The effect of hypertension on cerebrovascular carbon dioxide reactivity in atrial fibrillation patients

Atrial fibrillation (AF) and hypertension (HTN) are both associated with impaired cerebrovascular carbon dioxide reactivity (CVRCO2), an indicator of cerebral vasodilatory reserve. We hypothesised that CVRCO2 would be lower in patients with both AF and HTN (AF + HTN) compared to normotensive AF patients, due to an additive effect of AF and HTN on CVRCO2. Forty AF (68 ± 9 years) and fifty-seven AF + HTN (68 ± 8 years) patients underwent transcranial Doppler ultrasound measurement of middle cerebral artery blood velocity (MCA Vm) during stepped increases and decreases in end-tidal carbon dioxide (PETCO2). A cerebrovascular conductance index (CVCi) was calculated as the ratio of MCA Vm and mean arterial pressure (MAP). CVRCO2 was determined from the linear slope for MCA Vm and MCA CVCi vs PETCO2. Baseline MAP was higher in AF + HTN than AF (107 ± 9 vs. 98 ± 9 mmHg, respectively; p < 0.001), while MCA Vm was not different (AF + HTN:49.6 [44.1–69.0]; AF:51.7 [45.2–63.3] cm.s−1; p = 0.075), and CVCi was lower in AF + HTN (0.46 [0.42–0.57] vs. 0.54 [0.44–0.63] cm.s−1.mmHg−1; p < 0.001). MCA Vm CVRCO2 was not different (AF + HTN: 1.70 [1.47–2.19]; AF 1.74 [1.54–2.52] cm/s/mmHg−2; p = 0.221), while CVCi CVRCO2 was 13% lower in AF + HTN (0.013 ± 0.004 vs 0.015 ± 0.005 cm.s−1.mmHg−1; p = 0.047). Our results demonstrate blunted cerebral vasodilatory reserve (determined as MCA CVCi CVRCO2) in AF + HTN compared to AF alone. This may implicate HTN as a driver of further cerebrovascular dysfunction in AF that may be important for the development of AF-related cerebrovascular events and downstream cognitive decline. We demonstrated reduced cerebrovascular CO2 responsiveness in atrial fibrillation with hypertension (AF+HTN) vs. atrial fibrillation (AF). Furthermore, AF per se (as opposed to normal sinus rhythm) predicts reduced cerebrovascular CO2 responsiveness. Our findings suggest additional cerebrovascular dysfunction in AF+HTN vs. AF.


Introduction
Atrial fibrillation (AF) is a supraventricular arrhythmia affecting 0.5-1% of the global population [1], making it the most prevalent sustained cardiac rhythm disorder.Alarmingly, this number is projected to rapidly increase over the next 30 years as populations age and modern healthcare improves survival outcomes for acute cardiac events [1,2].
Hypertension (HTN) is a powerful risk factor for AF, and has previously been reported in 40-90% of AF cohorts [3].In addition, HTN is likely responsible for up to 50% of all AF diagnoses [3].Downstream complications of AF include thromboembolic events such as stroke, of which AF patients are five times more likely to develop [4], and neurodegenerative disorders including cognitive decline and dementia, even in the absence of overt stroke [5].Similarly, HTN is a known risk factor for stroke, present in ~60% of stroke occurrences [6], as well as cognitive impairment disorders [7].Possible mechanisms for this phenomena include silent cerebral infarction [8], inflammation [9] and cerebral hypoperfusion [10].
In AF, emerging evidence suggests that impaired cerebrovascular function may also contribute to this heightened risk of stroke and cognitive decline.It is currently unknown if HTN exacerbates cerebrovascular dysfunction in AF.Peripheral vascular dysfunction is well-documented in AF.Endotheliumdependent brachial artery flow-mediated dilatation is reduced in AF patients and coupled with increased levels of plasma von Willebrand factor, a known biomarker of endothelial dysfunction [11].Furthermore, increased oxidative stress [12] and the shift to a pro-inflammatory state [13] in AF patients synergistically worsen endothelial function.Moreover, hypertension evokes similar effects on peripheral vascular function.Structural vascular remodelling evoked by increased pressures and pulsatility increases oxidative stress and inflammation [14], causing vessel stiffening and subsequent endothelial dysfunction, as evidenced by reductions in nitric oxide (NO) production in HTN [15].Interestingly, peripheral vascular function was previously reported to not be different between a group of HTN patients and a group with both AF and HTN (AF + HTN) [16], likely due to the shared vascular pathologies between AF and HTN.Evidence of cerebrovascular dysfunction also exists in AF.
Cerebrovascular CO 2 reactivity (CVR CO2 ), defined as the change in (Δ) cerebral blood flow (CBF) versus Δ partial pressure of CO 2 (P a CO 2 ), is widely recognized as a marker of cerebral vasodilatory reserve, and has been associated with increased risk of mortality when impaired [17].AF patients exhibit worsened CVR CO2 compared to healthy controls, indicative of cerebrovascular dysfunction [18].Furthermore, neurovascular coupling, whereby cerebral blood flow is increased to support neuronal metabolic needs, is reduced in AF [19].CVR CO2 responses have also been reported to be worsened in hypertension [20] although this is not a universal finding [21].The combined effects of AF and HTN on CVR CO2 are yet to be determined; thus, it is still unknown whether their effects on CVR CO2 are additive or occlusive when concurrently present.
The aim of this study was to assess whether AF patients with HTN exhibit worsened cerebral vasodilatory reserve (i.e., poorer CVR CO2 ) compared to normotensive AF patients.Considering the current evidence indicating the effects of AF and HTN independently on CVR CO2 , we hypothesized that AF patients with hypertension would exhibit lowered CVR CO2 compared with normotensive AF patients.

Graphical Abstract
We demonstrated reduced cerebrovascular CO 2 responsiveness in atrial fibrillation with hypertension (AF+HTN) vs. atrial fibrillation (AF).Furthermore, AF per se (as opposed to normal sinus rhythm) predicts reduced cerebrovascular CO 2 responsiveness.Our findings suggest additional cerebrovascular dysfunction in AF+HTN vs. AF.

Participant characteristics
Ninety-seven participants were included across two AF groups: AF (n = 40) and AF + HTN (n = 57).Of these, we included fifteen participants (which met the current inclusion criteria) from a previous study in which we used the same experimental approaches and outcome measures to cross-sectionally analyse cerebral vasodilatory reserve in three groups (AF, HTN and healthy controls) [18].Patients were recruited from the following sites:

Experimental measures
A thorough participant history was recorded, including medication use, alcohol consumption and smoking habits.Using the recorded participant characteristics, each participant was calculated a CHA 2 DS 2 -VASc score.Anthropometric measures included height, weight, waist (umbilical level) and hip (femoral trochanter level) circumference.Noninvasive supine brachial BP was measured using an automated oscillometric device (M2, Omron, Kyoto, Japan).Beat-to-beat BP was continuously recorded by finger photoplethysmography (Finometer MIDI, Finapres Medical Systems, Amsterdam, The Netherlands), and heart rate (HR) by lead II ECG (BioAmp, ADInstruments, Dunedin, New Zealand).An oronasal mask connected to a heated pneumotach (Hans Rudolph, Kansas City, KS, USA) was used to record minute ventilation.A sampling line connected the oronasal mask to a capnograph (RespSense, Nonin Medical, Plymouth, MN, USA) for recording of end-tidal partial pressure of CO 2 (P ET CO 2 ).Bilateral 2 MHz transcranial Doppler probes (Doppler BosX, DWL, Sipplingen, Germany) were used to insonate the cerebral arteries fixed at their respective temporal windows, providing mean blood flow velocity (V m ) for the left posterior cerebral artery (PCA) and the right middle cerebral artery (MCA).Insonation of the MCA was successful in all participants, but only achieved in forty-seven participants (48%) for the PCA.

Experimental protocol
Pre-study guidelines included no strenuous exercise, food and caffeine intake for 12 h prior to their study, no alcohol the day before and day of their study and to refrain from consuming medication (excluding anticoagulants) on the morning of study.Physiological assessments were undertaken with participants lying supine on a bed, with their head on a pillow.After instrumentation, a 10-minute baseline period was acquired whilst participants breathed room air.During this baseline, the mean of 3 brachial BP recordings was used to correct beat-to-beat BP obtained by finger photoplethysmography. Stepped increases and decreases in P ET CO 2 were then used to assess CVR CO2 .Hypercapnic steps involved participants breathing through a Douglas bag circuit containing 4% CO 2 , followed by 7% CO 2 .Gas mixtures both contained 21% oxygen (N 2 balanced).Each gas mixture was breathed for 4 min.Following hypercapnic steps, participants breathed room air to allow P ET CO 2 to return to baseline.They were then required to alter their respiratory rate and depth to lower their P ET CO 2 to an equal and opposite value to that achieved in the two hypercapnic steps.Each hypocapnic step was 2 minutes in duration.All participants completed the 4% hypercapnic step and its equal opposite hypocapnic step.Two participants did not complete the 7% hypercapnic step and one participant did not complete it's equal opposite hypocapnic step.These participants were removed from our CVR CO2 analyses.

Data analysis
The ratio of height and weight squared was used to express participants BMI.Waist to hip and waist to height ratios were calculated for each participant.A multi-channel data acquisition system (Powerlab 16/35 and Labchart Pro 8.1.13,ADInstruments) was used to record and digitize analogue signals at 1 KHz.HR was calculated on a beat-to-beat basis from the ECG.Mean arterial pressure (MAP) and was obtained by integrating the arterial BP waveform over the complete cardiac cycle on a beat-to-beat basis.Similarly, V m was obtained by integrating the MCA and PCA V m waveforms over the complete cardiac cycle on a beat-to-beat basis.Cerebrovascular conductance index (CVCi) was calculated as the ratio of MCA V m or PCA V m with MAP.CVCi was not calculated if either V m or MAP values were determined to be statistical outliers.Baseline measures were averaged over the entire 10-minute duration.Hypercapnic and hypocapnic step change measures were obtained from the final minute of the experimental period.The linear slopes of MCA V m and PCA V m versus P ET CO 2 across the range of hypo-and hypercapnic steps provided CVR CO2 .Outliers were defined as ± interquartile range (IQR)*2.2 and excluded from analyses.

Statistical analysis
Shapiro-Wilk test was used to assess data normality.Levene's test was used to assess homogeneity of variance between group data.Independent two-tailed Students t-test was implemented to analyse normally distributed continuous data and Mann-Whitney U test for non-normally distributed continuous data.Fisher's Exact Probability Test was used to analyse normally distributed categorical data and Pearson chi-square for non-normally distributed categorical data.Effect size for significant continuous variables reported with Cohen's d (d) and Phi (φ) for significant categorical variables.Two-way analysis of variance (ANOVA) was implemented to compare the effect of patient group and P ET CO 2 on MCA V m , MCA CVCi, PCA V m , PCA CVCi and MAP.For ANOVA, data normality was assessed by inspection of Q-Q plots for standardized residuals.Mauchly's test of sphericity was used to assess the assumption of sphericity.If the assumption of sphericity was violated, a Greenhouse-Geisser correction was applied.Post-hoc pairwise comparisons were performed upon identification of significant main effects using t-test with Bonferroni correction.A forced entry regression model was used to determine the influence of variables of interest on baseline and exponential CVR CO2 parameters (MCA V m , MCA CVCi, PCA V m and PCA CVCi).Multicollinearity

Participant characteristics
Age and the proportion of males and females was not different in the AF and AF + HTN groups (Table 1

Baseline cerebral hemodynamics
Figure 1 shows baseline cerebrovascular measures.MCA V m was numerically lower in AF + HTN (49.

Forced entry regression
Significant predictors of baseline variables and CVR CO2 as identified from forced entry linear regression are shown in Table 3. Fibrillation (vs normal sinus rhythm [NSR]) at the time of the experimental visit was associated with a lower MCA V m CVR CO2 .The presence of hypertension was associated with a lower baseline MCA CVCi, and being female was associated with higher baseline MCA V m and baseline MCA CVCi.None of our entered independent variables were found to be associated with baseline PCA V m , baseline PCA CVCi, MCA CVCi CVR CO2 , PCA V m CVR CO2 and PCA CVCi CVR CO2 .

Discussion
This is the first study to investigate whether the presence of HTN as a comorbidity in AF worsens CVR CO2 .The major novel finding of the present study is that CVCi CVR CO2 is lowered in AF + HTN compared to AF, and may implicate an additional level of cerebrovascular dysfunction in AF + HTN.This may be important for the development of cerebrovascular events and downstream cognitive decline commonly associated with AF.
Baseline MCA V m was not different between AF and AF + HTN groups, indicating that cerebral perfusion at rest is retained in AF + HTN similar to that in AF.It is well accepted that there is a rightward shift in the cerebral autoregulatory curve in HTN, whereby CBF is maintained similarly to that in normotensives with higher MAP [22].This may explain our conflicting baseline responses in terms of MCA Vm and CVCi.As a consequence of increased baseline MAP in AF + HTN, MCA CVCi is lowered, in order to buffer the transmission of peripheral BP to CBF.This may be explained due to HTN evoked remodelling, hypertrophy and stiffening within the cerebral arteries and arterioles, resulting in increased wall-lumen ratios and decreased vessel diameters [23].These effects on the vasculature may be explained by a number of interconnected processes.Oxidative stress is heightened in HTN, and promotes proliferation of vascular smooth muscle cells as well as extracellular matrix remodelling [14].Elevated reactive oxygen species (ROS) can also drive inflammation in HTN, resulting in further vascular hypertrophy.Indeed, elevated markers of inflammation, including C-reactive protein and interleukin-6, are independently associated with HTN [24].Furthermore, the direct mechanical effects of heightened pressure on vessel walls can promote hypertrophy and remodelling through activation of extracellular matrix protein cascades, whereby noncellular material accumulates within vessel walls [25].This can also drive ROS production, resulting in a pathological cycle of cerebrovascular remodelling.
Interestingly, in the present study we have shown a reduction in MCA CVCi CVR CO2 in AF + HTN but no difference in MCA V m CVR CO2 between AF and AF + HTN.This offers distinct insight from our previous work, where we compared CVR CO2 in separate groups with either AF, HTN, or healthy controls (i.e.normotensive and with a normal sinus rhythm), and demonstrated lower MCA V m CVR CO2 in AF versus HTN and healthy controls.The current study, in which all patients studied had AF, suggests that in AF + HTN, CO 2 -mediated increases in CBF per unit pressure are blunted.The use of a conductance index allows us to somewhat account for changes in perfusion pressure and provides more insight to vasodilator responsiveness.Lowered CVCi CVR CO2 may be indicative of additional cerebrovascular dysfunction evoked by HTN.The aforementioned HTN induced cerebrovascular remodelling will have functional consequences in terms of vascular responsiveness to CO 2 that may drive differences in CVCi CVR CO2 between our AF + HTN and AF groups.In addition, endothelial damage/dysfunction associated with oxidative stress, inflammation, and reductions in NO bioavailability [26] and known to be present in both HTN and AF may also contribute and summate to worsen CVCi CVR CO2 in AF + HTN.In contrast, brachial artery flow mediated dilatation, a marker of peripheral endothelial function, is not different in HTN versus AF + HTN [16], and might suggest a stronger contribution of HTN than AF to peripheral/cerebrovascular dysfunction [22].
Forced entry regression suggested patients that were fibrillating at the time of study (as opposed to being in NSR) exhibited lower MCA V m CVR CO2 .MRI and xenon inhalation studies have previously shown improved CBF in AF patients who underwent NSR restoration procedures [27,28], suggesting a direct effect of fibrillation on cerebral perfusion Fig. 2  per se.However, in earlier work [18] we observed that MCA V m was lower in AF patients that were fibrillating at the time of study compared to those in NSR, but MCA V m CVR CO2 was not different.Differences in sample size (n = 97 in the present study vs. n = 31 in Junejo et al. [18]) and the analytical approach employed, may partly explain the discrepant findings, and further work is required to better understand how AF per se affects CVR CO2 and by what mechanisms.
The anterior and posterior cerebral circulations differ in regards to CVR CO2 , with the posterior cerebral circulation exhibiting absolute CVR CO2 values ~50% lower than that seen in the anterior cerebral circulations [29,30], which our results generally agree with.Additionally, differences may exist in endothelial function between the two circulations, as endothelium-derived NO has been shown to play a greater role in vasodilatory responses in the anterior cerebral circulation [31].Notably, baseline PCA and PCA CVR CO2 measures were not different between AF and AF + HTN.This may indicate preserved endothelial function in the posterior cerebral circulation in AF + HTN vs. AF groups.
Our findings should be interpreted in light of several experimental considerations.While transcranial Doppler ultrasound offers real-time measurements of cerebrovascular function, is non-invasive and easy to implement during physiological interventions, its inability to directly measure blood flow is a limitation.Vessel diameter cannot be measured with transcranial Doppler, thus changes in cerebral blood velocity may not be directly indicative of CBF changes.However, as transcranial Doppler-derived V m changes have been significantly correlated with gold standard measures of CBF (i.e., gadolinium tracer MRI and arterial spin labelling MRI [32]) V m values are generally considered proportional to CBF.P ET CO 2 has a strong positive correlation with P a CO 2 [33], and acts as a reliable non-invasive surrogate measure, though we acknowledge its possible underestimation of P a CO 2 at rest [34].There is debate surrounding the optimal method for assessing cerebrovascular function [35].The advantages of using fixed concentrations of CO 2 (4 and 7%) as in the present study, is that the experimental set-up is relatively simple, inexpensive, convenient in the clinic, does not required a complex computerised gas delivery system, and requires minimal participant cooperation.This method has also shown good between-day test-retest reliability [18].Our primary measure of cerebrovascular responsiveness (i.e., CVCi CVR CO2 ) was derived from the linear slope of the relationship between P ET CO 2 and MCA CVCi for each patient and was significantly (p = 0.047) lower in AF + HTN Table 3 Forced entry multiple regression with baseline MCA V m (cm•s −1 ), baseline MCA CVCi (cm.s −1 .mmHg−1 ), baseline PCA V m (cm•s −1 ), baseline PCA CVCi (cm.s −1 .mmHg−1 ), MCA V m (cm.s −1 .mmHg−1 ), MCA CVCi (cm.s −1 .mmHg−1 ), PCA V m (cm.s −1 .mmHg−1 ) and PCA CVCi (cm.sThe effect of hypertension on cerebrovascular carbon dioxide reactivity in atrial fibrillation patients versus AF.While a tendency (p = 0.083) for an interaction between Group and P ET CO 2 was observed for MCA CVCi (Supplementary Fig. 1B) this did not reach significance.This discrepancy is likely caused by differences in the methods of data analysis used.We did not screen for the presence of subclinical brain infarcts, alhough patient history and clinical records did not suggest any presence of these.In addition, although participants were requested to refrain from consuming medications on the morning prior to their study visit, it is unlikely that this was sufficient to cause full washout prior to data acquisition.Finally, it is a limitation of the current study that a healthy control group (i.e., normotensive and NSR) was not recruited, however we believe that this does not diminish our ability to achieve the primary aim of the study which was to determine whether the presence of HTN in AF worsens cerebrovascular responsiveness.
In conclusion, this is the first study that demonstrates reduced CVR CO2 in AF patients with concurrent HTN as compared to normotensive AF patients.This may be important for clinical management due to the possible heightened risk of cerebrovascular events and downstream cognitive decline.

Fig. 1
Fig. 1 Baseline MCA V m (A), MCA CVCi (B), PCA V m (C), and PCA CVCi (D) in participants with AF or AF + HTN.Individual values plotted with bars displaying median and interquartile range for MCA V m MCA CVCi and PCA V m , and mean and standard deviation for PCA CVCi.AF atrial fibrillation; AF + HTN atrial fibrillation with hypertension; MCA V m middle cerebral artery mean blood velocity; PCA V m posterior cerebral artery mean blood velocity; MCA CVCi middle cerebral artery cerebrovascular conductance index; PCA CVCi posterior cerebral artery cerebrovascular conductance index

Table 1
Mean ± SD displayed for normally distributed and median [interquartile range] for non-normally distributed continuous variables.Frequency (percentage) displayed for categorical discreate variables.Independent samples t test or Mann-Whitney U test used to infer statistical differences for continuous variables and Pearson's or Fisher's Exact Probability test for categorical data.Cohen's d (d) and Phi (φ) reported as effect size for significant continuous and categorical variables, respectively.Patients in AF at the time of data acquisition are identified as 'Fibrillating'.Significance: P ≤ 0.05 AF atrial fibrillation, AF + HTN atrial fibrillation with hypertension, φ, Phi, BMI body mass index, BP blood pressure, HR heart rate, MAP mean arterial pressure, MoCA Montreal Cognitive Assessment and autocorrelation were tested to reduce the risk of regression model violations.Tolerance, variance inflation factor, and Durbin-Watson statistic were 0.570-0.950,1.053-1.755and 1.754-2.520respectively.Mean ± standard deviation (SD) are presented for normally distributed data, median [interquartile range] for non-normally distributed data and frequency (percentage) for categorical data.Statistical significance was considered as p < 0.05.Analysis was performed using SPSS, version 27 (IBM Corp., Armonk, NY, USA).

Table 2
Mean ± SD displayed for normally distributed and median [interquartile range] for non-normally distributed variables.Independent samples t-test or Mann-Whitney U test used to infer statistical differences.Cohen's d reported as effect size for significant variables.Significance: p ≤ 0.05 AF atrial fibrillation, AF + HTN, atrial fibrillation with hypertension, CVCi conductance index, MCA middle cerebral artery, PCA posterior cerebral artery, V m mean velocity