Incidence of Perioperative Stroke After Neck Dissection for Head and Neck Cancer

Abstract

Objective:

To determine the incidence of perioperative stroke in patients undergoing a neck dissection.

Summary Background Data:

The incidence of perioperative stroke in non-head and neck surgery is between 0.08 and 0.2%. In contrast, a critical review of the literature identified 2 studies stating the incidence of perioperative stroke in head and neck surgery to be 3.2% and 4.8%. The implications of these results are significant because they suggest a potential need for preoperative screening and/or intervention for carotid artery pathology.

Methods:

This historical cohort study was conducted using discharge data for all neck dissections performed in a geographically-defined health region in Alberta, Canada, from 1994 to 2002. Subjects were selected for study if they had an assigned ICD-9CM procedure code for a neck dissection at one of the region's 3 adult-care hospitals. Our main outcome measure was perioperative stroke.

Results:

Patients (n = 499) were identified as having had a neck dissection (mean age 56.5 ± 15.3 SD, 65.3% male). Seven patients had ICD-9CM codes for postoperative central nervous system complications (incidence of 1.4%). However, on chart review, only one had had a true perioperative stroke corresponding to an incidence of 0.2% (95% confidence interval 0.01, 1.12). No missed strokes were found in a confirmatory random review of 10% of charts.

Conclusions:

The incidence of perioperative stroke in this study is significantly lower than that previously stated in the literature. This suggests that preoperative screening and/or intervention for carotid artery disease may not be necessary in this patient population.

Two studies exist in the current literature that indicate that the incidence of perioperative stroke in patients undergoing a neck dissection is quite high: 3.2% and 4.8%. The implications of these results are significant because they suggest a need for preoperative screening and/or intervention for carotid artery pathology. A database analysis was performed on all neck dissections performed in a geographically defined region to verify the above results.

Perioperative stroke during head and neck surgery is a devastating complication with high morbidity and mortality. Carotid artery stenosis is the single most important risk factor for stroke.¹ Moderate carotid stenosis (50–79% of the vessel's diameter) carries an annual risk of stroke of 2.1%.^2,3 Just over 15% of these patients will progress to severe carotid stenosis (80–99%) and an annual risk of stroke of 6.9%.^2,3 Carotid artery stenosis is related to a number of risk factors, namely hypertension, peripheral vascular disease, diabetes mellitus, smoking, atherosclerotic heart disease, male gender, age, and external irradiation to the head and neck for malignancy.^4–6 Patients undergoing a neck dissection for head and neck cancer may have many of these risk factors, thus predisposing them to perioperative stroke. In addition, a neck dissection may involve hemodynamic instability, blood loss, and exposure and manipulation of the vascular and neurologic structures of the neck, all of which may increase the risk of perioperative stroke.

The reported incidence of perioperative stroke in non-head and neck surgery varies between 0.08% and 0.2%.⁷ During carotid endarterectomy, the risk has been reported to be between 0.9% and 3.6%.^7,8 After a critical review of the literature, only 2 studies were found that addressed the incidence of perioperative stroke during head and neck surgery. They reported the risk of stroke during neck dissection to be 3.2% and 4.8%.^7,9

The implications of these results are significant because they suggest a potential need for preoperative screening and/or intervention for carotid artery pathology.^10–15 We therefore initiated this study to verify the above results by determining the incidence of perioperative stroke after neck dissection for head and neck cancer in a large Canadian city.

METHODS

Selection of Patients and Variables

After approval from our Conjoint Health Research Ethics Board, all patients who underwent a modified or radical neck dissection at any 3 of the adult hospital sites in Calgary, Alberta, from 1994 to 2002 were identified. All data were extracted from the Calgary Health Region's centralized administrative hospital discharge database. We obtained the following patient information: age, gender, hospital number, date of admission, date of discharge, date of surgery, up to 16 diagnosis codes, and up to 10 procedure codes.

Using codes from the clinical modification of the International Classification of Diseases, 9th revision (ICD-9CM),¹⁶ we identified patients who had the following codes: radical laryngectomy (ICD-9CM code 30.4), radical neck dissection, not otherwise specified (40.40), and radical neck dissection, unilateral (40.41), and bilateral (40.42). Any patient who had undergone resection of the carotid artery was automatically excluded from the study (38.32, 38.42, and 38.62).

We also identified any variables that are risk factors for carotid artery stenosis. The following variables were isolated: diabetes (ICD-9CM codes 250.1–250.7), peripheral vascular disease (443.9), old myocardial infarction (412), and cerebrovascular disease (430–438). Perioperative stroke was our main outcome measure. The following diagnostic codes were flagged, and any charts with one of these codes was pulled for more detailed review: occlusion and stenosis of precerebral arteries (ICD-9CM code 433.x), occlusion of cerebral arteries (434.x), acute stroke (436), late effects of cerebrovascular disease (438.x), and iatrogenic/postoperative stroke (997.02). To ensure that no strokes were missed, we also included nervous system complication, unspecified (997.00), central nervous system complication (997.01), and other nervous system complications (997.09).

Chart Audit

All coded strokes were subsequently verified by detailed chart review to confirm the diagnosis code and to identify variables that are predictors of stroke. Disease-specific variables identifiable by ICD-9CM codes have already been outlined. We also identified the following variables: smoker, previous irradiation therapy to the neck, carotid bruit, modified versus radical neck dissection, hypertension, and length of surgery. A random chart review of 10% of all neck dissections without recorded stroke codes was also performed to confirm the absence of postoperative stroke among patients with no ICD-9CM diagnosis codes for stroke.

Statistical Analysis

The incidence of perioperative strokes was determined by dividing the number of perioperative strokes by the total number of neck dissections performed. Exact 95% confidence intervals were calculated. A priori, we had also considered doing regression analyses of baseline characteristics that are associated with (or “predictive of”) the occurrence of stroke. However, the low observed incidence of perioperative stroke precluded such an analysis.

RESULTS

From 1994 to 2002, 499 neck dissections were identified from the Calgary Health Region administrative hospital discharge database. The mean age of the patients was 56.5 years (±15.3 SD). The demographics of the study patients are shown in Table 1. The majority of patients (65.3%) were male. Pulmonary disease (18.2%) and hypertension (17.4%) were the most common comorbidities, followed by diabetes mellitus (4.8%), previous myocardial infarction (4%), congestive heart failure (3.2%), peripheral vascular disease (2.6%), and cerebrovascular disease (1.2%).

TABLE 1. Associated Morbidity in 499 Patients Undergoing a Neck Dissection

Seven patients had an ICD-9CM code for postoperative central nervous system complications, which raises the possibility of an incidence as high as 1.4% (Table 2). However, after detailed review of the pertinent charts, only 1 of these patients had suffered an iatrogenic perioperative stroke. The other 6 patients all had codes corresponding to central nervous system complication (ICD-9CM code 997.01) and other nervous system complications (997.09). Upon careful chart review, it was confirmed that these codes did not specify the occurrence of a stroke but indicated postoperative delirium, metabolic or anoxic encephalopathy, or nerve palsy. Furthermore, any computed tomography scan of the head completed on this subset of patients revealed negative results (ie, no evidence of stroke or other focal pathology). The incidence of perioperative stroke in these 499 patients was therefore only 0.2% (95% confidence interval 0.01%, 1.12%).

TABLE 2. Number of Patients With ICD-9CM Codes Corresponding to a Neurological Complication or a Stroke

A chart review was performed to assess the possibility that some perioperative strokes might have been missed. Fifty charts without a coded stroke were pulled (10% of our sample size), and each chart was reviewed for any evidence of perioperative stroke. No missed strokes were identified.

DISCUSSION

This is the largest series in the literature assessing the incidence of perioperative stroke after a neck dissection. We found the incidence to be 0.2%, with an upper 95% confidence limit of only 1.12%. Two articles exist that discuss the incidence of stroke in neck dissection.^7,9 Nosan et al⁹ described 5 patients who suffered a postoperative stroke after undergoing head and neck surgery. Their case series was drawn from a center where 105 patients had undergone a neck dissection between 1989 and 1991, yielding an incidence estimate of 4.8%. Rechtweg et al⁷ described the authors’ experience with simultaneous carotid endarterectomy and neck dissection in 3 patients requiring surgery for metastatic neck disease. In their discussion section, the authors report an unpublished personal communication from Yoo et al, who reviewed 441 patients undergoing neck dissections in Toronto, Canada, and found an incidence of perioperative stroke of 3.2%.

Our results are substantially different from those of the 2 aforementioned studies. The possible explanations for this are several-fold. First, it is clear that important limitations exist in the 2 previous studies. In the first article,⁹ the authors did not clearly state how they identified the 5 cases of perioperative stroke from the total number of cases performed at their hospital nor why they chose to include only those cases performed between 1989 and 1991. Furthermore, details of the 100 patients without stroke were not provided. Given the small number of total cases, confidence intervals (although not provided) are likely to be quite large. In the second article,⁷ the incidence of perioperative stroke was merely cited in the form of a personal communication between the author and a group of investigators. Therefore, we have no information regarding patient profiles, methods, or statistics. It is thus fair to conclude that the quoted incidences in the above 2 articles must be viewed with caution.

Second, it is possible that surgical technique was different in our region than in the regions where the 2 previous studies were performed. Variation in surgical technique might limit the amount of hyperextension and rotation of the neck, and therefore the incidence of carotid artery intimal tear(s) and thrombus formation or plaque ulceration from turbulent flow. The amount of carotid artery retraction might also contribute to the incidence of stroke, as this has the potential to dislodge a thrombus or plaque. However, it seems unlikely that a difference in surgical technique caused such dissimilarity in the incidence of perioperative stroke among the various studies. Subjects identified by our regional database were operated on by at least 6 different surgeons, 4 otolaryngology specialists and 2 general surgeons, in 3 different hospitals.

Our study has limitations as well. It is an historical cohort drawing on hospital discharge data, and it is certainly possible that data-entry errors may have occurred. Although the sensitivity of stroke codes is not perfect, studies in the literature suggest that the sensitivity is one of the best of any condition in ICD-9CM.¹⁷ Furthermore, our chart audit provided reassurance that no missed strokes occurred. Of 50 charts, none had a missed stroke code. Although an initial incidence of coded neurologic events of 1.4% was found, chart review revealed that only 1 of these 7 patients had had a true perioperative stroke. The other 6 patients all had ICD-9CM codes for nervous system complications, all of which were attributable to postoperative delirium, metabolic or anoxic encephalopathy, or nerve palsy.

In conclusion, our study suggests that the incidence of perioperative stroke in patients undergoing a neck dissection for head and neck cancer is only 0.2%, with an upper 95% confidence limit of 1.12%. This is significantly lower than the previously stated risks of 3.2% and 4.8%.^7,9 A risk of 0.2% is equal to the risk of perioperative stroke in patients undergoing non-head and neck surgery.⁷ This therefore suggests that special preoperative screening and/or intervention for carotid artery pathology in this patient population is not needed, as it would be unlikely to provide much benefit in the face of such a low perioperative stroke rate.