Discussion
To our knowledge, this study is the first systematic review on this topic using the GRADE approach. It included four RCTs (693 infants) and one cohort study (371 infants) and demonstrated that permissive hypercapnia did not significantly reduce or increase any infants’ outcomes assessed (death, IVH, BPD, NEC, PVL, and NDI) compared with the normocapnia strategy (very low-to-low certainty of evidence). However, the largest RCT (Thome 2015) 17 reported a higher rate of NEC in the permissive hypercapnia group than in the normocapnia group, which requires further investigation.
The strengths of our systematic review, compared with previous ones, include (1) the use of the GRADE approach to evaluate the certainty of evidence for each outcome, (2) the use of the definition of BPD at a postmenstrual age of 36 weeks, not at 28 days after birth, and (3) the inclusion of both RCTs and observational studies. The GRADE approach was developed as a transparent method for grading the certainty of evidence for systematic reviews and guidelines.12 Although previous systematic reviews that evaluated the effect of permissive hypercapnia on BPD assessed the risk of bias in the included studies, they did not use the GRADE approach 12.
Unlike the previous systematic reviews that used BPD diagnosed at 28 days of age, this systematic review defined BPD as the use of oxygen or respiratory pressure support [e.g., CPAP, high-flow nasal cannula at a rate of >2 L/min] at 36 weeks of corrected gestational age.21-23 Although defining BPD has been a controversial issue in recent years, the definitions or diagnoses near term or later (from 36 weeks to 40 weeks postmenstrual age or to 1 year of age) are considered better than those at 28 days or a month after birth because the former definitions predict long-term adverse consequences better than the latter.24 The last systematic review on this topic (Ma, 2016) 21 reported that permissive hypercapnia did not reduce BPD (RR 0.93; 95% CI 0.83–1.03) (Ma 2016); however, the authors did not clearly describe the definition of BPD. Based on the results of that systematic review (Ma, 2016), the outcome of BPD seemed to include both BPD diagnosed at 28 days of age and that diagnosed at a postmenstrual age of 36 weeks together. Using the BPD definition at a postmenstrual age of 36 weeks, we found that hypercapnia did not reduce BPD. In addition to the small sample size, one of the other potential explanations for the negative finding for BPD is that permissive hypercapnia used might not always achieve low tidal volume enough to prevent BPD. However, none of the trials included in this systematic review measured and compared the tidal volumes of the two groups. All the included trials reported comparable peak inspiratory pressure (PIP) measurements between the two groups, except for one RCT (Thome 2015), which indicated that the target PaCO2 level might be insufficient to achieve low PIP or small tidal volume enough to prevent BPD.
This systematic review found that permissive hypercapnia did not have significant adverse neurological effects (severe IVH and NDI), consistent with the findings of previous systematic reviews.21,23 The increased risk of IVH due to permissive hypercapnia was also concerning in this review. High PaCO2 and low pH dilate cerebral blood vessels, increase cerebral blood flow, and cause fluctuations in cerebral blood flow, which may predispose infants to IVH.6,25, 26 Moreover, cerebral autoregulation was attenuated in the presence of hypercapnia. CO2 cerebral blood flow reactivity is more robust than pressure flow 27 and there is a steep positive linear relationship between PaCO2 and cerebral blood flow on postnatal days 2–4.28 One potential reason for the lack of increase in IVH with permissive hypercapnia is that the target range of PaCO2 in the permissive hypercapnia groups might be insufficient to increase IVH. One RCT (Mariani 1996) and one cohort study included in this review used a target PaCO2range of less than 55 mmHg in the permissive hypercapnia group. Moreover, in other studies, actual values of PaCO2 did not exceed 55 mmHg despite the target PaCO2, except in one study (Thome 2015). Even the Thome 2015 study did not achieve the target range in the permissive hypercapnia group, and the mean PaCO2 exceeded 55 mmHg only after day 4. Therefore, it is possible that the actual PaCO2 values of most infants in the permissive hypercapnia group did not reach the threshold (51–55 mmHg) that was reported to increase cerebral blood flow in previous studies. 28-31 . Another potential explanation for the lack of an increase in IVH is that most of the included studies administered sodium bicarbonate to correct acidosis. The correction of acidosis might attenuate the adverse effects of hypercapnia, although rapid infusion of sodium bicarbonate may increase the risk of IVH.32, 33
In this meta-analysis, the effect estimates (RR) indicated a potential increase in NEC in infants with permissive hypercapnia, although the difference was not statistically significant, with its 95% CI just crossing the neutral effect estimate (RR 1.69 [95% CI: 0.98–2.91]). Furthermore, the largest RCT (Thome, 2015) reported a significant increase in NEC in the permissive hypercapnia group. It is important to note that the RCT with an increased NEC in the permissive hypercapnia group (Thome 2015) used the longest intervention period (14 days after birth) and the highest target range of permissive hypercapnia (PaCO2 of 65–75 mmHg) among all eligible studies. The effect of lengthy permissive hypercapnia with a high PaCO2 target range on the NEC rates requires further investigation. An ongoing RCT aimed to enroll 160 preterm infants at <37 gestational weeks (HYFIVE; NCT02799875). The trial compares two different levels of target PCO2 and pH (PCO2 60–75 mmHg and pH≧7.2 versus PCO240–55 mmHg and pH ≥ 7.25) between 7 and 14 days in terms of outcomes such as alive ventilator-free days in 28 days, death, BPD, etc. This trial may provide further information on the optimal PaCO2 level and period in preterm infants.
Our systematic review has several limitations. First, the sample sizes of the included studies did not reach the optimal information size for all outcomes assessed (Supplemental Table D). Therefore, the negative findings in the meta-analyses may simply be due to the insufficient sample size. This limitation was reflected in the certainty of evidence downgraded for imprecision. Second, there was clinical heterogeneity among the eligible studies, in which there was a wide variation in the range of targeted PaCO2 and period for the interventions (permissive hypercapnia or normocapnia) (Table 1), although the statistical heterogeneity was low (I2=0–28%). Lastly, the included studies were relatively old, and the findings of the studies may not be applicable to current clinical practice. In particular, respiratory management strategies for premature babies have been rapidly changing over the last decade, and non-invasive ventilation and non-invasive surfactant administration avoiding mechanical ventilation using an endotracheal tube are increasingly used.34, 35 Therefore, infants who are on mechanical ventilation and are eligible for permissive hypercapnia may be sicker in the current clinical practice than those in previous trials.