Bertges DJ, Muluk V, Whittle J, Kelley M, MacPherson DS, Muluk SC. Relevance of Carotid Stenosis Progression as a Predictor of Ischemic Neurological Outcomes. Arch Intern Med. 2003;163(19):2285-2289. doi:10.1001/archinte.163.19.2285
The progression of carotid stenosis may be a better predictor of adverse neurological outcomes than a single measurement of stenosis in asymptomatic patients.
Retrospective review of prospectively collected data from a noninvasive vascular surgery laboratory between 1988 and 1997 at a Veterans Affairs Medical Center. A total of 1701 carotid arteries from 1004 asymptomatic patients were prospectively followed by duplex ultrasonographic scanning. Carotid arteries treated with endarterectomy were excluded. The main outcome measures were ipsilateral transient ischemic attack (TIA) and cerebrovascular accident (CVA).
The baseline degree of carotid stenosis was less than 50% of artery diameter in 75% of patients. The annual rates of ipsilateral TIA and CVA were each 3.3%. When categorized with respect to carotid artery, the annual rates of ipsilateral TIAs and CVAs were 2.0% and 2.1%, respectively. Univariable Cox proportional hazards modeling showed that both baseline carotid stenosis and progression of stenosis were significant predictors of the composite outcome TIA and CVA, as well as the outcome CVA alone. In multivariable modeling, the progression of carotid stenosis was a highly significant predictor of the composite outcome TIA and CVA (risk ratio [RR], 1.68; P<.001) and of CVA alone (RR, 1.78; P<.001). However, baseline stenosis was found to be a significant predictor of time to the combined outcome (RR, 1.29; P = .01) but not of CVA alone. Clinical risk factors did not add any additional predictive information.
The progression of carotid stenosis assessed by serial duplex scanning is a better predictor of ischemic neurological events than a single measurement of stenosis.
SEVERITY OF CAROTID stenosis is known to be an important predictor of transient ischemic attacks (TIAs) and cerebrovascular accidents (CVAs) in both symptomatic and asymptomatic patients.1- 5 For symptomatic patients with severe disease, clinical decision making is often guided by the finding that there is a significant reduction in CVA risk following carotid endarterectomy (CEA).1,3 However, the benefit of CEA is not as great for asymptomatic patients2,6; therefore, better prediction of future ischemic neurological events would be of great value in improving patient selection for intervention.
Although a single determination of the degree of carotid stenosis by duplex ultrasonography has predictive value, there is much room for improving the ability to predict neurological outcomes.2,5 The predictive value of serial duplex studies has been directly or indirectly addressed,5,7- 14 but conflicting conclusions have been reached by different authors. The present study is one of a series of reports analyzing a large cohort of patients undergoing prospective serial duplex surveillance at a large Veterans Affairs hospital. In previous work,15,16 we identified clinical and duplex findings that predict the time to progression of carotid stenosis. In this report we analyze the relationship of the progression of carotid stenosis to the occurrence of ischemic neurological outcomes. Our goal was to determine whether progression of carotid stenosis is a useful predictor of adverse neurological events. A related objective was to determine the utility of serial duplex surveillance of asymptomatic patients. Our strategy was to compare the predictive value of static determinations of stenosis with that of serially updated measures of stenosis.
From September 1988 to September 1997, 6775 carotid duplex studies were performed in 4171 patients at the noninvasive vascular laboratory of the Pittsburgh Veterans Affairs Medical Center. We identified 1004 patients who (1) were asymptomatic at the time of the initial study, (2) had at least 1 follow-up study more than 6 months after the baseline study, and (3) had at least 1 carotid artery in which CEA had not been performed. Asymptomatic was defined as the absence of TIA, amaurosis fugax, or CVA in the 6 months prior to the baseline study. Patients are referred to the laboratory by internists and surgeons for various indications. The laboratory schedules routine follow-up appointments for all patients at intervals of 12 to 18 months regardless of clinical status, because of a policy to prospectively track the clinical and ultrasonographic course of these patients' condition. Data on carotid arteries that had undergone CEA prior to the baseline study were excluded from analysis. In addition, the data of patients who underwent CEA after the baseline study were censored from this analysis at the time of CEA. Data were included up to the time of CEA if the patient had a serial duplex follow-up evaluation prior to surgery. Therefore, TIA or CVA occurring as a consequence of CEA was not included. As noted in a previous publication,16 our policy in the last 8 to 10 years has been to offer CEA to good-risk patients with severe (≥80%) stenosis. However, several severe asymptomatic lesions were followed up in patients who declined surgery or in patients with significant medical risk factors. Carotid arteries occluded at baseline were excluded from the analysis.
At each visit to the vascular laboratory a registered nurse obtained a detailed neurological history as well as yes/no responses to questions about smoking, hypertension, hyperlipidemia, diabetes, angina, and myocardial infarction. Blood pressure measurements were obtained from both arms, and a carotid duplex study was performed using an ultrasound color Doppler system (Accuson 128XP; Accuson Inc, Mountain View, Calif). The degree of internal carotid artery (ICA) stenosis was determined from velocity criteria validated at our institution by comparison with contrast angiographic assessment.16 The degree of carotid stenosis was determined from the ICA to common carotid artery (CCA) ratio of the peak systolic velocity. The degree of stenosis, as percentage of artery diameter, corresponds to the following ICA/CCA ratios: none (ratio, 0.1-1.4); mild (15%-49%; ratio, 1.5-1.9); moderate; (50%-79%; ratio, 2.0-3.9); severe (80%-99%; ratio ≥4.0); and occluded (no flow in ICA). Using angiography as gold standard, sensitivity and specificity ranged from 70% to 99%. We found excellent overall agreement between duplex and angiographic studies (κ
Recorded clinical end points were hemispheric or retinal TIAs and CVAs. Transient ischemic attack was defined as a transient (<24 hours) episode of monocular blindness or hemispheric neurological deficit. Cerebrovascular accident was defined as blindness or hemispheric deficit persisting for more than 24 hours. For this report, only data on the first TIA and/or first CVA ipsilateral to each carotid artery followed up by serial study were analyzed. We separately examined the outcomes of first ipsilateral CVA and first ipsilateral CVA or TIA.
Neurological outcome data were derived from vascular laboratory records, based on the neurological history taken by the nurse. In addition, medical records were reviewed for clinical follow-up to supplement vascular laboratory records when they were available only in electronic format. The neurological history recorded by the nurse was validated by a blinded examination of 241 medical charts randomly selected. For the purposes of this validation, the patients were categorized into 1 of 5 mutually exclusive groups (left hemisphere TIA, left hemisphere CVA, right hemisphere TIA, right hemisphere CVA, and no event). In the case of multiple events affecting the same patient, the first identified event was used. The same classification scheme was applied based on the vascular laboratory records, and the resulting 5 × 5 contingency table was used to calculate a kappa statistic (κ = 0.81; 95% confidence interval, 0.72-0.89), indicating excellent agreement between the nurses' neurological history and the history determined from medical record review.17
As described previously, patient deaths (n = 362) during the period of the study were tracked using a validated Veterans Affairs national database, the Beneficiary Identification and Record Locator Subsystem, intended to track veteran mortality and beneficiaries.18 For patients who died during the period of the study, hospital records were examined to determine whether the patient had a CVA or TIA between the time of the last vascular laboratory study and death. We were able to obtain necessary records for 71% of the deaths.
We used statistical methods for censored survival data, measuring survival time as the time from first study to outcome. The followed-up vessel was censored at the time of CEA, the date of last available complete clinical data, or the date of study termination (September 1997). Our goal in this analysis was to examine the predictive value of stenosis data from the vascular laboratory. Therefore, if the last available clinical follow-up visit was at the time of a vascular laboratory study, the stenosis data from that study were excluded from analysis but the clinical information was used. Cox regression was used to determine if baseline stenosis or stenosis progression were significant predictors of time to outcome.
Because the data collected represented measurements from carotid arteries on both sides in many patients, the SDs were adjusted for clustering by patient using the Huber-White robust estimator of variance for the Cox model.19 The correlation between stenosis on the left and right sides among patients having both carotid arteries under study was low (r = 0.14). Thus, the adjustment for clustering by patient using the Huber-White robust estimator of variance was considered to be a sufficient strategy to control for covariance between sides.
Baseline stenosis was measured as a continuous variable with values reflecting the category of stenosis; stenosis was also used as a time-dependent covariate representing the change in stenosis from baseline to the current time at each repeated assessment.
Descriptive statistics were calculated using SPSS software (SPSS Inc, Chicago, Ill). Life table and proportional hazards analysis were performed with Stata software (Stata Corporation, College Station, Tex). Life table plots were carried out to the time point when the SD reached 10% of the value of the survival distribution function. Statistical significance was inferred at the .05 level in all cases.
The study prospectively followed 1701 carotid arteries in 1004 patients. With the exception that most of these patients were male, baseline demographic and clinical parameters and baseline carotid artery stenosis distribution were similar to those of patients evaluated in most large peripheral vascular laboratories. Demographic and clinical parameters are shown in the following tabulation:
Baseline carotid artery stenosis distribution is shown in the following tabulation:
The mean time between the baseline study and the last duplex study was 2.1 years (range, 0.5-10 years), and the mean number of studies per carotid artery was 3 (range, 2-15). The mean interval between successive duplex scans was 15.4 months, and the mean interval between baseline study and last complete clinical follow-up was 2.7 years (range, 0.5-10.5 years). Among patients who had 1 or more events, the mean time between the last duplex study and the composite TIA or CVA end point was 6.2 months, and the mean time between the last duplex study and the CVA end point was 8.0 months.
The distribution of baseline stenosis values, shown in the above tabulation, indicates that 75% of the carotid arteries that were followed up had less than 50% stenosis at the initial duplex study. Only 10% of patients had a severe carotid artery lesion, with 80% to 99% stenosis at baseline. The progression of stenosis by category of baseline stenosis and mean time to progression are shown in Table 1.
As shown in Table 2, the annual event rate was approximately 2% each for CVA and TIA ipsilateral to the carotid arteries under study. An annual event rate of 3.3% each for ipsilateral CVA and TIA was observed when events were categorized by patient. Several clinical and demographic factors were examined as potential predictors of the composite outcome "TIA or CVA" and CVA alone. We performed the analysis separately for the left and right carotid arteries, and the results were similar to those obtained when we analyzed all carotid arteries together (left and right treated as independent variables).
In Cox regression analysis, baseline stenosis was found to be a significant predictor of time to the combined outcome "TIA or CVA," but not of time to CVA alone (Table 3). In contrast, a change in category of carotid stenosis (a time-dependent measure) was a highly significant predictor of both outcomes. Appropriate tests of the proportional hazards assumption as well as an examination of the survival distribution function indicated that the proportional hazards model was appropriate for both baseline and change in stenosis variables.
We also examined each of the following clinical variables in multivariable modeling with the stenosis variables: age, smoking status, race, diabetes, hypertension, blood pressure, angina history, myocardial infarction history, and serum cholesterol level. None of these variables had any significant predictive value in multivariable modeling, indicating that they did not add any predictive information.
Based on the model, the predicted cumulative event rates, by year of follow-up, for "TIA or CVA" and CVA alone are shown in Table 4 and Table 5. For each baseline category of stenosis, the finding of stenosis progression resulted in an increasing rate of neurological events. As an example, the tables show that progression from none/mild to moderate stenosis resulted in adverse neurological event rates similar to those of a severe stenosis that did not progress.
The randomized clinical trials comparing CEA with medical therapy have focused exclusively on a single measurement of carotid stenosis. Although asymptomatic patients with severe stenosis benefit from CEA, the benefit is less in them than in symptomatic patients.1,2 A better understanding of the risk associated with progression of carotid stenosis may improve selection criteria for asymptomatic patients. The objective of this study was to determine whether the progression of carotid stenosis, as detected by serial duplex ultrasound evaluation, is a better predictor of adverse ipsilateral neurological events than a single, initial measurement of stenosis.
In 2 previous reports we identified clinical and duplex study variables that predict the time to progression of carotid artery stenosis. Clinical etiologic factors of progression included pulse pressure and high-density lipoprotein levels.15 Baseline ICA stenosis was the most important duplex finding that predicted the progression of stenosis.16 To our knowledge, this is the largest study to date that describes neurological outcomes after long-term serial duplex surveillance of asymptomatic patients. The study is limited by selection bias in that the population was based on patients referred to the vascular laboratory of our institution. However, they were representative of patients with cerebrovascular disease except for the higher rate of smoking and predominance of male patients that is inherent in the Veterans Affairs population. The demographic and comorbid conditions analyzed did not add any predictive value to that of the stenosis variables.
The study allowed us to determine the effect of the progression of carotid artery stenosis on neurological events because most patients (75%) had an initial stenosis less than 50% of artery diameter. Some adverse neurological events could have been due to other causes because we did not exclude patients with valvular heart disease or arrhythmias. However, we limited the outcomes to retinal or hemispheric events to focus on events in the normal carotid distribution. At our institution, good-risk, asymptomatic patients with severe stenosis (80%-99%) are offered CEA. This in itself did not affect the number of neurological events detected because data on carotid stenosis were excluded from the analysis after the performance of CEA. High-risk patients and patients who refused CEA were still available for ongoing analysis. The annual rate of of 3.3% for first TIA or CVA was similar to that reported by others.2,20- 22 We also analyzed the composite outcome of TIA or CVA to improve the statistical power of the analysis.
The most important finding of the study is that the progression of carotid stenosis is a highly significant predictor of subsequent ipsilateral neurological events (Table 3). Progression of carotid stenosis predicted time to CVA while baseline stenosis did not (P = .11). The fact that baseline stenosis was not predictive of CVA is not necessarily surprising because the cohort was dominated by patients with baseline stenosis less than 50% of artery diameter and because the follow-up interval was long. The RR for first TIA or CVA and for first CVA was higher for progression of stenosis than for any single measure of stenosis. The RRs shown in Table 3 represent the effect of a 1-unit change in stenosis. The risk rises in a multiplicative fashion for patients with 2- or 3-unit changes.
These results are in contrast to the findings of Lewis et al5 who reported a secondary analysis of the Asymptomatic Cervical Bruit Study (ACBS). The authors concluded that progression of carotid stenosis was a poor predictor of neurological events and suggested that serial duplex ultrasonography was not beneficial in asymptomatic patients. The discrepancy can be explained by 2 factors. First, our patients had a higher rate of progression. As previously reported,16 the annualized risk of stenosis progression, defined as an increase in 1 stenosis category on duplex scanning, was 9%, compared with approximately 6% in the ACBS. A second possible explanation for the difference in findings is that a greater proportion of our study vessels had baseline lesions in the none-to-mild (<50%) stenosis category. Three quarters of the carotid arteries in the present study were in this category at baseline, compared with half of the arteries in the ACBS. This could reduce the predictive value of baseline stenosis and increase the predictive value of stenosis progression. Finally, the larger size of the present study may also increase the power to detect the predictive power of progression.
Over the 10-year period of this study, 38% of patients with mild stenosis progressed to moderate stenosis. An interesting finding was that progression of stenosis from mild to moderate was associated with the same relative risk of adverse neurological events as a severe stenosis that did not progress. While we do not necessarily advocate CEA in these patients, the progression of stenosis should alert the clinician to a significant risk of TIA or CVA. In future trials, the progression of carotid stenosis may serve as a better selection criterion for CEA for asymptomatic patients.
In constructing the model, we assumed that each progression to a higher category was associated with the same added risk of a neurological event. It is possible that progression from a mild to moderate stenosis carries a different risk than progression from a moderate to severe stenosis. The assumption of equal added risk for progression at different levels is inherent in the proportional hazards model, and it appears justified by the fact that standard statistical tests demonstrate that the model is appropriate for our data.
Over the study period, 34% of all neurological events were associated with progression of carotid stenosis. Because the mean time between the last duplex scan and the index neurological event was between 6 and 8 months, and our mean study-to-study interval was 15 months, we may have missed identifying the progression of stenosis before an event occurred in a significant number of patients. Thus, the true risk of progression may be higher than reported here. Our data are also not well suited to measure the effect of the rate of progression of carotid stenosis. It is possible that lesions that progress rapidly over a short time present a higher risk. The effect of rate of progression on outcome deserves further study. Some data were inevitably lost in this study because of patients who received their care at other institutions. We believe that this had only limited impact, however, because we had a large number of patients who continued their follow-up in our laboratory over a long time. Another cause of lost data is patient death due to causes that we were unable to ascertain. We were able to identify the cause of death of 71% of the patients who died during the study period. A final limitation of our study is that we cannot comment on whether serial duplex surveillance can improve outcomes in a cost-effective manner. Additional study will be needed to make this determination.
The progression of carotid stenosis is a highly significant predictor of subsequent adverse neurological events in asymptomatic patients. These data may indicate a benefit for following up asymptomatic patients with serial duplex surveillance. While the initial degree of carotid stenosis is important, patient selection for CEA may be improved by factoring the progression of stenosis into clinical decision making.
Corresponding author: Satish C. Muluk, MD, Division of Vascular Surgery, Allegheny General Hospital, 320 E North Ave, Pittsburgh, PA 15224 (e-mail: firstname.lastname@example.org).
Accepted for publication December 19, 2002.