, Fanaroff A. Safety, reliability, and validity of a physiologic defi nition of bronchopulmonary dysplasia. J Perinatol 2003; 23: 451–56.

25 Papile LA, Burstein J, Burstein R, Koffl er H. Incidence and evolution of subependymal and intraventricular hemorrhage: a study of infants with birth weights less than 1,500 gm. J Pediatr 1978; 92: 529–34.

26 The International Classifi cation of Retinopathy of Prematurity revisited. Arch Ophthalmol 2005; 123: 991–99.

27 Bell MJ, Ternberg JL, Feigin RD, et al. Neonatal necrotizing enterocolitis. Therapeutic decisions based upon clinical staging. AnnSurg 1978; 187: 1–7.

28 Pampallona S, Tsiatis AA. Group sequential designs for one-sided and two-sided hypothesis testing with provision for early stopping in favor of the null hypothesis. J Stat Plan Inference 1994; 42: 19–35.

29 Wang R, Lagakos SW, Ware JH, Hunter DJ, Drazen JM. Statistics in medicine–reporting of subgroup analyses in clinical trials. N Engl J Med 2007; 357: 2189–94.

30 Zou G. A modifi ed poisson regression approach to prospective studies with binary data. Am J Epidemiol 2004; 159: 702–06.

31 Voigt M, Rochow N, Straube S, Briese V, Olbertz D, Jorch G. Birth weight percentile charts based on daily measurements for very preterm male and female infants at the age of 154-223 days. J Perinat Med 2010; 38: 289–95.

32 Jensen EA, Schmidt B. Epidemiology of bronchopulmonary dysplasia. Birth Defects Res A Clin Mol Teratol 2014; 100: 145–57.

33 Dargaville PA, Aiyappan A, Cornelius A, Williams C, De Paoli AG. Preliminary evaluation of a new technique of minimally invasive surfactant therapy. Arch Dis Child Fetal Neonatal Ed 2011; 96: F243–8.

34 Vendettuoli V, Bellu R, Zanini R, Mosca F, Gagliardi L, for the Italian Neonatal Network. Changes in ventilator strategies and outcomes in preterm infants. Arch Dis Child Fetal Neonatal Ed 2014; 99: F321–24.

35 Thome UH, Ambalavanan N. Permissive hypercapnia to decrease lung injury in ventilated preterm neonates. Semin Neonatal Med 2009; 14: 21–27.

36 Laff ey JG, Kavanagh BP. Hypocapnia. N Engl J Med 2002; 347: 43–53.

37 Fuchs H, Mendler MR, Scharnbeck D, Ebsen M, Hummler HD. Very low tidal volume ventilation with associated hypercapnia–eff ects on lung injury in a model for acute respiratory distress syndrome. PloS One 2011; 6: e23816.

38 Ambalavanan N, Carlo WA, Wrage LA, et al. PaCO2 in Surfactant, Positive Pressure, and Oxygenation Randomised Trial (SUPPORT). Arch Dis Child Fetal Neonatal Ed 2014; published online Nov 25. http://fn.bmj.com/content/early/2014/11/25/archdischild-2014-306802.abstract.

39 Strand M, Ikegami M, Jobe AH. Eff ects of high pCO2 on ventilated preterm lamb lungs. Pediatr Res 2003; 53: 468–72.

40 Laff ey JG, Honan D, Hopkins N, Hyvelin JM, Boylan JF, McLoughlin P. Hypercapnic acidosis attenuates endotoxin-induced acute lung injury. Am J Respir Crit Care Med 2004; 169: 46–56.

41 Ryu J, Heldt GP, Nguyen M, Gavrialov O, Haddad GG. Chronic hypercapnia alters lung matrix composition in mouse pups. J Appl Physiol 2010; 109: 203–10.

42 Vadasz I, Hubmayr RD, Nin N, Sporn PH, Sznajder JI. Hypercapnia: a nonpermissive environment for the lung. Am J Respir Cell Mol Biol 2012; 46: 417–21.

43 Doerr CH, Gajic O, Berrios JC, et al. Hypercapnic acidosis impairs plasma membrane wound resealing in ventilator-injured lungs. Am J Respir Crit Care Med 2005; 171: 1371–77.

44 Lang JD, Figueroa M, Sanders KD, et al. Hypercapnia via reduced rate and tidal volume contributes to lipopolysaccharide-induced lung injury. Am J Respir Crit Care Med 2005; 171: 147–57.

45 Briva A, Vadasz I, Lecuona E, et al. High CO2 levels impair alveolar epithelial function independently of pH. PLoS One 2007; 2: e1238.

46 Gephart SM, Spitzer AR, Eff ken JA, Dodd E, Halpern M, McGrath JM. Discrimination of GutCheckNEC: a clinical risk index for necrotizing enterocolitis. J Perinatol 2014; 34: 468–75.

47 Hagen EW, Sadek-Badawi M, Carlton DP, Palta M. Permissive hypercapnia and risk for brain injury and developmental impairment. Pediatrics 2008; 122: e583–89.

48 Fabres J, Carlo WA, Phillips V, Howard G, Ambalavanan N. Both extremes of arterial carbon dioxide pressure and the magnitude of fl uctuations in arterial carbon dioxide pressure are associated with severe intraventricular hemorrhage in preterm infants. Pediatrics 2007; 119: 299–305.

49 McKee LA, Fabres J, Howard G, Peralta-Carcelen M, Carlo WA, Ambalavanan N. PaCO2 and neurodevelopment in extremely low birth weight infants. J Pediatr 2009; 155: 217–21.

Comment

www.thelancet.com/respiratory Vol 3 July 2015 499

During the past decade the threshold of viability in extremely preterm infants has shifted to the lower gestational age of younger than 26 weeks postmenstrual age with subsequent improved survival. As a consequence, more of the survivors have to be treated for typical co-morbidities of extreme prematurity such as brochopulmonary dysplasia. The latter is associated with extreme prematurity and ventilator-induced lung injury. Data from small studies have suggested that permissive hypercapnia might reduce the incidence of lung injury, and as a result the German multicentre PHELBI study group1 embarked on a large randomised

controlled study of 830 extremely low birthweight infants testing two levels of partial pressures of carbon dioxide (PCO2) during the fi rst 2 weeks of life. The study was stopped prematurely after an interim analysis of 359 (23%) of 1534 infants screened in 53 months, in which investigators noted no diff erence in the primary outcome of death or moderate to severe bronchopulmonary dysplasia between groups.

We would like to commend the authors and the journal for publishing the no diff erence fi ndings as the report contains lessons to be learned. The study included preterm infants from 23 to 28 weeks gestation onwards.

Arg117His-CFTR status has direct relevance to newborn screening programmes for cystic fi brosis. Inclusion of Arg117His-CFTR in screening panels leads to identifi cation of infants who have a very low likelihood of developing cystic fi brosis lung disease, at least during childhood. Removing the Arg117His-CFTR mutation from newborn screening panels has been advocated.9 This position will need to be reassessed in view of Moss and colleagues’ results,6 since ivacaftor will probably play an important part in treating Arg117His/5T-CFTR and might well be of benefi t in those few patients with Arg117His/7T-CFTR who have lung disease.

These data6 are therefore an important contribution to the medical literature, since they provide clear support for the treatment of patients with Arg117His-CFTR. The results also show CFTR activity in children as well as in adults, providing a rationale for the treatment of some children. Whether patient-specifi c response to ivacaftor can be predicted from specimens studied ex vivo—eg, nasal epithelial cells or intestinal organoids—remains to be clarifi ed.10 Long-term studies to assess the safety and effi cacy of ivacaftor in general are also needed. In this respect, evidence that the drug can modify the rate of decline in lung function or the development of structural injury will be important. Finally, the results from Moss and colleagues’ study reinforce the notion of mutation-specifi c treatment of CFTR dysfunction—an important step on the road to personalised care for all individuals with cystic fi brosis.

Permissive hypercapnia in preterm infants: the discussion continues

Published OnlineJune 16, 2015http://dx.doi.org/10.1016/S2213-2600(15)00240-4

See Articles page 534

Frank J AccursoUniversity of Colorado and Children’s Hospital Colorado, Aurora, CO 80045, USAFrank.Accurso@ucdenver.edu

I have served on the Cystic Fibrosis Foundation Therapeutics, Inc./Vertex Pharmaceuticals, Inc. Joint Development Committee since 2003. The purpose of this committee is to improve the communication between the two groups. I represent Cystic Fibrosis Foundation Therapeutics, Inc. in these discussions and have no ties, fi nancial or otherwise, with Vertex Pharmaceuticals, Inc. I receive reimbursement from Cystic Fibrosis Foundation Therapeutics, Inc. for the twice a year in-person committee meetings only for plane fare, lunch, and parking. I do not receive any remuneration from Cystic Fibrosis Foundation Therapeutics, Inc. for time or eff ort. I receive no remuneration at all from Vertex Pharmaceuticals, Inc.

1 Rowe SM, Miller S, Sorscher EJ. Cystic fi brosis. N Engl J Med 2005; 352: 1992–2001.

2 Stoltz DA, Meyerholz DK, Welsh MJ. Origins of cystic fi brosis lung disease. N Engl J Med 2015; 372: 1574–75.

3 Van Goor F, Hadida S, Grootenhuis, et al. Rescue of CF airway epithelial cell function in vitro by a CFTR potentiator, VX-770. Proc Natl Acad Sci USA 2009; 106: 18825–30.

4 Ramsey BW, Davies J, McElvaney NG, et al. VX08-770-102 Study Group. A CFTR potentiator in patients with cystic fi brosis and the G551D mutation. N Engl J Med 2011; 365: 1663–72.

5 De Boeck K, Munck A, Walker S, et al. Effi cacy and safety of ivacaftor in patients with cystic fi brosis and a non-G551D gating mutation. J Cyst Fibros 2014; 13: 674–80.

6 Moss RB, Flume PA, Elborn JS, et al, on behalf of the VX11-770-110 (KONDUCT) Study Group. Effi cacy and safety of ivacaftor in patients with cystic fi brosis who have an Arg117His-CFTR mutation: a double-blind, randomised controlled trial. Lancet Respir Med 2015; 3: 524–33.

7 Kiesewetter S, Macek M Jr, Davis C, et al. A mutation in CFTR produces diff erent phenotypes depending on chromosomal background. Nat Genet 1993; 5: 274–78.

8 Witt DR, Schaefer C, Hallam P, et al. Cystic fi brosis heterozygote screening in 5161 pregnant women. Am J Hum Genet 1996; 58: 823–35.

9 Thauvin-Robinet C, Munck A, Huet F, et al. The very low penetrance of cystic fi brosis for the R117H mutation: a reappraisal for genetic counselling and newborn screening. J Med Genet 2009; 46: 752–58.

10 Dekkers JF, van der Ent CK, Beekman JM. Novel opportunities for CFTR-targeting drug development using organoids. Rare Dis 2013; 1: e27112.

Comment

500 www.thelancet.com/respiratory Vol 3 July 2015

The incidence of bronchopulmonary dysplasia is highest in infants younger than 28 weeks and therefore this was an appropriate target population. Of note is the long duration of recruitment of more than 4 years. Clinical practice might have changed during this time especially with the increasing use of non-invasive means of respiratory support to wean or avoid intratracheal ventilation altogether.2 During non-invasive ventilator support PCO2 concentrations are established by the breathing eff orts of newborn babies and clinicians are less able to aff ect the concentrations directly. Future studies will need to consider this growing group of patients and would benefi t from new non-invasive techniques such as plethysmogram analysis of pulse oximetry traces to measure spontaneous breathing rates together with PCO2 concentrations.3

In the PHELBI study, PCO2 target concentrations were aimed at three increasing levels in the fi rst 14 days of life.1 The study design might have inadvertently aff ected the clinician’s decision to ventilate newborn babies to achieve the PCO2 targets during the study period. It was not possible to mask the clinicians to group allocations. After 14 days, 25% of infants were still ventilated. In the present trend of neonatal clinical care to wean to non-invasive ventilation as soon as possible, many tertiary care centres would consider this intubation rate to be high.2

The PHELBI study group targeted the lung injury aspect of bronchopulmonary dysplasia by using low tidal volume ventilation strategies without controlling

for it with volume-targeted ventilation methods that help to reduce lung injury. Results of recent animal studies have also shown that lung injury starts early at resuscitation at birth and could be improved by allowing redistribution of placental blood through delaying cord cutting and initiating lung expansion fi rst.4 Findings of the study by Polglase and colleagues4 in preterm lambs showed a smoother transition to extrauterine life after birth if the lambs were given infl ation breath with their cord intact. This method enabled a smooth increase in pulmonary blood fl ow with expansion of lung alveoli.4 Benefi ts of redistribution of placental blood in preterm infants such as better adaptation after birth, less need for blood transfusion, less incidence of intraventricular haemorrhage and necrotising enterocolitis have been widely described and therefore have been incorporated into international guidelines on newborn resuscitation.5–8

The PHELBI study group did not collect information about whether the recruited infants received stand-ardised delivery room management and sub sequent stabilisation to minimise lung injury, including any means of redistribution of placental blood before enrolment into the study. Future studies of the reduction of lung injury and the development of bronchopulmonary dysplasia should include this aspect of resuscitation of the preterm infant. This approach would need antenatal consent to enrol the preterm infant into the trial. Findings of a recently published qualitative study of parents showed a positive attitude towards enrolling their unborn preterm baby into a randomised trial to study two diff erent ways of enhancing redistribution of placental blood at birth.9,10 Increasing data seem to suggest that the lower the gestational age the more varying practices between hospitals will aff ect the outcome of preterm infants.11 With the participation of many centres in increasingly larger randomised controlled trials over a long time (like the PHELBI study), accounting for comparative eff ectiveness between centres (ie, alternative standards of care, assessing outcomes important to individuals, and incorporating varied settings and participants)becomes an impor tant part of outcome assessment.12

Much still needs to be learnt about lung injury and the development of bronchopulmonary dysplasia in extremely low birthweight infants. The pattern of respiratory distress syndrome in these infants has

Roge

r Job

/Rep

orte

rs/S

cienc

e Ph

oto

Libr

ary

Comment

www.thelancet.com/respiratory Vol 3 July 2015 501

About 5–10% of patients on mechanical ventilation will have persistent respiratory failure necessitating prolonged mechanical ventilation.1 This condition is part of the larger syndrome of chronic critical illness, in which critically ill patients have continuing organ failures leading to protracted periods of organ support.2 Both prolonged mechanical ventilation and chronic critical illness are important public health problems. In the USA, for example, more than 380 000 individuals are estimated to have chronic critical illness every year, with costs exceeding US$50 billion annually.3

A meta-analysis of studies reporting clinical outcomes in prolonged mechanical ventilation published in The Lancet Respiratory Medicine further highlights the challenges posed by these patients.4 Emily Damuth and colleagues4 systematically reviewed 124 studies from 16 diff erent countries worldwide, with sobering results. Pooled mortality at hospital discharge was 26%, 57% of

patients were liberated from mechanical ventilation by the end of their hospital stay, and 22% of patients were discharged home. Perhaps most concerning was that 59% of patients were dead at 1 year.

These fi ndings should serve as a wake-up call to clinicians, hospital administrators, and health policy makers involved in the care of patients with prolonged mechanical ventilation. Eff orts are urgently needed to increase liberation rates and improve survival. Unfortunately, the way to proceed in this area is not at all clear. Despite the increasing recognition of the immense clinical and fi nancial burden chronic critical illness puts on health systems, evidence-based strategies to guide clinical care in this population are lacking. This situation is the real wake-up call—not that outcomes must be improved, but that how to improve them is simply not known. Immediate, targeted research is needed to fi ll this knowledge gap

Improving outcomes in prolonged mechanical ventilation: a road map

Published OnlineMay 21, 2015http://dx.doi.org/10.1016/S2213-2600(15)00205-2

See Articles page 544

Jade

/Ble

nd Im

ages

/Cor

bis

changed over the past 10 years. The introduction of antenatal steroids has changed the incidence of severe respiratory distress syndrome. Researchers can now look at other preventive measures at the time of resuscitation at birth,5 which might have more eff ect than PCO2 concentrations in the fi rst day of life or diff erent ventilation strategies such as low volume ventilation or patient synchronised ventilation. The PHELBI study provides the basis for further discussions on how to design future study protocols. Optional antenatal consent will probably increase recruitment in trials of extremely low birthweight infants of 23 to 28 weeks gestation.

Heike Rabe, Jose Ramon Fernandez-AlvarezBrighton and Sussex Medical School and University Hospitals, Academic Department of Paediatrics, Royal Alexandra Children’s Hospital, Brighton BN2 5BE, UKHeike.Rabe@bsuh.nhs.uk

We declare no competing interests.

1 Thome UH, Genzel-Boroviczeny O, Bohnhorst B, et al, for the PHELBI Study Group. Permissive hypercapnia in extremely low birthweight infants (PHELBI): a randomised controlled multicentre trial. Lancet Resp Med 2015; published online June 16. http://dx.doi.org/10.1016/S2213-2600(15)00204-0.

2 Fernandez-Alvarez R, Gandhi RS, Amess PN, et al. Heated humidifi ed high fl ow nasal cannula versus low fl ow nasal cannula as weaning mode from nasal CPAP in infants ≤28 weeks gestation. Eur J Pediatr 2014; 173: 93–98.

3 Wertheim D, Olden C, Symes L, et al. Monitoring respiration in wheezy preschool children by pulse oximetry plethysmogram analysis. Med Biol Eng Comput 2013; 51: 965–70.

4 Polglase GR, Dawson JA, Kluckow M, et al. Ventilation onset prior to umbilical cord clamping (physiological-based cord clamping) improves systemic and cerebral oxygenation in preterm lambs. PLoS One 2015; 10: e0117504.

5 Sweet D, Carnielli V, Greisen G, et al. European Consensus Guidelines on the management of neonatal respiratory distress syndrome in preterm infants: 2010 Update. Neonatology 2010; 97: 402–17.

6 The American College of Obstetricians and Gynecologists. Committee on Obstetric Practice. Timing of umbilical cord clamping after birth. Obstet Gynecol 2012; 120: 1522–26.

7 The National Institute for Health and Care Excellence. Intrapartum care: care of healthy women and their babies during childbirth. December, 2014. http://www.nice.org.uk/guidance/cg190 (accessed Feb 25, 2015).

8 WHO. Every Newborn: an action plan to end preventable deaths. http://www.everynewborn.org/Documents/Full-action-plan-EN.pdf (accessed March 10, 2015).

9 Ayers S, Sawyers A, During C, et al. Parents report positive experiences about enrolling babies in a cord-related trial before birth. Acta Pediatr 2015; 104: e164–e170.

10 Rabe H, Jewison A, Fernandez Alvarez R, et al. Milking compared with delayed cord clamping to increase placental transfusion in preterm neonates. A randomized controlled trial. Obstet & Gynecol 2011; 117: 205–11.

11 Rysavy MA, Li L, Bell EF, et al. Between-hospital variation in treatment and outcomes in extremely preterm Infants. New Engl J Med 2015; 372: 1801–11.

12 Lagatta J, Uhing M, Panepinto J. Comparative eff ectiveness and practice variation in neonatal care. Clin Perinatol 2014; 41: 833–45.