Determining Cerebrovascular Sensitivity to pH Using Continuous Arterial Blood Gas Analysis—A Prospective Observational Study

BACKGROUND: Although carbon dioxide (CO 2 ) is widely regarded as a driver for cerebral vasomodulation, often termed cerebrovascular CO 2 -reactivity, experimental evidence has shown that this response is primarily mediated through changes in pH rather than CO 2 . This complicates the management of cerebral blood flow (CBF) during the treatment of acid-base disorders with partial respiratory or metabolic compensation, when the CBF has normalized together with pH equilibration, despite persistent hyper- or hypercapnia. Although a pH-guided approach would therefore be preferred, no quantitative estimate of the relationship between pH and CBF has currently been reported in the literature. Therefore, this study aims to estimate the cerebrovascular sensitivity to pH and explore whether changes in CBF velocity (CBFV) are more closely related to arterial pH or CO 2 . METHODS: In patients undergoing cardiac surgery, CBFV and continuous in-line arterial blood gas monitoring, incorporated in the cardiopulmonary bypass circuit, were analyzed. Mixed-effects models were constructed with percentage change in CBFV as response variable and arterial pH and/or CO 2 as explanatory variable(s). Model fit of the explanatory variables was compared using marginal and conditional R 2 . Improvement of model fit and parsimony were assessed with the likelihood-ratio test (LRT) and change in Akaike information criterion (ΔAIC), respectively. RESULTS: Thirteen patients were included in the analysis with a total of 1960 datapoints. CBFV decreased by 30.6% per 0.1 increase in arterial pH (95% confidence interval [CI], 28.9–32.2; P < .0001), and increased by 1.95% per mm Hg increase in PaCO 2 (95% CI, 1.76–2.14; P < .0001). Arterial pH explained a greater proportion of variance than PaCO 2 (marginal R 2 : 0.444 vs 0.294; conditional R 2 : 0.744 vs 0.530). Furthermore, adding arterial pH to the null CO 2 -model considerably improved model fit (LRT = 616.4) and parsimony (ΔAIC = −615), whereas adding arterial CO 2 to the pH-model barely improved model fit (LRT = 4.6) and parsimony (ΔAIC = −3). CONCLUSIONS: This exploratory study provides a quantitative estimate of the relationship between CBFV and pH, and suggests that changes in CBFV are more closely related to arterial pH than to CO 2 . This estimate of the cerebrovascular sensitivity to pH was used to describe a feasible pH-guided approach in the management of CBF during the treatment of partially compensated acid-base disorders.