Worldwide, >650,000 new cases of head and neck cancer are diagnosed each year.1 Among these patients, 20% to 30% will have distal metastases in addition to local or regional recurrences.2,3 The treatment of advanced head and neck cancer, depending on the subsite of disease, has trended from surgery to radiation therapy or chemoradiation as the initial modality in an attempt to preserve aerodigestive structure and functionality.4 Patients with persistent disease after initial treatment or with recurrence may subsequently undergo surgery.
Physicians have long been aware that head and neck radiation therapy (HNRT) impairs sympathovagal balance and carotid sinus baroreflex5–8 and therefore influences blood pressure (BP) and heart rate (HR) regulation. Radiation-induced impairment of hemodynamic regulation is a long-term dynamic process. However, most patients suffer only subtle damage to hemodynamic regulation after HNRT, and changes in their BP and HR in response to emotional and/or physical stress are largely regulated. General anesthesia has a mixed influence on sympathetic neural outflow and hemodynamic stability. Positive pressure ventilation may activate the sympathetic nervous system to compensate for decreased cardiac filling resulting from positive intrathoracic pressure. On the other hand, intravenous administration of propofol has not only decreased sympathetic nervous system activity but also blunted the HR response to hypotensive challenge.9 Volatile anesthetics, except for desflurane that transiently increase sympathetic nervous system activity via a quick concentration surge, decrease peripheral vascular tone and suppress baroreflex responses to hypotension.10 Therefore, patients who suffer from subtle damage to hemodynamic regulation induced by HNRT may have profound BP and HR changes in response to noxious stimulation under general anesthesia. However, the impacts of HNRT on the regulation of HR and BP under general anesthesia remain unclear.
We retrospectively reviewed a group of patients with oral cavity or oropharyngeal malignancies initially treated with HNRT with or without concurrent chemotherapy and subsequently treated with surgery. We then compared BPs and HRs during anesthesia induction, skin incision, and anesthesia emergence with the corresponding BPs and HRs in patients treated upfront with surgery. Our primary goal was to assess the associations of general anesthesia with intraoperative hemodynamic measurements in patients with a history of HNRT.
The institutional review board of the University of Texas MD Anderson Cancer Center (UTMDACC) approved this study. Patients were identified by a head and neck surgery registry at UTMDACC including 4011 adults (≥18 years of age) who had primary oral cavity or oropharyngeal cancer at all stages. All of the patients underwent surgery under general anesthesia with tracheal intubation from 2007 to 2012. The study exclusion criteria consisted of documented autonomic neuropathy, cardiomyopathy with impaired left ventricular ejection fraction due to any cause (ischemic-, nonischemic-, and chemotherapy-related factors), concurrent HNRT or chemotherapy at the time of surgery, the presence of an implanted cardiac device (a pacemaker or defibrillator), atrial fibrillation, and fiber-optic airway intubation while awake or asleep.
Of the 4011 patients, 236 (30 female and 206 male) who received intensity-modulated radiation therapy alone or concurrently with chemotherapy before surgery and met our inclusion criteria were identified and placed in the treatment group. Also, 1640 patients (192 female and 1448 male) in the registry who had surgery only and met the inclusion criteria for the study were identified and formed a pool for selection of controls. However, for some of these patients, electronic data on confounding factors potentially related to HR and BP changes, including history of chemotherapy and daily antihypertensive use, were unavailable for retrieval. Therefore, a cohort was not readily available for use as a control group, and a control group could not be randomly created without replacement of the missing data. To avoid manual data clarification and replacement for all the patients in the pool, age, sex, and body mass index (BMI) were chosen as matching factors with a ratio of 1:1 to select 236 controls. The missing data on the controls were handpicked from corresponding database including chemotherapy and the list of home medications and placed in the analysis data after matching. The matching range for age was ±5 years, and the treatment and control groups were matched according to BMI at 5 levels: up to 25 kg/m2, 26 to 30 kg/m2, 31 to 35 kg/m2, 36 to 40 kg/m2, and >40 kg/m2. Each patient in the treatment group was successfully matched with a control patient.
Patient characteristics, including age, sex, BMI, American Society of Anesthesiologists (ASA) physical status score, type of daily oral antihypertensive taken (β-blockers versus non–β-blockers), history of chemotherapy, history of HNRT (bilateral versus unilateral), and interval from HNRT to surgery, were collected. Baseline hemodynamic variables for each patient, including systolic BP (SBP), diastolic BP (DBP), and HR measured during preoperative assessment (typically 1–2 days before surgery), were retrieved from the anesthesia database at MD Anderson. Also, all 3 of these hemodynamic variables collected electronically during surgery at 5-minute intervals were obtained from the same database to obtain the baseline variables. BP values measured using BP cuffs and HR readings recorded using a pulse oximeter were the primary sources for these variables. Missing BP and HR data were replaced with the nearest available readings at corresponding time points. To avoid the influences of volume shift on the HR and BP changes during surgery, data on hemodynamic variables 30 minutes after anesthesia induction and 15 minutes after first skin incision were collected (significant bleeding is unlikely in this interval). In addition, owing to large variation in hemodynamic changes in response to anesthetic withdrawal, BP and HR data obtained 20 minutes before tracheal extubation were included in the study.
Anesthetic data were not included in our analysis. However, as per our routine practice, the agents typically used for anesthesia induction are propofol and fentanyl; rocuronium (or succinylcholine) is used to facilitate endotracheal intubation. Anesthetic maintenance is achieved with an inhaled agent. For patients undergoing head and neck surgeries, desflurane is the most commonly used inhaled agent in our practice.
Summary statistics were calculated, including mean values, standard deviations, and ranges for continuous variables such as age, BMI, and interval from radiation therapy to surgery. The frequency counts and percentages for categorical variables such as sex and irradiation status were summarized. The Fisher exact test or a χ2 test was used to evaluate associations between 2 categorical variables. The Wilcoxon rank sum test was used to evaluate differences in continuous variables between patient groups. Mean and standard error plots over time for each BP and HR measurement were generated for the treatment and control groups. A multivariable model of repeated measures was used to examine the associations of HNRT with intraoperative HR and BP after adjusting for baseline HR and BP, the time effect, and other covariates potentially associated with intraoperative HR and BP changes, including use of β-blockers, history of chemotherapy, and ASA score. BP and HR measurements were collected every 5 minutes for 30 minutes during anesthetic induction, for 15 minutes during incision, and for 20 minutes during emergence. Time was treated as a continuous variable in the model. The baseline measurements of BP and HR were not included in the outcome vector and only used as adjustment for baseline. Linear mixed-effects models with unstructured covariance structure (UN) were used for BP and HR outcomes. We explored 3 covariance structures: unstructured covariance structure (UN), compound symmetry, and autoregressive (1), and chose the most appropriate structure to better fit the data. In our analysis, unstructured covariance structure (UN) had lower values for AIC and BIC compared with compound symmetry and autoregressive (1). The interaction between treatment and time was assessed and included in the multivariable model if the interaction was significantly associated with BP or HR. The relationships between treatment and outcomes were assessed by collapsing over time.
All tests were 2-sided. P values < .05 were considered statistically significant. All analyses were conducted using the SAS (version 9.4; SAS Institute, Cary, NC) and S-Plus (TIBCO Software, Palo Alto, CA) statistical software programs.
In total, we included 472 patient records in our retrospective review. The median age of the patients was 58.7 years (range, 31.0–84.0 years). Half of them received HNRT alone or with concurrent chemotherapy as the initial treatment and subsequently underwent surgery, whereas the other half received upfront surgery. The patient distribution according to sex was equal in the 2 groups: 30 female (13%) and 206 male (87%) patients. The mean (± standard deviation) ages of the patients in the treatment and control groups were 59.1 ± 9.0 and 58.3 ± 9.2 years, respectively (P = .364). The mean (± standard deviation) BMIs in the treatment and control groups were 25.57 ± 4.66 and 25.45 ± 4.40 kg/m2, respectively (P = .873). The standardized differences in age and BMI between the 2 groups were 8.32% and 7.41%, respectively. We identified no intraoperative mortality or cardiac arrest information in the data.
The distribution of the 2 groups according to baseline characteristics is presented in Table 1. Compared with the control group, the treatment group had higher overall ASA scores (P = .0412). Also, 28% of the control patients and 17% of the treatment patients were taking β-blockers (P = .0041). In addition, 45% of the control patients and 28% of the treatment patients were taking multiple antihypertensives (P = .0001).
Of the 236 patients with a history of HNRT, 86% underwent chemoradiation, with the remaining 14% undergoing HNRT only. Furthermore, 214 patients (91%) underwent bilateral irradiation, whereas 22 patients (9%) underwent unilateral treatment. The median interval from irradiation to surgery was 93 days (range, 21–2121 days). Additionally, 39 patients (17%) in the control group received preoperative chemotherapy. Age, sex, and BMI were well balanced between the 2 groups. We adjusted for the unbalanced covariates, consisting of ASA score, β-blocker use, and history of chemotherapy, in our multivariable model.
The baseline hemodynamic variables in the 2 groups are summarized in Table 2. In comparing the baseline BPs and HRs in the 2 groups, we found that the treatment group had a significantly higher mean resting HR (P = .0012) and significantly lower mean SBP (P < .0001). However, the difference in mean DBP was not significant (P = .6411).
Figure shows mean and standard error plots of BP and HR measured in real time in the 2 patient groups at baseline, anesthesia induction, skin incision, and anesthesia emergence. The P values were based on a model of repeated measures collapsed over time for comparison of the 2 groups.
During anesthesia induction, the patients in both groups underwent continuous declines in BP during the first 15 minutes after onset of anesthetic induction. Patients given HNRT presented with significantly lower SBPs than did the corresponding control patients (P < .0001; column 1 of Figure). The DBP values did not differ significantly between the 2 groups (P = .6382). The HR values during anesthesia induction also did not differ significantly between the 2 groups (P = .4732). All of the patients presented with mild elevations in HR at 5 minutes after anesthesia induction, reflecting HR acceleration due to endotracheal intubation. Whereas the control patients had a mean (± standard deviation) HR increase of 4.4 ± 14.1 bpm, the patients given HNRT had a mean (± standard deviation) HR increase of 1.9 ± 12.5 bpm (P = .0474) indicating that the patients in the treatment group had decreased HR responses to endotracheal intubation. The difference in HR at 5 minutes after anesthetic induction between the groups was −3.0 (95% confidence interval [CI], −5.7 to −0.2).
During incision, the BP readings in the treatment group were significantly lower than those in the control group (P < .0001 in SBP, P = .0162 in DBP; column 2 of Figure). The largest BP discrepancy between the 2 groups was at the time of skin incision, indicating that the patients in the treatment group had reduced hypertensive responses to incision. The differences in the HR readings between the 2 groups were not statistically significant. However, the HRs in the treatment group were below the baseline at all time points, whereas those in the control group were above the baseline during the first 10 minutes.
During emergence from anesthesia, all of the patients had continuous increases in HR and BP toward the time of tracheal extubation, reflecting increased HR and BP responses to anesthesia withdrawal (column 3 of Figure). However, the SBP readings were markedly lower in the treatment group than in the control group at all time points (P = .0066), whereas DBPs (P = .4711) and HRs (P = .5113) did not differ between the 2 groups.
Use of a mixed model to examine the association of daily antihypertensive use, chemotherapy use, and ASA score respectively with intraoperative HR and BP demonstrated that neither β-blocker nor multiple antihypertensive use daily was significantly associated with HR or BP (P > .05). The ASA scores and overall effects of chemotherapy on BP and HR during anesthesia were not significant (P > .05).
We used multivariable mixed models to examine the associations of the treatments (HNRT plus surgery versus surgery alone) with HRs and BPs after adjusting for baseline HR and BP, time, daily β-blocker use, chemotherapy, ASA score, and interaction of time and treatment (if the interaction was statistically significant) (Table 3). HR values were significantly associated with treatment (HNRT plus surgery versus upfront surgery). Patients who received previous HNRT had significant decreases in HR at anesthesia induction (−2.21 [95% CI, −4.42 to −0.01]; P = .0492) and incision (−2.66 [95% CI, −5.61 to −0.16]; P = .0373) but did not significantly differ in HR during anesthesia emergence (−3.01 [95% CI, −6.31 to 0.29]; P = .0740) compared with the patients in the control group. Likewise, the patients who had undergone HNRT had significant decreases in SBP during anesthesia induction (−6.88 [95% CI, −10.94 to −2.78]; P = .0011) and incision (−15.87 [95% CI, −20.45 to −11.29]) but no significant decreases in SBP during anesthesia emergence (−5.42 [95% CI, −10.94 to 0.11]; P = .0547) compared with the patients in the control group. The association between HNRT and DBP decreases was only significant during incision (−6.50 [95% CI, −9.47 to −3.53]; P < .0001).
Baroreflex sensitivity is directly connected to HR response to arterial pressure changes.11,12 HNRT is known to impair baroreflex sensitivity and therefore attenuate neuronal responses to BP and HR changes.13 Although full-blown baroreflex failure is rare, decreased baroreflex sensitivity after HNRT is common. Timmers et al7 studied 12 patients who underwent bilateral neck irradiation and found that all of them had markedly decreased HR responses to phenylephrine challenge. Likewise, Huang et al14 studied 89 patients after they underwent neck irradiation and found that none of them had HR responses to deep breathing or the Valsalva maneuver. However, none of the patients in either study presented with daily hemodynamic instability. Applying these findings to our patients who received previous HNRT, although their baseline HR and BP were grossly within normal limits, the significantly increasing baseline HR and decreasing SBP plus markedly few patients taking antihypertensives, including β-blockers, were highly suggestive of pathologic changes in HR and BP regulation. Other authors have reported that patients with compromised baroreflex sensitivity had elevated HR and BP.15 However, our patients presented with increased HR but decreased SBP. We speculate that hypovolemic hypotension resulting from poor oral intake due to pain or difficulty swallowing was the cause of the SBP differences between our patients and the patients in their study. Indeed, both clinical reports6,8,12 and our experience indicated that lightheadedness or dizziness and orthostatic hypotension are common in patients with a history of HNRT. The BP changes due to reduced baroreflex sensitivity in the present study were likely superseded by hypovolemic hypotension in patients who underwent HNRT. However, regardless of the reason, preoperative hypotension markedly decreased the tolerance to anesthetics in patients with a history of HNRT.
While the impact of HNRT on baroreceptor sensitivity is generally accepted, the effects of HNRT on hemodynamic stability under general anesthesia are less clear; identifying this impact was the impetus for this study. We found that compared with BP changes in the controls, patients who had a history of HNRT presented with significant intraoperative hypotension. Although differences in the measured HRs between the 2 groups were not statistically significant, HRs in the patients with a history of HNRT decreased significantly from baseline, indicating that anesthetics impose not only hypotensive but also negative chronotropic effects on patients with a history of HNRT. Furthermore, that significantly fewer patients in the treatment group than in the control group took β-blockers supported that the negative chronotropic effect did not result from daily pharmacologic management but rather from organic injuries. Most anesthetics attenuate baroreflex in a dose-dependent manner by interfering with sympathetic neural outflow.10,16 However, as we observe a decrease of HR in response to administration of a phenylephrine bolus and increasing HR after increased blood loss during anesthesia in patients with an intact baroreflex, the depressive effects on baroreflex of anesthetics should be only partial, and the coupling of baroreceptor stimulation and effector responses is grossly maintained under general anesthesia.17 Therefore, we expect to see some interplay between HR and BP changes during noxious stimulations such as endotracheal intubation and skin incision. Overall negative chronotropic HR responses and a lack of corresponding BP elevation in response to noxious stimulation in patients given HNRT were highly suggestive of impairment of baroreflex-mediated sympathetic cardiac acceleration and sympathetic muscle nerve activity in these patients. Decreases in HR and BP in addition to lack of appropriate hemodynamic responses to surgical stimulation or volume changes in a patient with preexisting hypovolemia during anesthesia may quickly result in sustained hypoperfusion. Indeed, others have reported severe hypotension and bradycardia or cardiac arrest under general anesthesia in this population.18,19
The most significant finding in our study was that general anesthesia induced a significant negative chronotropic effect on HR in patients with a history of HNRT. This was seen during anesthetic induction and skin incision, when HRs in the treatment group generally remained below baseline, whereas those of the control group remained above baseline. Although autonomic neuropathy and baroreflex damage are different clinical entities, in our study, the BP and HR changes during anesthesia in patients given HNRT were similar to the hemodynamic responses to anesthesia typically seen in patients with diabetic autonomic neuropathy. Given their impaired autonomic HR and BP regulation, diabetic patients are prone to development of intraoperative hypotension as well as bradycardia.20,21 Therefore, the strategies used for perioperative hemodynamic management in diabetic patients may also be effective in patients given HNRT.
The significant impact of general anesthesia on hemodynamic stability through a negative chronotropic effect in patients who underwent HNRT should heighten an anesthesia provider’s awareness and make one prepared for rapid development of substantial bradycardia in these cases, particularly those with concurrent hypotension. Earlier intervention for bradycardia may prevent development of hemodynamic instability.
Because cancer therapies are highly patient selective, we could not randomly assign our patients to different treatment groups. In addition, owing to the uneven distribution of confounding factors in our relatively small sample size, we could not always appropriately manage the bias due to treatment indications and subject distributions to produce a valid comparison between the 2 groups.
In summary, the most significant finding in our study was that anesthetics imposed a significant negative chronotropic effect on HR in patients with a history of HNRT. Preoperative hypovolemia exacerbates the hypotensive effects of general anesthesia in this population. When significant bradycardia and hypotension arise altogether, substantial hemodynamic deterioration may develop rapidly. Therefore, one should be watchful for bradycardia in patients who had undergone HNRT, particularly the cases with concurrent systemic hypotension. Early intervention for bradycardia is warranted to avoid hemodynamic instability.
Name: Gang Zheng, MD.
Contribution: This author helped with study design, data collection/verification, discussion, and writing and revision of the manuscript.
Name: Wenli Dong, MD, MS.
Contribution: This author helped with study design, data analysis, discussion, and writing and revision of the manuscript.
Name: Carol M. Lewis, MD, MPH.
Contribution: This author helped with study design, data contribution, discussion, and revision of the manuscript.
This manuscript was handled by: Scott M. Fishman, MD.
1. Marur S, Forastiere AAHead and neck cancer: changing epidemiology, diagnosis, and treatment. Mayo Clin Proc. 2008;83:489–501.
2. Licitra L, Felip EESMO Guidelines Working Group. Squamous cell carcinoma of the head and neck: ESMO clinical recommendations for diagnosis, treatment and follow-up. Ann Oncol. 2009;20(suppl 4):121–122.
3. Forastiere A, Koch W, Trotti A, Sidransky DHead and neck cancer. N Engl J Med. 2001;345:1890–1900.
4. Gokhale AS, Lavertu PSurgical salvage after chemoradiation of head and neck cancer: complications and outcomes. Curr Oncol Rep. 2001;3:72–76.
5. Timmers HJ, Karemaker JM, Lenders JW, Wieling WBaroreflex failure following radiation therapy for nasopharyngeal carcinoma. Clin Auton Res. 1999;9:317–324.
6. Sharabi Y, Dendi R, Holmes C, Goldstein DSBaroreflex failure as a late sequela of neck irradiation. Hypertension. 2003;42:110–116.
7. Timmers HJ, Karemaker JM, Wieling W, et al.Arterial baroreflex and peripheral chemoreflex function after radiotherapy for laryngeal or pharyngeal cancer. Int J Radiat Oncol Biol Phys. 2002;53:1203–1210.
8. Ketch T, Biaggioni I, Robertson R, Robertson DFour faces of baroreflex failure: hypertensive crisis, volatile hypertension, orthostatic tachycardia, and malignant vagotonia. Circulation. 2002;105:2518–2523.
9. Sellgren J, Ejnell H, Elam M, Pontén J, Wallin BGSympathetic muscle nerve activity, peripheral blood flows, and baroreceptor reflexes in humans during propofol anesthesia and surgery. Anesthesiology. 1994;80:534–544.
10. Neukirchen M, Kienbaum PSympathetic nervous system: evaluation and importance for clinical general anesthesia. Anesthesiology. 2008;109:1113–1131.
11. Robertson D, Hollister AS, Biaggioni I, Netterville JL, Mosqueda-Garcia R, Robertson RMThe diagnosis and treatment of baroreflex failure. N Engl J Med. 1993;329:1449–1455.
12. Mano T, Iwase SSympathetic nerve activity in hypotension and orthostatic intolerance. Acta Physiol Scand. 2003;177:359–365.
13. Timmers HJ, Wieling W, Karemaker JM, Lenders JWCardiovascular responses to stress after carotid baroreceptor denervation in humans. Ann N Y Acad Sci. 2004;1018:515–519.
14. Huang CC, Huang TL, Hsu HC, et al.Long-term effects of neck irradiation on cardiovascular autonomic function: a study in nasopharyngeal carcinoma patients after radiotherapy. Muscle Nerve. 2013;47:344–350.
15. Smit AA, Timmers HJ, Wieling W, et al.Long-term effects of carotid sinus denervation on arterial blood pressure in humans. Circulation. 2002;105:1329–1335.
16. Ebert TJ, Harkin CP, Muzi MCardiovascular responses to sevoflurane: a review. Anesth Analg. 1995;81:S11–S22.
17. Keyl C, Schneider A, Hobbhahn J, Bernardi LSinusoidal neck suction for evaluation of baroreflex sensitivity during desflurane and sevoflurane anesthesia. Anesth Analg. 2002;95:1629–1636.
18. Truong AT, Sturgis EM, Rozner MA, Truong DTRecurrent episodes of asystole from carotid sinus hypersensitivity triggered by positioning for head and neck surgery. Head Neck. 2013;35:E28–E30.
19. Sorgente A, Sarkozy A, Chierchia GB, de Asmundis C, Capulzini L, Brugada PCarotid sinus hypersensitivity following radiotherapy delivery in a patient with bilateral glomus jugular tumour. J Cardiovasc Med (Hagerstown). 2011;12:675–677.
20. Burgos LG, Ebert TJ, Asiddao C, et al.Increased intraoperative cardiovascular morbidity in diabetics with autonomic neuropathy. Anesthesiology. 1989;70:591–597.
21. Maser RE, Lenhard MJCardiovascular autonomic neuropathy due to diabetes mellitus: clinical manifestations, consequences, and treatment. J Clin Endocrinol Metab. 2005;90:5896–5903.