Late permissive hypercapnia and respiratory stability among very preterm infants: a pilot randomised trial

Abstract

Objective

Determine if targeting higher transcutaneous carbon dioxide improves respiratory stability among very preterm infants on ventilatory support.

Design

Single-centre pilot randomised clinical trial.

Setting

The University of Alabama at Birmingham.

Patients

Very preterm infants on ventilatory support after postnatal day 7.

Interventions

Infants were randomised to two different transcutaneous carbon dioxide levels targeting 5 mm Hg (0.67 kPa) changes with four sessions each lasting 24 hours for 96 hours: baseline-increase-baseline-increase or baseline-decrease-baseline-decrease.

Main outcome measures

We collected cardiorespiratory data evaluating episodes of intermittent hypoxaemia (oxygen saturations (SpO₂)<85% for ≥10 s), bradycardia (<100 bpm for ≥10 s), and cerebral and abdominal hypoxaemia on near-infrared spectroscopy.

Results

We enrolled 25 infants with a gestational age of 24 w 6 d±11 d (mean±SD) and birth weight 645±142 g on postnatal day 14±3. Continuous transcutaneous carbon dioxide values (56.8±6.9 in the higher group vs 54.5±7.8 in the lower group; p=0.36) did not differ significantly between groups during the intervention days. There were no differences in intermittent hypoxaemia (126±64 vs 105±61 per 24 hours; p=0.30) or bradycardia (11±16 vs 15±23 per hour; p=0.89) episodes between groups. The proportion of time with SpO₂<85%, SpO₂<80%, cerebral hypoxaemia or abdominal hypoxaemia did not differ (all p>0.05). There was moderate negative correlation between mean transcutaneous carbon dioxide and bradycardia episodes (r=−0.56; p<0.001).

Conclusion

Targeting 5 mm Hg (0.67 kPa) changes in transcutaneous carbon dioxide did not improve respiratory stability among very preterm infants on ventilatory support but the intended carbon dioxide separation was difficult to achieve and maintain.

Trial registration number

NCT03333161.

INTRODUCTION

Intentionally reducing the intensity of mechanical ventilation while accepting a higher partial pressure of carbon dioxide (PaCO₂) is a commonly used lung-protective strategy in the neonatal intensive care unit.^1–3 As well as permitting a lower intensity of mechanical ventilation, permissive hypercapnia may increase respiratory drive.⁴ It has been demonstrated that CO₂ inhalation decreases apnoeic episodes among preterm infants.⁵ In addition, apnoea may be more likely to occur among infants breathing closer to their apnoeic CO₂ threshold, which is higher among preterm infants.^6–8 Targeting PaCO₂ above the apnoeic threshold among preterm infants on positive pressure respiratory support increases respiratory muscle activity/drive of the diaphragm and accessory respiratory muscles⁶ and may decrease apnoea and resultant intermittent hypoxaemia.

Alternatively, decreasing respiratory support and alveolar ventilation to achieve hypercapnia may result in respiratory instability among infants with respiratory failure. Apnoea of prematurity is more frequent among infants with altered chemoreceptor responses including those with lower alveolar ventilation, higher alveolar CO₂ and a decreased respiratory responsiveness to CO₂.⁷ The objective of this study was to test the hypothesis that targeting a higher transcutaneous CO₂ (TcCO₂), a surrogate marker of PaCO₂, compared with a lower TcCO₂ decreases apnoea, bradycardia and intermittent hypoxaemia and improves respiratory stability among very preterm infants on positive pressure respiratory support.

METHODS

Trial design, setting and participants

This single-centre trial was conducted at The University of Alabama at Birmingham. Preterm infants from 22 weeks and 0 days to 28 weeks and 6 days of gestational age and with a birth weight from 401 to 1000 g who were already enrolled in the multicenter Pre-Vent Apnoea observational study ( Clinicaltrials.gov identifier: NCT03174301) were screened for eligibility. We included infants if they remained intubated or on nasal intermittent positive pressure ventilation (NIPPV) after postnatal day 7 with a baseline pH>7.25 and PaCO₂>40 mm Hg (5.3 kPa) and had a TcCO₂ monitor that was correlated within 5 mm Hg (0.67 kPa) of the PaCO₂. We excluded infants judged too unstable by the clinician, infants with major congenital malformations, infants with a terminal illness or decision to withdraw or limit care, and infants whose parents had refused or withdrawn informed consent.

Recruitment and randomisation

After informed consent was obtained, infants were randomised using a 1:1 parallel allocation to one of the two intervention sequences. Randomisation order was computer generated with block sizes of two to four by NA. The allocation was concealed using sequentially numbered opaque sealed envelopes until enrolment by DL.

Interventions

Infants were randomised to two different levels of permissive hypercapnia targeting 5 mm Hg (0.67 kPa) changes from baseline TcCO₂ in one of two sequences with four sessions lasting 24 hours each, for a total of 96 hours. The first 24 hours of data collection was used to establish the baseline mean TcCO₂ (0–24 hours). Over the next 3 days, we evaluated interventions as follows; either targeted to increase (24–48 hours) followed by return to baseline (48–72 hours), followed by increase (72–96 hours), referred to as the higher group, or decrease-baseline-decrease referred to as the lower group. For example, infants randomised to increase-baseline-increase initially had their TcCO ₂ limits (based on correlation with daily capillary blood gas) increased to target 5 mm Hg (0.67 kPa) above baseline by reducing the intensity of mechanical ventilation for 24 hours. This was followed by a decrease in TcCO₂ limits to target a 5 mm Hg (0.67 kPa) reduction (back to baseline) by increasing the intensity of mechanical ventilation for the next 24 hours. Then, we increased the TcCO₂ limits to target a 5 mm Hg (0.67 kPa) above baseline by reducing the intensity of mechanical ventilation for 24 hours. Ventilation could be adjusted routinely every 4 hours or more frequently if clinically indicated but changes were not mandated. We recommended decreasing pressure/volume before the rate to increase the TcCO₂ while the rate was increased before pressure/volume to lower the TcCO₂.

These intervention limits were chosen to allow us to determine the effects of altering PaCO₂ on respiratory patterns while maintaining the pH at or above 7.20. TcCO₂ alarm limits used an upper limit set 15 points above the lower limit during the intervention period. Study team members updated the TcCO₂ monitor alarm limits to target the intended PaCO₂ and informed members of the clinical care team of the target range for each 24-hour period. Both groups were connected to dedicated study monitors (Philips IntelliVue MP50), which monitored heart rate (HR), respiratory rate (RR) and oxygen saturation (SpO₂), and INVOS 5100C Cerebral/Somatic Oximeters, which measured cerebral and abdominal near-infrared spectroscopy (NIRS) using infant/neonatal sensors (Medtronic, Dublin, Ireland). TcCO₂ was measured using the SenTec Digital Monitoring System (SenTec Inc, Fenton, Missouri, USA). The study monitors were connected to the central alarm system during the 96-hour study period. Masking of the intervention was not performed as TcCO₂ monitors are used in routine clinical care among ventilated infants in our neonatal intensive care unit. However, the outcome assessors (AN and IA) were masked to the study group assignment.

Measures

We collected real-time cardiorespiratory data using ixTrend (ixitos GmbH, Berlin, Germany) from the study monitors.⁹ The primary outcome measure was the number of episodes of intermittent hypoxaemia (SpO₂<85% for ≥10 s) per 24 hours. Secondary outcome measures included the number of episodes of severe intermittent hypoxaemia (SpO₂<80% for 10 s) per 24 hours, the number of episodes of bradycardia (HR<100 bpm for ≥10 s) per 24 hours and the proportion of time with cerebral (<55%) and abdominal hypoxaemia (<40%).

Statistical analysis

A sample size of 24 infants was planned a priori for this pilot randomised clinical trial.^{10 11} We planned to enrol up to 35 patients if needed to achieve the planned sample size to account for attrition due to incomplete data collection. A data and safety monitoring board reviewed all adverse event data from this randomised trial at 6 monthly intervals or sooner for predefined major adverse events.

We used the intention-to-treat principle to compare primary and secondary outcomes. We analysed numeric data using MATLAB (MathWorks, Natick, Massachusetts, USA) and used SAS V.9.4 for all statistical analyses. To compare baseline characteristics between infants on higher TcCO₂ and lower TcCO₂, we used Wilcoxon test for count and continuous variables and Fisher’s exact test for categorical variables. For data collected by session, we fitted a generalised linear mixed model with random intercept to account for the repeated measures for the same subject over time using gamma distribution for continuous outcomes being modelled, negative binomial for count outcomes and beta distribution for proportions. The main goal was to test for group effects. However, we first checked if there was evidence that the group effects changed over time by fitting models with group, time and group by time interaction. Since the interaction term was not significant (indicating no evidence of group effect, if any, changing over time), the final random intercept models fitted only included group effect. We used Pearson’s correlation to determine the strength and direction of the relationship between outcome measures. A sensitivity analysis was performed removing data from days when infants received non-invasive positive pressure ventilation. No adjustment for multiple testing was done. A p value of <0.05 was considered statistically significant.

RESULTS

We randomised 29 very preterm infants (figure 1). Fifteen infants were randomised to higher TcCO₂ and 14 were randomised to lower TcCO₂. Three infants randomised to higher TcCO₂ and one infant randomised to the lower TcCO₂ group were excluded from the analysis as insufficient data were available for comparison (figure 1). Three of these infants were placed on continuous positive airway pressure and one was transferred to another facility during the intervention. We analysed data from the remaining 25 infants with sufficient data for comparison. Infants were studied on postnatal day 14±3 and had a gestational age of 24 weeks and 6±10 days (mean+SD) and birth weight 645±142 g (table 1). There were no differences in clinical characteristics, baseline ventilator parameters or baseline blood gas values.

Figure 1.

Flow diagram showing the number of infants screened, randomised and analysed. CPAP, continuous positive airway pressure.

Table 1.

Baseline characteristics and clinical demographics of study participants

	Higher TcCO2 (N=12)	Lower TcCO2 (N=13)
Gestational age, weeks and days, mean±SD (days)	25 0/7±14/7	24 5/7±8/7
Birth weight, grams, mean±SD	654±167	637±120
Days after birth at study entry, days±SD	14±3	14±3
Male sex, no (%)	8 (67)	6 (46)
Race
Black, no (%)	10 (83)	7 (54)
White, no (%)	2 (17)	6 (46)
Antenatal steroids, no (%)	12 (100)	13 (100)
Histological chorioamnionitis, no (%)	6 (50)	7 (54)
Caesarean section delivery, no (%)	10 ((83)	10 (77)
Surfactant, no (%)	11 (92)	13 (100)
Days ventilated before study entry, days, median (range)	12 (6–15)	13 (5–17)
Mean airway pressure (cmH2O), mean±SD	8±1	8±1
Positive end-expiratory pressure (cmH2O), mean±SD	5±1	5±1
Ventilator rate (breaths per minute),* mean±SD	42±23	31±13
FiO2, mean±SD	49±16	38±13
pH, mean (SD)	7.31±0.07	7.30±0.05
PaCO2 (mm Hg), mean (SD)	55±8	53±7
HCO3 (mmol/L), mean (SD)	27±4	26±4

^*Excluded high-frequency ventilation rate.

The number of days on invasive mechanical ventilation or high-frequency ventilation during the intervention did not differ (table 2). There was a difference in the change in level of support (changing between NIPPV, invasive mechanical ventilation and high-frequency ventilation) between groups with four infants in the higher TcCO₂ group escalating support and two infants in the lower TcCO₂ group deescalating support during the study period.

Table 2.

Respiratory support during the study period and blood gas data by permissive hypercapnia target group^*

	Higher TcCO2 (N=12)	Lower TcCO2 (N=13)	P value
Number of days on Invasive ventilation, n (%)	12 (48)	13 (52)	0.99†
0	2 (17)	2 (15)
1	1 (8)	1 (8)
2	9 (75)	10 (77)
Number of days on high frequency, days (%)	12 (48)	13 (52)	0.220†
0	10 (83)	13 (100)
1	1 (8)	0 (0)
2	1 (8)	0 (0)
Change in support during study period, n (%)			0.03†
Escalated support	4 (33)	0 (0)
Deescalated support	0 (0)	2 (15)
No change	8 (67)	11 (85)
Mean airway pressure (cmH2O), mean±SD	9±2	8±2	0.29‡
Positive end-expiratory pressure (cmH2O), mean±SD	5±1	5±1	0.93‡
Ventilator rate (breaths per minute),§ mean±SD	45±21	37±21	0.40‡
FiO2, mean±SD	48±24	44±21	0.65‡
pH, mean±SD	7.29±0.08	7.30±0.07	0.51‡
PaCO2 (mm Hg), mean±SD	54±7	52±9	0.49‡
HCO3 (mmol/L), mean±SD	26±4	26±4	0.74‡

^*Observed mean and SD.

^†Fisher’s exact test.

^‡P values based on random effects mixed model with only group effect based on gamma distribution.

^§Excluded high-frequency ventilation rate.

Based on the repeated measure models, ventilator and blood gas parameters did not differ between groups during the intervention period (table 2). There were no differences in episodes of intermittent hypoxaemia (126±64 vs 105±61 per 24 hours; p=0.30), severe intermittent hypoxaemia (42±34 vs 37±25 per 24 hours; p=0.74) or bradycardia (11±16 vs 15±23 per 24 hours; p=0.89) between higher and lower TcCO₂ target groups during the intervention days (table 3). The proportion of time with SpO₂ <85% (8%±7% vs 7%±7%; p=0.47), SpO₂ <80% (3%±3% vs 2%±2%; p=0.65), cerebral hypoxaemia (25%±28% vs 21%±30%; p=0.61) or abdominal hypoxaemia (16%±21% vs 15%±24%; p=0.93) did not differ between groups during the intervention days. Continuous TcCO₂ values (56.8±6.9 in the higher group vs 54.5±7.8 in the lower group; p=0.36) did not differ significantly between groups during the intervention days.

Table 3.

Outcome measures by permissive hypercapnia target group during intervention days

	Higher TcCO2 target	Lower TcCO2 target	P value
No of episodes of intermittent hypoxaemia per 24 hours, mean±SD	126±64	105±61	0.30*
Percentage of time SpO2<85%, mean±SD	8±7	7±7	0.47*
No of episodes of severe intermittent hypoxaemia per 24 hours, mean±SD	42±34	37±25	0.74†
Percentage of time SpO2<80%, mean±SD	3±3	2±2	0.65‡
No of episodes of bradycardia per 24 hours, mean±SD	11±16	15±23	0.89†
Percentage of time cerebral near-infrared spectroscopy <55%, mean±SD	25±28	21±30	0.61‡
Percentage of time abdominal near-infrared spectroscopy <40%, mean±SD	16±21	15±24	0.93‡
TcCO2 (mm Hg), mean±SD	56.8±6.9	54.5±7.8	0.36*

^*P values based on random effects mixed model with only group effect based on gamma distribution.

^†P values based on random effects mixed model with only group effect based on negative binomial distribution.

^‡P values based on random effects mixed model with only group effect based on beta distribution.

Overall, there was moderate negative correlation between mean TcCO₂ and episodes of bradycardia (r=−0.56; p<0.001) (figure 2), but no correlation was noted between TcCO₂ and intermittent hypoxaemia or severe intermittent hypoxaemia (online supplemental figure S1). There was weak positive correlation between mean TcCO₂ and the percentage of time with cerebral (r=0.26; p=0.009) and abdominal hypoxaemia (r=0.27; p=0.009) (online supplemental figure S1). There was also a positive correlation between mean TcCO₂ and FiO₂ (r=0.36; p<0.001), mean TcCO₂ and mean airway pressure (r=0.48; p<0.001), and mean TcCO₂ and bicarbonate (r=0.47; p<0.001) (online supplemental figure S2). In addition, there was moderate positive correlation between mean daily TcCO₂ and once per day PaCO₂ (r=0.60; p<0.001) (online supplemental figure S2).

Figure 2.

Correlation between episodes of bradycardia (<100 beats per minute for ≥10 s) per day and mean daily transcutaneous carbon dioxide (TcCO₂). There was moderate negative correlation between mean TcCO₂ and episodes of bradycardia (r=−0.56; p<0.001).

In a sensitivity analysis excluding infants who were not on invasive mechanical ventilation, there remained no difference in episodes of intermittent hypoxaemia, severe intermittent hypoxaemia, the proportion of time with low SpO₂, cerebral hypoxaemia, abdominal hypoxaemia or episodes of bradycardia (results not shown). In this sensitivity analysis, the ventilator settings, the mean TcCO₂, PaCO₂, pH and bicarbonate also did not differ.

DISCUSSION

This trial of late permissive hypercapnia targeted different ranges of TcCO₂ among very preterm infants on positive pressure respiratory support. Targeting a slightly higher TcCO₂ did not alter measures of respiratory stability. However, the intended TcCO₂ separation was not realised as achieved TcCO₂ regressed to the baseline and did not differ between groups. Infants in both groups had frequent episodes of intermittent hypoxaemia, cerebral hypoxaemia and abdominal hypoxaemia. The positive correlations between mean TcCO₂ and FiO₂, mean airway pressure, and cerebral and abdominal hypoxaemia indicate that higher TcCO₂ was a marker for increased respiratory illness severity. Bradycardia events occurred relatively frequently in many infants, and higher achieved levels of TcCO₂ were associated with fewer episodes of bradycardia.

Limitations

This pilot trial of 25 very preterm infants did not use a cross-over design and was not powered to detect potentially important clinical differences related to levels of permissive hypercapnia.^{1 12–15} Enrolled infants had two periods of intervention in order to assess the feasibility of making relatively small intentional alterations in TcCO₂ targets and also to increase the study power. It is possible that the 5 mm Hg (0.67 kPa) intended change may have been difficult to achieve because of the inherent random variation in respiratory drive and expected fluctuations in carbon dioxide due to compensatory changes in spontaneous ventilation, episodes of cessation of spontaneous ventilation or changes unrelated to intervention such as endotracheal tube secretions or shifting atelectasis. In addition, the small sample size increases the risk of confounding and the inclusion of infants on NIPPV may have impacted the ability to target TcCO₂ as it may be difficult to influence the carbon dioxide levels because the mechanisms causing hypoxaemia and bradycardia, such as apnoea, asynchrony or forced expirations, may differ between intubated and non-intubated infants. ^{16 17} Future physiologic studies targeting late permissive hypercapnia to assess control of breathing may limit recruitment to infants on invasive mechanical ventilation, require tighter TcCO₂ ranges, the use of different absolute pCO₂ ranges rather than changes from baseline, mandated ventilator changes to achieve the target range and/or the use of a no-man’s-land to achieve separation.

In this study, we measured cerebral oxygenation but did not assess cerebral perfusion. Large fluctuations in carbon dioxide, whether higher or lower, are associated with brain injury among extremely preterm infants, possibly due to impaired cerebral autoregulation in this population.¹⁵ Higher thresholds of carbon dioxide with lower limits of pH than we used have been trialled previously within the first 2 weeks after birth but were associated with an increased risk of necrotising enterocolitis.¹⁴ Our study used transient mild changes in targeted carbon dioxide levels after 7 days of age, which is beyond the highest risk period of intraventricular haemorrhage. In addition, we used a pH control lower limit of 7.20 that has been tested in a large randomised clinical trial and did not result in adverse effects.¹⁸ We avoided hypocapnia, which may diminish cerebral blood flow by using a lower TcCO₂ limit of 40 mm Hg.

Permissive hypercapnia is widely used in neonatology,^{18 19} but the effects on respiratory stability have not been studied. We used continuous TcCO₂ monitors that correlated with daily blood gas measurements alongside detailed continuous physiologic data to assess respiratory control and oxygenation. It is known that preterm infants on supplemental oxygen therapy have frequent episodes of hypoxaemia.⁹ In the current study, despite being on positive pressure respiratory support, infants in both groups had frequent episodes of intermittent hypoxaemia, severe intermittent hypoxaemia, cerebral hypoxaemia and abdominal hypoxaemia. Previous studies have found that intermittently measured PaCO₂ levels above 55 mm Hg were associated with modestly higher cerebral saturations on NIRS in very preterm infants.²⁰ It is also known that small changes in FiO₂ modulate NIRS measures,²¹ but in our study the difference in FiO₂ and the proportion of time with hypoxaemia was not significant.

We found a moderate negative correlation between TcCO₂ and episodes of bradycardia. It is possible that achieving carbon dioxide levels above the apnoeic threshold⁸ decreases episodes of apnoea and associated bradycardia among ventilated very preterm infants. Carbon dioxide administration has been shown to reduce apnoea among spontaneously breathing preterm infants.^{5 22} As most infants were intubated in the current study, it is likely that centrally mediated responses to CO₂ rather than known differences in CO₂ thresholds of upper airway muscles⁶ were responsible. However, future randomised controlled trials of late permissive hypercapnia among intubated preterm infants focused on bradycardia are needed to determine causality. Finally, we noted a high rate of refusal of informed consent possibly due to the lack of anticipated benefit for participants.

CONCLUSION

Targeting higher carbon dioxide levels did not improve respiratory stability among very preterm infants on positive pressure respiratory support although the intended TcCO₂ separation was not achieved. Future studies of permissive hypercapnia may consider stricter mandated interventions and target ranges to increase the TcCO₂ separation.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Permissive hypercapnia is a commonly used lung-protective strategy in the neonatal intensive care unit.

• The effect of altering carbon dioxide levels on respiratory stability among ventilated very preterm infants is not known.

WHAT THIS STUDY ADDS

• Targeting higher carbon dioxide levels did not improve respiratory stability among very preterm infants on positive pressure respiratory support but the intended change in carbon dioxide levels was not achieved.

• There was moderate negative correlation between higher levels of transcutaneous carbon dioxide and fewer episodes of bradycardia.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• Randomised controlled trials of late permissive hypercapnia focused on bradycardia, and associated outcomes among intubated preterm infants may require stricter mandated interventions and TcCO₂ target ranges.

Data availability statement

Data are available upon reasonable request.