Reply to Kikutani et al.: Early PaCO2 Changes after Initiating Extracorporeal Membrane Oxygenation: Considerations for Future Research.

From the Authors:

We would like to thank Dr. Kikutani and colleagues for their thoughtful comments on our article on the association between early changes in PaCO and neurological complications in patients on extracorporeal life support (ECLS) (1). Here, we will try to address them. First, as stated in the discussion, we fully acknowledge that the availability of only two blood gases is the main limitation of the study. Indeed, this may result in an underestimation of the maximal change in PaCO in the first 24 hours, as illustrated by Dr. Kikutani’s data. It is unclear, however, if a transient drop in PaCO, more likely to be missed by the Extracorporeal Life Support Organization (ELSO) data, is more harmful than a sustained decrease, which is more likely to be adequately captured. More granular data would be needed to better evaluate the impact of different types of changes in PaCO over time. The main challenge, however, is that neurological complications are relatively infrequent, and large sample sizes would be needed to provide adequate power to detect a relatively small effect size.

Second, we performed an analysis of the association between a PaCO drop >50% and neurological complications stratified by baseline PaCO subgroups as requested (Figure 1). Visual inspection of the forest plot suggests a more pronounced effect in patients with baseline hypocapnia or severe hypercapnia (U-shaped relationship), which goes against Dr. Kikutani’s hypothesis of reduced cerebrovascular consequences of changes in PaCO in patients with the most severe hypercapnia. In the stratified analysis, the Breslow-Day test did not suggest significant heterogeneity (P = 0.718), and the Mantel-Haenszel estimate of the common odds ratio was 1.45 (95% confidence interval, 1.13–1.89; P = 0.005), which is in line with other analyses. The observed effect of changes in PaCO in the extremes of baseline PaCO could be explained by a reset in the range of PaCO cerebrovascular response (2). Although chronic conditions such as chronic obstructive pulmonary disease have been associated with blunted cerebrovascular reactivity, the sensitivity of cerebral blood flow to acute changes in both O2 and CO2 is increased by sustained exposure (48 h) to hypoxemia (3).

Figure 1.

Unadjusted odds ratio of neurological complications associated with a relative PaCO drop >50% stratified by baseline PaCO subgroup. OR = odds ratio.

Third, we agree that known risk factors for neurological complications, including hypertension, hyperlipidemia, diabetes mellitus, atrial fibrillation, and chronic use of antithrombotic therapy, could have been included in the multivariate analysis if they had been available. However, that data is not reliably recorded in the ELSO registry. Moreover, to act as confounding factors, those variables should not only be associated with the outcome variable (neurological complications) but also the exposure variable relative change in PaCO (Rel∆CO2). It is improbable that factors such as hyperlipidemia or atrial fibrillation are associated with Rel∆CO2, making them unlikely to be true confounders. Rel∆CO2 is mainly determined by baseline PaCO as well as ventilator and ECLS parameters selected by the clinical team. We thus think that the association found between Rel∆CO2 and neurological complications is robust despite not controlling for certain known risk factors. Since the publication of our article, a similar association has now also been reported in patients on venoarterial extracorporeal membrane oxygenation by another group also using data from the ELSO registry (4).

What remains unclear, in our opinion, is the optimal rate of PaCO correction after initiation of ECLS to prevent neurological injury. It is possible that there is no one-size-fits-all target and that PaCO correction should be individualized to optimize cerebral blood flow by using neuromonitoring such as near-infrared spectroscopy or transcranial doppler.