Unilateral Carotid Body Resection in Resistant Hypertension

Study design and patients

We present pooled results from 2 independent centers in which the primary endpoints were safety and feasibility of uCB excision in patients with drug-resistant hypertension. Secondary endpoints were to assess the proportion of patients showing reductions in ambulatory blood pressure (ABP) and MSNA. Inclusion criteria were: age between 18 to 75 years; taking ≥3 antihypertensive medications, including a diuretic agent, at maximal tolerated dose; no evidence of causes of secondary hypertension following thorough biochemical, clinical, and imaging assessment (detailed in Supplemental Table 1); office systolic blood pressures (OSBPs) ≥150 mm Hg; daytime mean ABP ≥135 mm Hg (Table 1, Supplemental Table 2). There was no control group because this was a first-in-man, safety and feasibility study. Exclusion criteria are detailed in Supplemental Table 3. A total of 113 patients were recruited from specialist hypertension clinics at Bristol, United Kingdom; Gdansk, Poland; and London, United Kingdom. Following screening in Bristol and Gdansk, 15 patients (7 men and 8 women) were eligible, including 3 participants who had undergone renal denervation and 2 with sleep apnea who were on continuous positive airway pressure. All patients gave written informed consent. The study complied with the Declaration of Helsinki, was approved by the Central Bristol research ethics committee (12/SW/0277) and the Medical University of Gdansk Independent Bioethics Commission for Research (NKBBN/398/2012), and was registered at ClinicalTrials.gov (NCT01745172 and NCT01729988). Informed consent was obtained from all patients.

Study protocol

Procedure

Surgery

The surgical removal of the CB was performed following the procedure described by Winter (24). Under either general or local anesthesia with sedation, an incision was made over the anterior aspect of the sternocleidomastoid muscle, one-third of the distance between the angle of the mandible and the clavicle, and over the region of the carotid bifurcation as identified via ultrasound/computed tomography angiography. The sternocleidomastoid muscle was retracted laterally along with the internal jugular vein to expose the carotid bifurcation. By gentle retraction of the external carotid artery (in some cases, the superior thyroid artery was cut to enhance retraction), the intercarotid septum was exposed. Tissue within this septum was isolated from the internal and external carotid arteries, and a ligature was placed at the saddle of the bifurcation. The septal tissue was excised as close to the ligature as possible. The surgeries in some cases involved ligature of the pedicle, and in others cautery over the saddle.

Histology

Resected tissue was fixed with 10% formalin, embedded in paraffin, sectioned (50- to 100-μm thick), and stained with hematoxylin and eosin. The presence of glomus cells and organ margins were checked and photomicrographs were taken.

Analytical methods and statistical analysis

All analyses were conducted blind with respect to the sequence of visits and whether the patient was a BP responder or nonresponder (see the following text). All measurements of MSNA burst area and BRS were calculated by importing acquired data into MATLAB 8.0 (2012b, The MathWorks, Natick, Massachusetts), and statistical tests of these measures were conducted in Prism version 2.0 (GraphPad Software, Inc., La Jolla, California). Results are presented as mean ± SEM. A 2-sided probability value of p < 0.05 was considered statistically significant. Statistical comparisons of groups were assessed by 1-way analysis of variance (ANOVA) for parametric data or Kruskal-Wallis ANOVA on ranks for data that were not normally distributed. A 2-way ANOVA with Sidak (between groups) and Tukey (within groups) post hoc tests was conducted to compare data across and within groups at different follow-ups. Serial within-group comparisons were subjected to repeated measures ANOVA. All variables analyzed with ANOVA passed the D’Agostino & Pearson normality test and also Bartlett’s test for equal variance. All variables passed normality except for day ambulatory systolic blood pressure (ASBP), which was analyzed with Kruskal-Wallis test and expressed as median (first quartile, third quartile). Medications were calculated as a percentage of the maximal recommended dose for hypertension. For each patient, the whole drug equivalents (sum of the percentage of the maximal recommended dose) were calculated on the basis of doses quoted in the British National Formulary.

Safety

There were 2 serious adverse events consisting of prolonged hospitalization of patients with BP that was difficult to control. One of the events occurred shortly after the CB removal procedure, and this event was judged by the Clinical Events Committee (CEC) to be “possibly related” to the unilateral removal of the CB. In the other patient, multiple hospitalizations occurred for BP control before and after uCB removal, and the hospitalizations bore no consistent temporal relationship to the CB removal. These hospitalizations were, therefore, judged by the CEC to be “unrelated” or “unlikely to be related” to the uCB.

In 1 patient with pre-existing moderate obstructive sleep apnea (OSA), sleep-disordered breathing (SDB) worsened after uCB. This was not noted as an adverse event by the study site. The CEC felt, however, that given that SDB was a pre-existing disease in this patient and the apnea-hypopnea index increased from 20 events/h at baseline to 74 events/h 3 months post-carotid body removal, worsening SDB should be noted as an adverse event. The patient was treated with continuous positive airway pressure, and the apnea-hypopnea index decreased substantially. The adverse events and polysomnography are shown in Supplemental Tables 4 and 5, respectively.

Patients maintained their ventilatory response to hypoxia as the average baseline HVR (−0.4 ± 0.1 l/min/% SpO₂) (Table 1) was not changed after uCB resection at any time point. There were no changes in breathing patterns during sleep after uCB (Supplemental Table 5), and although blood oxygen fell to lower minimal levels during desaturation episodes (from 87 ± 1% to 81 ± 1%; p < 0.05), there were no changes in the apnea-hypopnea index, apnea duration, baseline blood oxygen saturation, and average blood desaturation (Supplemental Table 5).

Feasibility

On the basis of its visualization on computed tomography scans (Figure 1), a CB was resected from either the right (n = 11) or left side (n = 4). Characteristic glomus tissue was found subsequently in histological sections of the resected specimen in all (Figure 1) but 1 patient; in this patient we observed no adverse effects, changes in BP, or changes in any of the other measured variables at all follow-up visits.

BP, medication, and autonomic indexes in all patients (n=15)

There was no significant change in either ASBP or OSBP at any time following uCB resection compared with the corresponding values at screening and baseline (Figure 2, Table 1) (n = 15); a similar finding was observed for HBP (Supplemental Figure 1). Notably, at all follow-up examinations there was no statistically significant difference between screening and baseline OSBP or 24-h ASBP (Table 1), indicating stability of BP in the run-up to uCB resection. Also, there were no changes in MSNA (frequency or incidence), BRS, heart rate variability, HVR and minute ventilation (Table 1), or medication between screening and baseline (Supplemental Table 2). We next determined if there was a proportion of patients that showed a reduction in BP and whether this correlated with any other measured variable.

Proportionating and characterizing responding patients

We defined a responder as a participant with evidence of glomus cells in the resected tissue (as determined histologically) and a ≥10-mm Hg drop in ABP at the 3-month follow-up visit to allow time for patient recovery and the resolution of adverse events. A nonresponder was defined as having evidence of glomus cells in the resected tissue, but did not show a ≥10-mm Hg fall in ambulatory BP at 3-month follow-up. On the basis of these definitions, there were 8 responders (i.e., 53.3%; 95% confidence interval: 26.6% to 78.7%) who showed significant reductions in ASBP at 3 months and 6 nonresponders (Figure 3); glomus cells were not found in 1 patient (see the previous text). No patient with a left CB resected and/or prior renal nerve denervation responded. In the 8 responders, day ASBP decreased at the 3-month (−23 ± 3 mm Hg; p = 0.0005), 6-month (−26 ± 4 mm Hg; p = 0.0021), but not 12-month (−12 ± 8 mm Hg; p = 0.22) follow-up compared with screening (Figure 3A). Night-time ASBP was also reduced at 3-month (−20 ± 4 mm Hg; p < 0.0243), 6-month (−16 ± 5 mm Hg; p < 0.047) and at 12-month follow-ups (−15 ± 6 mm Hg; although p = 0.067) compared with screening (Figure 3B) with 24-h ASBP following a similar time-course (Figure 3C). Comparing both day and night ASBP between responders and nonresponders, there were significant differences between patient groups at all time points (Figures 3A and 3B) (p < 0.05 to p < 0.01). Regarding OSBP, responding patients exhibited a reduction at 1 month (−52 ± 8 mm Hg; p = 0.006), 3 months (−46 ± 8 mm Hg; p < 0.005), and 6 months (−35 ± 7 mm Hg; p < 0.0088), but not at 12 months compared with both screening and baseline (Figure 3D). There was no change in the variability of ASBP. For ASBP and OSBP, there were no sex-related differences (Fisher exact test). The HBP was also reduced in responders but not nonresponders at 3 months (Supplemental Figure 1).

Antihypertensive medications

Among responders, across all study time points, there was a significant difference in whole-dose equivalents (WDE) (p = 0.0009; 4.5 ± 0.6 WDE at screening and baseline, falling to 3.5 ± 0.6 WDE by 6 and 12 months). There was also a trend toward reductions in both the number of medications (p = 0.06) and medication classes (p = 0.06). The were no changes in WDE (0.98), the number of medications (p = 0.15), or medication classes (p = 0.15), among nonresponders.

MSNA following CB resection

Neither baseline MSNA burst frequency nor incidence differed between responders and nonresponders (Table 2, Figures 4A and 4B) (p = 0.31 and p = 0.46, respectively). However, compared with baseline, responders exhibited a decrease in total MSNA burst area/min at 3 months (−374 ± 102%·s/min; p = 0.0137) and 6 months (−520 ± 135%·s/min; p = 0.0296), but not at the 12-month follow-up (p = 0.74) (Figure 4C). Furthermore, total MSNA burst area/min in the responders was lower than in nonresponders at 3-month (−374 ± 108%·s/min vs. 281 ± 174%·s/min; p < 0.0154), 6-month (−165 ± 135%·s/min vs. 16 ± 193%·s/min; p < 0.025) and at 12-month follow-up (−166 ± 38%·s/min vs. 386 ± 123%·s/min; p = 0.06) (Figure 4C). In contrast, total MSNA burst area/min compared with baseline did not change at any follow-up time in the nonresponders (Figure 4C).

Baroreflex modulation of MSNA

In responders, BRS improved as revealed by a decrease at 6-month (−1.50 ± 0.18%·s/mm Hg; p < 0.0245), but not at 12-month follow-up (−0.96 ± 0.63%·s/mm Hg; p = 0.52) compared with baseline (Figure 4D). There was no change in BRS relative to baseline in nonresponders at any time point (Figure 4D) (p > 0.05).

Characteristic distinctions between responders and nonresponders

Before surgery and compared with nonresponders, responders had a higher hypoxic ventilatory response (p < 0.027) and faster ventilatory frequency (p < 0.025) (Table 2) at baseline consistent with higher peripheral chemoreflex sensitivity and drive, respectively. Moreover, they consistently had the right carotid body removed.

CB dysfunction prevalence in hypertension

Across all patients, there were no changes in ABP or OBP after uCB resection. This was not surprising given that hypertension has a heterogeneous etiology and chemoreflex hyper-reflexia was reported in only a subset of patients with cardiovascular disease (18). Hence, we performed a post hoc analysis using a BP reduction of >10 mm Hg in daytime ASBP at 3 months post-uCB resection to identify if a proportion of participants had responded, and if so, whether they exhibited any distinct physiological features. We found a substantial reduction in daytime ASBP in 8 patients (>−20 mm Hg) that persisted for at least 6 months. The BP response after uCB resection wanes at 12 months, which may reflect compensation including that from the other CB. Notably, the level of BP in responders remained significantly lower relative to nonresponders at 12 months. It is encouraging that in these responding patients, medications were reduced compared with baseline, which may have blunted the magnitude of the BP response. Longer-term follow-up data is now crucial to determine the persistence of the BP-lowering effect following uCB resection. Nevertheless, the reduction in both ABP and medication suggests that CB modulation therapy may go beyond a pharmacological treatment for a subset of patients.

Mechanism of action of CB resection

In the responders (but not the nonresponders), MSNA total activity was reduced over a similar time course to the ABP, indicating reduced vasomotor tone; this is consistent with data from hypertensive rats (13) and data seen transiently using hyperoxia to inhibit CB afferent activity in humans with hypertension (12). We also noted an improvement in MSNA BRS in the responder but not in the nonresponder patients, which could contribute mechanistically to the lowering of BP. This finding is consistent with the known antagonism between peripheral chemoreceptor and baroreflex function (10).

The elevated HVR and higher respiratory frequency in the responders versus nonresponders is supportive of aberrant hyper-reflexia and high CB drive, respectively. The elevated HVR is consistent with peripheral chemoreceptor hyper-reflexia in hypertension, as reported in animals (13) and humans with hypertension 8, 11. It may be possible to use HVR and respiratory frequency (and hyperoxia to depress the CB) (12) to pre-select the patients with hypertension who are most likely to benefit from CB modulation therapy, assuming that their arteries and arterioles are able to vasorelax.

Study limitations

As a first-in-man, safety and feasibility study, we did not include a control group. However, in the 1 patient in whom we found no evidence of glomus cells in the resected tissue, we failed to see a fall in BP. Notwithstanding the recruitment difficulties associated with a surgical intervention, we acknowledge that the absence of a control arm and the low number of patients are limitations of the present study. A percutaneous catheter-based approach to ablate the CB selectively may be better tolerated, for example.

Although patients were fully assessed in specialist hypertension clinics before study entry, drug non-adherence cannot be ruled out. Notably, we fully adhered to the established drug adherence procedures at the time the study was initiated (see the Methods sections). There were no statistical differences between screening and baseline OSBP, which may partly overcome concerns regarding a Hawthorne effect and regression to the mean. Comparing total MSNA from the same patient at different times can be problematic, but we controlled for this by referencing to the largest burst in the neurogram.

Reasons for dysfunctional CB

The issue of why the CB develops hypersensitivity is unclear, but may include hypoperfusion as suggested in heart failure (2). In the hypertensive condition, this may be a result of stenosis caused by atheroma or arteriole hypertrophy; the latter may be induced by elevated sympathetic activity to the CB and/or angiotensin II. Inflammation is also present in the CB of hypertensive rats (28) and may trigger release of cytokines, chemokines, and reactive oxygen species that could also contribute to its hyperactive state. In rodents subjected to chronic intermittent hypoxia, a condition causing hypertension and reflected in humans with sleep apnea, there is a change in the balance of gasotransmitters including H₂S and CO (29) and purinergic mechanisms (30); whether these occur in the CBs of human patients with hypertension is unknown.

Turning to the nonresponders, there are several explanations for lack of BP improvement. Anatomical variations in the distribution of the CB may result in insufficient glomus tissue being resected. We acknowledge the heterogeneous etiology of hypertension and would not expect a BP response in individuals with nondysfunctional CBs or mechanisms unrelated to CB dysfunction. Even in patients with CB dysfunction, it may be that the contralateral CB predominates in driving up BP. We do not rule out stiffened arteries that are unable to regain compliance even with a reduction in sympathetic tone.