chat analysis of the influence of specific risk factors on late results after carotid endarterectomy

CHAT analysis of the influence of specific risk factors on late results after carotid endarterectomy Eugene F. Bernstein, MD, PhD, Jeffrey H. Kaplan, MD, Terese E. Scala, James A. Koziol, PhD, and Ralph B. Dilley, MD, La Jolla, Calif.

The CHAT classification separates various current and historical presentations of cerebrovascnlar disease in an effort to determine important prognostic clues for management and prognosis. To evaluate known risk factors for late stroke and death, we followed up for an average of 44 months 633 patients who had undergone 714 carotid operations. We analyzed the indication for surgery (by CHAT) and the effect of preoperative risk factors (age, hypertension, cardiac disease, tobacco use, diabetes, hyperlipidemia, renal disease, pulmonary disease, and total risk factor score) on the end points of late stroke and death. Ipsllateral stroke was uncommon after carotid endarterectomy: with life-table analysis, the probability of late stroke at 5 years after carotid endarterectomy was 3%. Among the 127 patients with amaurosis fugax, the incidence of late stroke and of mortafity was a combined total of 1% per year, and the 17 patients who had been first seen with permanent ocular stroke (blindness) fared equally well. The 28 patients who were first seen with vertebrobasilar symptoms and were treated by carotid endarterectomy also fared particularly well, with no late strokes or deaths within the first 5 years. Logistic regression analyses revealed that the various indications for carotid endarterectomy were associated with differing patterns of risk factors as siown~ficant predictors of late stroke or death. For patients first seen with asymptomatic lesions, only diabetes was an important predictor for late stroke (p -- 0.05) and renal disease was the only marker for early death (p = 0.05). On the other hand, those factors were not significant risk factors for patients first seen with amaurosis fugax, for whom tobacco use was a negative predictor for stroke (p -- 0.06) and male gender a negative predictor for early death (p = 0.03). After cortical transient ischemic attacks and carotid endarterectomy, there were no risk factors predictive of late stroke or of death. For patients with prior stroke, age was a very strong predictor of stroke (p -- 0.01) and both age and a history of cardiac disease were significant risk factors for early death (p -- 0.007). In contrast to the results in reports of patients treated medically for transient ischemic attacks and stroke, we found that several risk factors appeared to play relatively minor roles. In conclusion, stroke after carotid endarterectomy was uncommon, least common after ocular symptoms, and most likely after permanent cortical stroke. Specific risk factors were less important for patients after carotid endarterectomy than for the medically treated stroke patient. (J VAse SURG 1992;16:575-87.)

The CHAT classification of vascular disease was designed to separate various current (C) and histor- ic (H) presentations o f disease and their separate angiographic (A) and target (T) organ manifestations

From the Division of Vascular and Thoracic Surgery, Scripps Clinic and Research Foundation, La Jolla.

Presented at the Seventh Annual Meeting of the Western Vascular Society, Maui, Hawaii, Jan. 11-15, 1992.

Reprint requests: Eugene F. Bernstein, MD, Scripps Clinic and Research Foundation, 10666 N. Torrey Pines Rd., La Jolla, CA 92037.

24/6/39071

from each other with the aim of identifying specific factors or groups of factors that would be important in determining appropriate management approaches or prognosis. When initially described, the system provided information about the improved outlook of patients with amaurosis fugax compared with that o f patients with cortical transient ischemic attacks (TIA). These two groups of patients had previously been grouped within the overall classification o f TIA. ~ In addition, the prognosis for patients without symptoms but with prior contralateral symptoms was shown to be different from that of patients

575

5 7 6 Bernstein et al.

Journal of VASCULAR

SURGERY

Table I. The CHAT classification of stroke*

C Current status (< 1 yr)

symptoms Vascular territory

H History (> 1 yr)

Symptoms of vascular territory

A Artery

Site Pathology

0 - Asymptomatic

1 - Brief stroke TIA ( < 24 hr)

2 - T e m p o r a r y stroke with fifll recovery (24 hr to 3 wk)

3 - Permanent stroke, mi- nor ( > 3 wk)

4 - Permanent stroke, major ( > 3 wk)

5-Nonspec i f ic dysfunction 6 - Improving stroke 7 - Fluctuating stroke 8 - Deteriorating stroke

a - Carotid ocular (amaurosis fugax)

b - Carotid cortical

Same categories as current status

c - Vertebrobasilar i-5, a-e

d - O t h e r focal

e - Diffuse Subscript s is used to indicate prior operation

0 - No lesion

1 - Appropriate lesion

2 - L e s i o n only in another vascular pathway

3 - Combined, appropriate lesion and lesion in another vascular pathway

a - Arteriosclerosis

c - Cardiogenic embolus

d - Dissection

e - Aneurysm

f-- Fibromuscular

r - Arteritis t - Trauma

AVM, Arteriovenous malformation; significant, 50% stenosis or disease thought to be the source of symptoms. *After Baker JD, Rutherford RB, Bernstein EF, et al. J VASC SURG 1988;8:721-9.

with a completely negative contralateral history of cerebrovascular disease.

The CHAT classification system was adopted by the Subcommittee on Reporting Standards for Cere- brovascular Disease of the Society for Vascular Surgery/International Society for Cardiovascular Surgery in 19887 In addition, that committee advocated the use of a risk factor severity scale for the cerebrovascular area that had been proposed by the Subcommittee on Reporting Standards for Lower Extremity Ischemia. The availability of these two schemes-a classification of the presentation of cerebrovascular disease and a risk factor classification- suggested the evaluation of the relation of specific risk factors and a total risk factor score with the long-term outlook after successful carotid endarterectomy (CEA). This approach seemed particularly appropriate because of the growth of the number of reports emphasizing the importance of such risk factors on the likelihood of future stroke and early death. 4s In addition, a larger data base with longer follow-up was available to provide long-term data in our patients.

METHODS

To evaluate the relation of known risk factors with late stroke and early death after CEA, we analyzed the indication for surgery by CHAT in 633 patients who had undergone 714 carotid operations (CEA) and had been followed up in the Vascular

Registry of the Scripps Clinic and Research Foun- dation. All operations were performed by two surgeons who have used the same operative indications and technical surgical methods since the open- ing of the institution for surgery in 1978. All patients who had carotid surgery also underwent a completion angiographic study to document the technical status of the repair at the completion of their operative procedure. Other details of operative tech- nique have been published previously. 9

The chart of each patient in the registry was reevaluated, and a CHAT classification and risk factor grading was performed as of the time of initial clinical presentation (Table I). Risk factors studied included diabetes, tobacco use, hypertension, cholesterol lev- els, and cardiac, renal, and pulmonary disease. A cardiac workup was routinely based on a history or abnormal electrocardiogram (ECG). The grading scale used was that recommended by the Subcom- mittee and ranged from 0 to 3 for each factor (Table II). s A total risk factor (TOT) score was determined by adding the grade score for each of the seven risk factors. In addition, analyses of the effect of age (by decade) and of gender were performed. Follow-up of all vascular surgical patients in our institution is a routine function of the ongoing vascular registry. All patients were requested to be seen at 3, 6, and 12 months and annually thereafter. Both a clinical examination and duplex study were performed at each visit. All diagnoses of stroke were

Volume 16 Number 4 October 1992 CHAT analysis of risk factors after CEA 577

T Target

Site Pathology

0 - N o lesion

1 -Appropr ia t e lesion

2 - L e s i o n only in another vascular territory

3 - Combined, appropriate lesion and lesion in another vascular territory

h - Hemorrhage

i - Infarct

j - L a c u n a r

m - A V M

n - N e o p l a s m

q - Other r - Retinal embolism

confirmed by a neurologist. All analyses were based on the end points of ipsilateral stroke (after hospital discharge from carotid surgery), death, and stroke- free survival.

All data concerning the condition of the patients when first seen, details of the operative procedure, early morbidity, and late follow-up were entered in a VAX model 11/750 computer (Digital Equipment Corp., Maynard, Mass.) of the General Clinical Research Center of the Scripps Clinic with the use of CLINFO software (BBN Software Products Corp., Cambridge, Mass.) for clinical data management and analysis. Life-table analyses were performed according to the product limit method of Kaplan and Meier 1° and comparisons between groups were made with Gehan's modification of the Wilcoxon testY Categorical data in contingency tables were analyzed with Pearson's chi-square test.

Logistic regression analyses were performed to assess the relative significance of the various risk factors as predictors of outcome events at 5 years, ipsilateral stroke, and death) 2 For these analyses, all graded risk factors were dichotomized into 0-1 (negative-positive) variables. In particular, this facil- itated the reporting of odds ratios and the associated 95% confidence intervals for these variables. The estimated odds ratio approximated how much more likely it was that the particular outcome (stroke or death) would occur among those with the factor present (i.e., positive) than among those with the factor absent (i.e., negative).

RESULTS

Risk factor incidence. In the 633 patients who were operated on and studied for an average of 44 months thereafter, late ipsilateral stroke was observed in 16 and late contralateral stroke was noted in 33. Four of these contralateral strokes occurred in the group of 81 patients in whom the contralateral side had also been subjected to endarterectomy and 29 were seen in hemispheres that had not been operated on. In the total patient group, the incidence of hypertension was 59.7%, tobacco use 52.9%, cardiac disease 47.3%, hyperlipidemia 38.8%, pulmonary disease 30.6%, diabetes 15.0%, and renal disease 14.2%.

Univariate CHAT analysis. An initial assessment of the Kaplan-Meier life-table analyses according to the major clinical symptom at presentation is summarized in Table III, in which the 5-year probability of remaining stroke-free, of surviving, and of surviving stroke-free is documented. The table also separates the patients according to the CHAT scheme. The group of patients free from symptoms is first presented as a single entity, and then is subdivided into subgroups on the basis of their history of contralateral symptoms. The small group with a history of contralateral permanent stroke had a higher likelihood of late stroke, compared with those with no history of contralateral disease (Fig. 1).

Patients who were first seen with TIA were subdivided by CHAT into those with amaurosis fugax, anterior cortical lesions, and vertebrobasilar symptoms. In addition, varying grades of preoperative stroke were also separated. The most striking finding of this analysis was the relative absence of ipsilateral stroke in all groups, with the probability of remaining stroke-free in all patients 97% at 5 years. The worst outlooks for recurrent stroke were that after surgery for an asymptomafic lesion in patients with a history of a contralateral cortical stroke with permanent deficit and that after cortical stroke with minor residual deficit as the indication for ipsilateral carotid surgery. Survival data document a mortality rate of approximately 5% per year in patients with all indications for carotid surgery, except for those with minor ocular deficit, in which no deaths occurred (p < 0.001).

TOT score. The likelihood of stroke or death by 5 years after CEA as a function of the TOT score was examined. In Table IV, these results are summarized for all patients and indicate no effect on stroke for those with a TOT score from 0 to 9 and no statistically significant effect with any score. On the other hand, there was a significant effect of each

578 Bernstein et al.

Journal of VASCULAR

SURGERY

Table II. Risk factor categories ~

0 1 2 3

None Adult onset, no insulin Adult onset, insulin Juvenile onset controlled

None (or abstinence None currently, absti- Current, < one Current _> 1 pack/day > 10 yr) nence 1-10 yr pack/day (or absti-

nence < 1 yr) Hypertension Distolic < 90 mm Hg Easily controlled, single Requires two drugs

drug Hyperlipidemia Cholesterol and tfiglyc- Mild elevation, diet

erides within controlled normal limits for age

&symptomatic, normal ECG

Diabetes

Tobacco use

Cardiac &symptomatic, remote MI ( > 6 mo), occult MI by ECG

Types, II, III, or IV re- quiring strict diet control

Stable angina, controlled ectopy, asymptomatic arryrtmfia, drug-compensated congestive failure

Renal No renal disease, creati- Creatinine 1.5-3.0 Creatinine 3.1-5.9 nine < 1.5 mg/dl, mg/dl, clearance 30- mg/dl, clearance 15-30 clearance > 50 49 ml/min ml/min ml/min

Pulmonary &symptomatic, normal &symptomatic or mild Between 1 and 3 chest x-ray film, PFT dyspnea on exertion, > 80% of predicted mild x-ray parenchy-

mal changes, PFTS 65%-80% of predicted

> Two drugs or uncon- trolled

Requires drug control

Unstable angina, symptomatic or poorly controlled ectopy or dysrythmia, poorly compensated failure, MI < 6 mo

Creatinine >6.0 mg/dl, clearance < 15 ml/min, dialysis or transplant

Vital capacity < 1.85 L, FEV 1 < l . 2 L o r <35% of predicted, maximal voluntary ventilation < 28 L/nfin or < 50% of predicted, Pco 2 > 45 mm Hg, sup- plemental 02 required, pulmonary hypertension

M/, Myocardial infarction; PFT, pulmonary function tests; FEV, forced expiratory volume in 1 second; Pco2, carbon dioxide pressure. ~From Baker JD, Rutherford RB, Bernstein EF, et al. J VASC SURG 1988;8:721-9.

increment of the TOT score on the likelihood of early death (p < 0.01) and consequently on stroke-free survival (p < 0.01).

When all those patients operated on for asymptomatic lesions were grouped together (Table V), there was no effect on stroke related to TOT score. However, in groups limited to patients with global asymptomatic states (e.g., no symptoms in either hemisphere, CHAT classification CoHo), the patients with TOT scores of 4 to 5 and 10 to 14 were significantly more likely to experience late stroke (p < 0.01). An analysis of the TOT score in patients with prior minor cortical stroke (CHAT classification C3b) failed to show any influence of risk factor on late stroke. In contrast to the relative absence in each of these TOT analyses of an effect of risk factors on late stroke, there was a highly significant effect of TOT score on the probability of early death.

Univariate risk factor analysis. The effect of each of the specific risk factors was then analyzed for each clinical presentation by CHAT. In Table VI the effect of varying risk grades for tobacco use, hypertension, and cardiac disease are presented for the

patients without symptoms. There was no obvious or statistically significant effect of any of these individual factors on the likelihood of stroke, survival, or stroke-free survival. Similar results were observed in those patients with cortical TIA (Table VII) and all degrees of stroke (Table VIII).

In Table IX, CHAT analyses of the data with arterial and target organ risk factors are summarized. None of the comparisons indicated any significant risk factor predictive value for stroke after CEA. However, more advanced degrees of arterial disease (CHAT classification A3) and a positive computed tomographic (CT) or magnetic resonance imaging (MRI) scan for infarction (CHAT classification Tli) were correlated with an increased probability of early death.

Multivariate .logistic analysis. Results of the logistic regression analyses are presented for two outcomes at 5 years and for different CHAT classi- fications in Tables X (stroke) and XI (survival). In each instance, the logistic model fit the data adequately, as assessed by the usual goodness-of-fit chi-square test. Data are presented as standardized

Volume 16 Number 4 October 1992 CHAT analysis of risk factors a#er CEA 579

.Q

.Q o n

1.0

0 .9

0.8

0.7

0 .6

0.5

Stroke-free

...c.o.....a . . . . . . . . . . c o . o

1 ' ; . . . . !

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . t . °

I I i i I i I i I I

0 12 24 36 48 60 72 84 96 100 120 132

Months

Fig. 1. Kaplan-Meier curves depict probability of stroke and of survival in those patients first seen with asymptomatic stenoses who had no history o f any contralateral neurologic symptoms (CoHo) and those who had asymptomatic stenoses with history of contralateral stroke (Coil3).

Table HI . Probability o f stroke-free survival 5 years after CEA*

Probability at 5 years

CHAT Conventional Stoke-free Indication class class No. pa~ents SmTke-free Survival survival

&symptomatic, all cases C O &symptomatic, no prior history CoHo &symptomatic, contralat, amaurosis Coil l &symptomatic, contralat. TIA C0H 2 &symptomatic, contralat, perm. stroke C0H s

Amaurosis fugax Cla Cortical TIA Clb Vertebrobasilar TIA C k

Temp. stroke with flail recovery C2a. c Stroke, minor deficit, ocular Ca, Stroke, minor deficit, cortical C3b

Misc. indications in smaller groups All indications

&symptomatic 180 0.95 0.79 0.75 &symptomatic 144 0.95 0.83 0.80 &symptomatic 6 1.00 0.83 0.83 &symptomatic 8 1.00 0.54 0.54 &symptomatic 15 0.80 0.66 0.53

TIA 125 0.97 0.79 0.81 TIA 155 0.96 0.70 0.70 TIA 29 1.00 0.69 0.68

Stroke, RIND 27 1.00 0.66 0.66 Stroke 16 1.00 1.00 1.00 Stroke 47 0.89 0.67 0.59

43 615 0.97 0.76 0.75

p = 0.01

RIND, Reversible ischemic neurologic deficit. *Excludes nine cases of permanent operative strokes out of 714 operations (1.3%).

coefficients, i.e., coefficients divided by standard error, for each variable in the relevant logistic regression equation. Positive coefficients denoted increased risk of stroke or death. A standardized coefficient o f less than - 2 or greater than + 2 was significant at approximately the p = 0.05 level. Sex was coded 0 for female and 1 for male, so that negative coefficients indicated males have a decreased risk relative to females for either stroke or death. Data are also presented as p values and odds ratios and lower and upper bounds to 95% confidence intervals for the odds ratios.

In the group free of symptoms, only diabetes had

a significant effect on the likelihood of late stroke (odds ratio 18.8). In amaurosis fugax, no factor was significant, but tobacco use approached this level as a negative predictive factor for stroke (p = 0.06). In the cortical TIA group, there were no significant factors in predicting late stroke, and the same was true when all TIAs, including vertebrobasilar cases, were analyzed. However, when the initial clinical presentation was stroke, age was a strongly predictive factor for late stroke (to = 0.01) with an odds ratio o f 13.6. Borderline, but not significant, results were also noted for tobacco (p = 0.08), cardiac (p = 0.14), and renal (p = 0.13) risk factors.


Journal of VASCULAR

SURGERY

Table IV. Probability of stroke-free survival 5 years after CEA

Indication No. risk factors No. patients Stroke-free


Survival Stroke -free survival

All 0 to 3 253 0.94 4 to 5 172 0.94

6 87 0.91 7 to 9 94 0.94

10 to 14 24 0.83

0.86 I 0 . 8 1 ~ 0.75 O.73 ~ 0.68 0.66 0.66 0.63 0.35 0.28

Braces indicatep < 0.01.

Table V. Probability of stroke-free survival 5 years after CEA


Indication No. risk factors No. patients S t roke- f ree Survival Svroke.froe survival

&symptomatic (Co)

&symptomatic (CoHo)

Stroke, minor, cortical (Csb)

0 to 3 63 0.96 n- 0.95 m- 0.91 4 to 5 53 0.86 [L 0.85 [L. 0.80

6 27 0.95 L 0.75 L 0.69 -1 7 to 9 23 1.00 0.54 0.54 |

10 to 14 16 0.88 0.43 0.36 _]

Total 182 Oto3 45 1.00 I 0.97 ~ 0.97 ~ 4 to 5 47 0.86 0.83 0.79

6 24 0.94 0.64 0.71 7 to 9 16 1.00 0.70 0.70

10 to 14 11 0.82 0.58 0.49 Total 143

0 to 3 21 0.85 0.83 ~ I 0.69 ~] 4 to 5 12 1.00 0.69 0.64

6 8 0.88 0.27 0.27 7 to 9 8 0.71 0.27 0.00

10 to 14 0 N/A N/A N/A Total - ~

Braces indicatep < 0.01. N/A, Not applicable.

Table VI. Probability of stroke-free survival 5 years after CEA: All asymptomatic (Co)


Specific risk factor Degree of risk No. pa in t s Stroke-free Survival Stroke-free survival

Tobacco 0 77 0.91 0.82 0.78 1 30 0.95 0.84 0.79 2 23 0.91 0.79 0.75 3 54 0.95 0.70 0.68

Hypertension 0 66 0.89 0.90 0.81 1 59 0.98 0.85 0.83 2 37 0.94 0.63 0.60 3 21 1.00 0.58 0.58

Cardiac 0 95 0.98 0.89 0.88 1 28 0.96 0.91 0.87 2 39 0.82 0.61 0.48 3 23 0.96 0.60 0.58

None of the risk factors predicted early death in the patients free of symptoms, but renal disease approached significance (Table XI). In patients with amaurosis fugax only female gender predicted early death (p = 0.03)• With cortical TIA, age was a positive predictive factor• However, when the entire

renal, and cardiac disease all were significant predictors of early death. In the patients initially being seen with stroke, age was a very strong predictor of early death (p = 0.007, odds ratio 10.6), as was cardiac disease (p = 0.006, odds ratio 14.3), and renal disease approached significance (p = 0.056, odds


Table VII. Probability of stroke-free survival 5 years after CEA: Cortical TIA (Cxb)


Specific risk factor Degree of risk No. patients Stroke-free Survival Stroke-free survival

Tobacco 0 79 0.93 0.70 0.69 1 23 0.91 0.60 0.57 2 16 0.87 0.67 0.60 3 42 0.97 0.77 0.78

Hypertension 0 58 0.95 0.71 0.69 1 48 0.94 0.67 0.64 2 44 0.94 0.67 0.67 3 10 0.75 0.56 0.58

Cardiac 0 84 0.96 0.80 0.78 1 24 0.96 0.57 0.57 2 41 0.90 0.61 0.60 3 11 0.67 0.37 0.58

Table VIII . Probability of stroke-free survival 5 years after CEA: All strokes (C 2 and C3)


Specific risk fiwtor Degree of risk No. patients Stroke-free Survival Stroke-free survival

Tobacco 0 41 0.89 0.60 0.51 1 12 0.92 0.83 0.73 2 16 0.94 0.72 0.71 3 21 0.90 0.87 0.82

Hypertension 0 37 0.92 0.68 0.63 1 32 0.88 0.70 0.59 2 20 0.95 0.81 0.81 3 1 0.00 0.00 0.00

Cardiac 0 50 0.90 0.85 0.77 1 14 0.93 0.84 0.84 2 24 0.92 0.39 0.33 3 2 1.00 1.00 1.00

Table IX. CHAT analysis of arterial (A) and target organ (T) factors on late stroke and death after CEA

Probability of difference at 5 years (p value)

Groups at risk No. patients Stroke-free Survival Svroke-free survival

All patients A 1 vs A s 372/219 0.42 0.02 0.03 T o vs T 2 127/37 0.97 0.11 0.05 CHAT analysis CoAl vs CoA3 104/714 0.60 0.02 0.09 CoAl vs C2A 1 104/93 0.76 0.02 0.09 C1A1 vs C1A 3 91/34 * 0.08 0.04 C2A 1 vs C2A 3 93/52 0.84 0.94 0.90

CaT o vs C:Tli 21/2 * * * C2T 0 w3 C2Tli 43/6 * 0.23 0.26 CIT . vs C2T:i 21/43 * 0.96 0.98

A1, Appropriate unilateral arterial lesion;As, bilateral arterial lesion; Co, asymptomatic stenosis; C1, ocular TIA (amaurosis fugax); C2, cortical TIA; To, negative CT/MRI scan; T l i , positive and appropriate CT/MRI scan infarction. *Inadequate number of events m permit statistical evaluation.

DISCUSSION

The importance of specific risk factors in the likelihood of the development of stroke has been strongly supported in the literature by studies of patients at risk. Recent summaries of data in patients

at risk have been presented by Wolf et al.,4 Dyken et al.,s and Whisnant,6 and all conclude that there is a strong effect of age, gender, race, and a history of prior stroke on the probability of a subsequent stroke (Table XII). In addition, treatable factors, including

5 8 2 Bernstein et al.

Journal of VASCULAR

SURGERY

Table X. Logistic analysis of the influence of specific risk factors on the probability of late stroke after CEA

Clinical presentation

A_9~nptomatic (Co) Amaurosis fugax (C1,)

Odds Lower Upper Odds Lower Upper Risk factor CoeffSE p value ratio BND* BND* Coq7SE p value ratio BND BND

Gender -0 .245 0.81 0.9 0.51 7.17 - 0 . 7 1 4 0.48 0.4 0.42 E0.01 4.44 Age 0.979 0.33 1.9 0.51 7.17 -0 .276 0.78 0.7 0.81 E0.01 6.67 CHAT pos. history -0 .823 0.41 0.5 0.95 E0.01 2.64 0.543 0.59 1.7 0.23 12.8 Diabetes 2.180 0.03 18.8 1.31 269 0.000 1.00 1.0 0.91 1.06 Tobacco 0.039 0.97 1.0 0.27 3.90 - 1.910 0.06 0.1 0.11 E0.01 1.09 Hypertension - 1.730 0.08 0.4 0.11 1.15 1.300 0.19 4.4 0.46 41.2 Cholesterol 0.964 0.34 1.8 0.56 5.48 - 0.961 0.34 0.3 0.22 E0.01 3.76 Cardiac 1.360 0.17 2.3 0.69 7.53 1.290 0.20 3.9 0.48 32.2 Renal - 0.408 0.68 0.7 0.94 E0.01 4.55 0.000 1.00 1.0 0.14 E0.01 6.03 Pulmonary 0.417 0.68 1.9 0.51 7.17 - 0 . 9 5 4 0.34 0.3 0.22 3.80

CoeffSE, Coefficient divided by standard error; BND, bound; E, error. *95% confidence intervals for odds ratio.

Table XI. Logistic analysis of the influence of specific risk factors on the probability of survival after CEA


Avymptornatic (Co) Amaur0s/sfugax (C1.)

Odds Lower Upper Odds Lower Upper Risk factor CoeffSE p value ratio BND* BND* CoeffSE p value ratio BND BND

Gender - 1.200 0.23 0.6 0.21 1.46 - 2 . 1 6 0 0.03 0.2 0.51 E0.01 0.88 Age 0.566 0.57 1.4 0.45 4.16 1.130 0.26 2.1 0.57 7.83 CHAT pos. history 0.909 0.36 1.6 0.58 4.42 -0 .063 0.95 1.0 0.23 4.04 Diabetes - 0.357 0.72 0.6 0.51 E0.01 7.93 0.434 0.66 2.3 0.49 E0.01 109 Tobacco 0.539 0.59 1.4 0.45 4.12 - 1.460 0.14 0.3 0.60 E0.01 1.53 Hypertension 1.350 0.18 2.0 0.72 5.63 0.082 0.94 1.1 0.27 4.10 Cholesterol - 0 . 2 4 0 0.81 0.9 0.38 2.15 0.585 0.56 1.5 0.40 5.48 Cardiac 1.420 0.16 2.0 0.76 5.20 0.545 0.59 1.4 0.41 4.72 Renal 1.960 0.05 3.0 0.99 8.96 1.170 0.24 2.5 0.53 11.3 Pulmonary 0.792 0.43 1.5 0.57 3.73 1.390 0.17 2.5 0.68 0.09

CoeffSE, Coefficient divided by standard error; BND, bound; E, error. *95% confidence intervals for odds ratio.

hypertension, cardiac disease, prior TIA, diabetes, and an increased hematocrit have been strongly implicated. In the current study, however, after surgical correction of local carotid artery disease, none of these factors appeared to play a significant role in predicting further stroke.

While these data may appear to be contrary to the well-established role documented for risk factors in the pathogenesis of stroke, two other reports offer support for the present data. Whisnant and W e i b e r s , 7

in studying patients after TIA, reported no effect of cardiac disease or diabetes on subsequent stroke, although these had been documented as very important in studies of patients with no prior neurologic

events. In that study, the effects of age, gender, and race were not reported.

In a publication from the Stroke Data Bank, Sacco et al. s evaluated standard risk factors in patients who had sustained a stroke and survived, and only hypertension and diabetes were significant determinants of future stroke. Age, gender, race, cardiac disease, prior TIA, and increased hematocrit were not significant predictors. Those results are similar to the data in the present study. One significant difference between the patients studied by Sacco et al.s and the present report is that all of the patients who were first seen with stroke in this report sustained only minor strokes with full or suhstarttial recovery and the



Cortical TIA (C1~) All TIAs (C1) Stroke (Ca, a., ab)

Odds Lower Upper Odds Lower Upper Odds Lower Upper CoeffSE p value ratio BND BND CoeffSE p value ratio BND BND CoeflSE p value ratio BND BND

-0 .453 0.65 0.7 0.20 2.72 - 1 . 1 9 0 0.23 0.6 0.22 1.44 - 0 . 4 5 9 0.65 0.6 0.59E0.01 5.91 1.560 0.12 2.5 0.78 8.21 1.010 0.31 1.5 0.66 3.59 2.520 0.01 13.6 1.71 108

- 1 . 0 2 0 0.31 0.5 0.12 2.00 - 0 . 4 1 6 0.68 0.8 0.30 2.21 1.450 0.15 3.8 0.60 24.5 -0 .288 0.77 0.7 0.52E0.01 9.03 -0 .302 0.76 0.7 0.81E0.01 5.91 0.000 1.00 1.0 0.02 66.47 -0 .395 0.69 0.8 0.20 2.95 - 0 . 7 4 7 0.46 0.7 0.24 1.92 1.770 0.08 8.8 0.76 102

0.537 0.59 1.4 0.39 5.08 0.769 0.44 1.4 0.58 3.52 -0 .451 0.65 0.7 0.96E0.01 4.38 -0 .891 0.37 0.6 0.17 1.96 -1 .400 0.16 0.5 0.21 1.30 -0 .008 0.99 1.0 0.21 4.65 - 0 . 7 3 9 0.46 0.6 0.15 2.36 0.072 0.94 1.0 0.41 2.60 1.480 0.14 5.2 0.56 47.1 - 0 . 1 6 7 0.87 0.9 0.14 5.26 - 1 . 1 4 0 0.25 0.4 0.84E0.01 1.93 1.510 0.13 5.5 0.58 51.9

1.120 0.26 2.0 0.60 6.55 1.270 0.20 1.8 0.73 4.34 -1 .060 0.29 0.4 0.50E0.01 2.5


Corucal TIA (Cjb) All TIAs (C~) Stroke (Ca, ~., ab)

Odda Lower Upper Odd~ Lower Upper Odds Lower Upper Coef/SE p value ratio BND BND CoeJTSE p value ratio BND BND CoeJTSE p value ratio BND BND

0.519 0.60 1.1 0.45 3.96 -0 .761 0.45 0.8 0.37 1.55 -0 .215 0.83 0.8 0.15 4.71 1.890 0.06 2.6 0.96 6.76 1.960 0.05 1.9 1.00 3.66 2.710 0.01 10.6 1.86 60.6

-0 .279 0.78 0.9 0.30 2.47 0.157 0.88 1.1 0.52 2.16 1.050 0.29 2.3 0.48 10.6 - 1 . 2 0 0 0.23 0.2 0.14E0.01 2.87 -0 .845 0.40 0.4 0.68E0.01 2.93 0.022 0.98 1.0 0.18 5.95

0.159 0.87 1.1 0.37 3.18 - 0 . 6 5 0 0.52 0.8 0.36 1.68 0.664 0.51 1.9 0.27 13.8 0.401 0.69 1.2 0.44 3.48 -0 .040 0.97 1.0 0.51 2.00 -2 .020 0.04 0.2 0.33E0.01 0.978

-0 .838 0.40 0.7 0.25 1.75 -0 .338 0.74 0.9 0.47 1.71 -0 .769 0.44 0.6 0.13 2.46 1.580 0.11 2.4 0.81 7.06 2.030 0.04 2.0 1.02 3.98 2.770 0.01 14.3 2.1 96.7 1.100 0.27 2.0 0.57 7.20 2.230 0.03 2.5 1.11 5.59 1.910 0.06 9.0 0.91 89.4 1.140 0.25 1.7 0.67 4.44 2.560 0.01 2.3 1.21 4.43 - 0 . 3 1 4 0.75 0.8 0.14 4.26

surgical procedures dealt with atheroembolic and hemodynamically induced infarctions. On the other hand, the group reported by Sacco and co-workers included patients first seen with all etiologic factors for stroke, including both parenchymal and subarach- noid hemorrhage, lacuna, cardiac embolism, and other less common causes. Thus the general risk factor background of the two groups may have differed considerably.

The recent report of the European Carotid Surgery Trial also presented risk factor data for those patients allocated to receive no surgery. 13 Three adverse prognostic factors were noted in addition to severe carotid stenosis: history of stroke, residual neurologic deficit, and a positive CT scan for cerebral

infarction. Factors studied but not found to be significant included sex, age, obesity, blood pressure, cholesterol level, diabetes, peripheral vascular disease, and cardiac disease. Patients were not divided by type of clinical presentation with cortical or retinal symptoms, duration of symptoms, or history of contralateral disease. Among those patients allocated to surgery, no risk factor analysis was reported beyond the immediate postoperative period (30 days). These data, which confirm a relatively minor role for risk factors in the probability of stroke when accompa- nied by a high degree of carotid stenosis, support the data of Sacco et al. s documented after stroke and our own data after CEA, but are not in accord with the reports ofWolfet al.,4 Dyken et al.,5 and Whisnant. 6


Journal of VASCULAR

SURGERY

Table XII. Analysis of recent literature regarding stroke risk factors

Before any neurologic event A#~r TIA A#er prior s~oke

Wolff Dyken 6 Whimant 6 Whimant 7 Sacco 8 Risk factor* (1983) (1984) (I 990) (1987) (I 989)

Well-documented, not treatable Age Yes Gender Yes Race Yes Prior stroke Yes

Well-documented and treatable Hypertension Cardiac TIA Diabetes Increased hematocrit

Less certain factors Lipids and cholesterol Smoking Obesity Oral contraceptive

Doubtful factors Alcohol Coffee Pulmonary Pulmonary Renal

Striking Definite Not evaluated Yes Definite Not evaluated Yes Definite Not evaluated Many times Definite Not evaluated

Dominant Strong correlation Foremost Powerful Yes Very important Frequent Yes Very important Definite Yes Definite Related Substantial

Yes No Yes No Not evaluated

Conflicting Conflicting Possible Not evaluated Men >65 Not definite Possible Weak

Not important Possible Not evaluated Significant Yes Possible Not evaluated

Not evaluated Not evaluated Not evaluated Not evaluated Not evaluated

NS NS NS NS

0.01 NS NS 0.05 NS

Not evaluated Not evaluated Not evaluated Not evaluated

Not evaluated Not evaluated Not evaluated Not evaluated Not evaluated

NS, Not significant. *After Dyken ML, Wokf PA, Barnett HJM, et al. Stroke 1984;15:1105-11.

More detailed analyses of the various subgroups within the overall populations before and after stroke might resolve these differences. Similar data have been reported from the North American Symptom- atic Carotid Endarterectomy Trials (NASCET), wherein risk factors were found to be very important in those patients assigned to the medical group, in contrast to the surgical group in which risk factors had no obvious effect on the subsequent risk of stroke. 1.

The value of the CHAT analysis in this report was to separate patients within the traditional groups of "asymptomatic" stenosis, TIA, and stroke into several subgroups on the basis of other details of their current status and history. This analysis has confirmed the value of such an approach, since important and statistically significant differences have again been shown between such subgroups, which may weigh heavily in the decision to advise surgery and in the prognosis for the patient. For example, those patients free of symptoms with no prior history of contralateral neurologic symptoms had a much lesser probability of late stroke (96% stroke-free at 5 years) than those who were operated on for a contralateral

markedly different. Similar differences have been shown in the past between amaurosis fugax and cortical TIA. 1~ The differing patterns of risk factor importance brought out by the current data reem- phasize these separate clinical entities, which have been grouped together in the past. In addition, however, the potential of many further important distinctions still remains within the CHAT classification, if one could analyze a much larger data base with longer follow-up data. The value of the arterial and target organ subclassifications within each clinical and historical group cannot be evaluated adequately until a vastly larger data base is available, which would probably require pooling of data from a number of imtitutiom. We hope reports such as this will stimulate the cooperative reviews required to provide a detailed prospective analysis of the value of all of the CHAT factors and combinations in both medically and surgically treated patients.

CONCLUSIONS CHAT analysis and risk factor grading of patients

after CEA has permitted evaluation of the importance of standard risk factors on the probability ofhte

lesion opposite a prior stroke with permanent resin-! stroke and of survival after such surgery. The ual deficit (80% stroke-free at 5 years). In additio~ ~!:3) ~ddence of late stroke in the entire patient group, the likelihood of survival of the two groups was '~*/~gardless of clinical presentation, was 3% at 5 years,

Volume 16 Number 4 October 1992 CHAT analy~ of risk factors after CEA 585

by life-table analysis. There were no significant differences in this outcome regardless of the clinical preoperative presentation.

Specific risk factors and the TOT score did not predict late stroke after CEA, although they were powerful predictors of early death in every clinical subgroup analyzed. Logistic analysis indicated that few risk factors were important predictors of late stroke or early death and that these predictors were different for each clinical presentation subgroup.

REFERENCES 1. Betnstein EF, Browse NL. The CHAT classification of stroke.

Ann Surg 1989;209:242-8. 2. Hye RJ, Dilley RB, Browse NL, Bemstein EF. Evaluation of

a new classification of cerebrovascular disease (CHAT). Am J Surg 1987;154:104-10.

3. Baker JD, Rutherford RB, Bernstein EF, et al. Suggested standards for reports dealing with cerebrovascular disease. J VAsc SURG 1988;8:721-9.

4. Wolf PA, Karmel WB, Verter J. Current status of risk factors for stroke. Neurol Clin 1983;1:317-43.

5. Dyken ML, Wolf PA, Barnett HJM, et al. Risk factors in stroke. Stroke 1984;15:1105-11.

6. Whisnant JP. Classification of cerebrovascular diseases, Ill: stroke 1990;21:637-76.

7. Whisnant JP, Wiebers DO. Clinical epidemiology of transient

cerebral attacks (TIA) in the anterior and posterior cerebral circulation. In: Sundt T, ed. Occlusive cerebrovascular disease: diagnosis and surgical management. Philadelphia: WB Saundets, 1987:60-5.

8. Sacco RL, Foulkes MA, Mohr JP, et al. Determinants of early recurrence ofcerebralinfarction: the Stroke Data Bank. Stroke 1989;20:983-9.

9. Thomas M, Otis SM, Rush M, ZyroffJ, Dilley RB, Bernstein EF. Recurrent carotid artery stenosis following endarterectomy, Ann Surg 1984;200:74-9.

10. Kaplan EC, Meier P. Non-parametric estimation from incom- plete observations. J Am Stat Assoc 1958;55:457-81.

11. Gehan EA. A generalized Wilcoxon test for comparing arbitrary singly-censored samples. Biometrika 1965;52:203- 18.

12. Dixon WJ, Brown MB, Engelman L, Jennrich ill. Stepwise logistic regression. In: BMDP statistical software manual, vol 2. Berkeley, California: University of California Press, 1990: chap LR.

13. Warlow CP. MRC European Carotid Surgery Trial: interim results for symptomatic patients with severe (70-99%) or with mild (0-29%) carotid stenosis. Lancet 1991;337:1235-43.

14. North American Symptomatic Carotid Endarterectomy Trial Collaborators. Beneficial effect of carotid endarterectomy in symptomatic patients with high grade stenosis. N Engl J Med 1991;325:445-53.

Submitted Feb. 13, 1992; accepted May 1, 1992.

DISCUSSION

Dr. Jerry Goldstone (San Francisco, Calif.). As you heard from Dr. Bernstein, this manuscript describes the first time that the CHAT system has been used to analyze some of the risk factors that have been implicated in predicting stroke and death and some of the other factors related to cerebrovascular disease. The results notwith- standing, that is, that they did not find many positive correlations, which in itself was surprising, I think the importance of this report is the CHAT system itself. Although this method was presented originally in 1987 and then subsequently in 1989 and although most of us have known it was there, we have not really paid much attention to it.

In fact, at the end of their 1989 paper the authors concluded by stating: "The usefulness of such a dassifica- tion must be proven by clinical studies whose aim is to determine whether the subclassifications will permit a more accurate assessment of the risk of stroke or recurrent stroke and the relative risk and benefits of various forms of intervention both medical and surgical." The authors have answered their own challenge.

One of the important results from today's data is that after these 700 carotid operations, there was only a 3% late stroke rate at the end of 5 years. This is a recurrent stroke rate of only 0.6% per year, and to me this seems to be one

of the most important risk factors that was identified: the mere presence of a significant carotid bifurcation lesion. In fact, perhaps the most important method of risk factor reduction is the properly performed removal of that lesion in carefully selected patients. I think the key here is the "carefully selected patient" component, and the CHAT system is designed to allow us to do just that; that is, to select properly those patients who are likely to benefit, both in the short and long terms.

This CHAT system is on first glance rather complicated, but it is really quite logical, and the more you look at it, the more logical it becomes. It has several advantages that Dr. Bemstein has not gone into today, but that are well-documented in the manuscript and in his 1989 manuscript.

Something that is curious to me is that even though the joint national vascular societies, following the recommen- dation of Dr. Baker's committee, have adopted this system as the reporting standard for cerebrovascular disease, I have yet to see a paper in print that uses the CHAT system, and I would like to ask Dr. Bernstein if perhaps he can comment on the reasons for this.

Second, how complicated is this system in real life? That is to say, you used a large computer and complex software. Most of us don't have large computers or a bevy


Journal of VASCULAR

SURGERY

of people dedicated to data acquisition. Is the system difficult to code? How long does it take, and do you think it's practical on, for example, a small desktop computer?

At the end of this manuscript, the authors state, "We hope reports such as this will stimulate the cooperative reviews required to provide a detailed prospective analysis of the value of all of the CHAT factors and combinations." Dr. Bernstein, what specific recommendations or proposals do you have for this society? Do you think the Western Vascular Society should develop a vascular registry such as other regional vascular societies have done and use the CHAT system to see if, as you suggested, a larger population of patients would prove what has been carefully suggested in this study?

Finally, I for one would like to strongly endorse the method. It is the first system that has hopes for sorting out what is important and what is not important in extracranial cerebral vascular disease. We still need to know these things. The North American Symptomatic Carotid End- arterectomy Trial (NASCET) results have been published and the Veterans Affairs asymptomatic trial results will soon be published, which both show a benefit for carotid surgery, but these groups are only showing a benefit at the moment for certain selective groups of patients, and we need to know exactly what we're dealing with because clearly, as Dr. Bernstein indicated, if the patients will only survive a year, there will be strong pressures brought on us to recommend not operating on them, particularly if they are free of symptoms.

I urge all of you to read this manuscript carefully when it is published and to consider using this system in your own research efforts whether they be in a large university or in a private practice.

Dr. Eugene F. Bernstein. First, I would like to thank Dr. Goldstone for his kind words. Second, I would like to comment on the question of how hard CHAT is to use. I think the reason other people have not used it is that it is daunting to look at all of the subgroups and even to look at the classification table. People say to themselves, "There's no reason I 'm going to do that. It's too much work for the return."

The current international classification for lung cancer, which is used all over the world, is by TNM subsets. This is as daunting as and is analogous to the CHAT system. It is now the international standard of communication among people who study lung cancer because it is important to know whether you are in stage 3A or 3B.

What we are proposing is not something that's too tough to do or without value. I think it is analogous to what the cancer surgeons do, and they have learned that they can decide intelligently about a patient's treatment and prognosis by knowing more about the patient.

How complicated is it to use our system? Our assistant did a lot of this classification after perhaps a few hours of initial training, with only a little supervision thereafter. I would say an average patient could be accurately classified in less than 5 minutes of reviewing the chart.

You do not need more than a Macintosh computer to look at the data, depending on what kind of analysis you want. I f you already have a registry, it is simple to add the CHAT subgrouping to your classification. Then the data could be run through a large computer to analyze more information.

What could the Western Vascular Society do? If three or four of the larger participating groups in our Society were to look at their data with this system in a way that would permit us to combine their data, so that instead of a few hundred patients we would be analyzing a few thousand patients, we would really know if there is something to be gained by this kind of classification system.

Dr. Allen Wail (San Francisco, Calif.). Dr. Bernstein, in a recent retrospective review of community hospital carotid reconstructions of slightly more than 200 patients, a dominant risk factor seemed to be the individual surgeons' technical competence. I wonder how you might include individual surgeon variability in a CHAT study?

Dr. Bernstein. Well, that's difficult. Frankly, it has not been addressed by the cancer study group. Many other issues are difficult, and certainly this kind of proposal is presented only to be tested and modified by things that one would learn after testing it. I f experience were to show that this factor or any other factor that's not listed in the initial list was important, the factor would be added, just as the international TNM classification for cancer is regularly changed.

Dr. D. Eugene Strandness (Seattle, Wash.). I had an opportunity to read this manuscript. I thought it was fascinating, and I agree with many of Dr. Goldstone's comments. There are a couple of weaknesses that I see in

this classification system, and I think they are fundamental and pertain more to the patient who has not undergone CEA.

The two boil down to the arterial and target organ pathologic conditions. If you look at the NASCET results, the striking thing about the medically treated group is that most of those patients did not have a stroke. Why didn't they have a stroke? The fact of the matter is we don't know exactly what's going on in the arterial wall in that particular subset of patients.

The same problem applies to the target organ. We don't really have good methods of looking at the target organ, and I 'm not quite sure, Dr. Bernstein, that looking at the target organ on the basis of this classification scheme would do much good.

The other aspect of this report that concerns me relates to the role of diabetes as a risk factor. Patients with type II diabeties are known to have twice the stroke risk of those without diabeties and 8 times the prevalence of carotid artery disease. Yet in the group of patients with symptoms it did not seem to appear as a prominent risk factor. Is this observation true or simply a matter of small numbers? I wish you would comment on this issue of the arterial and target organ pathologic conditions.

Dr. Bernstein. The analysis we were able to do on patients who had undergone surgery permitted us to have


arterial pathologic evaluations based on an angiogram in every case. Because of the limit of time I only presented to you a two-line summary of the data analysis for the arterial pathologic condition, which merely showed that the more arterial disease you have, the more likely you are to die early. But once the artery is operated on, you are not any more likely to have an ipsilateral stroke on that side.

Regarding the target organ pathologic condition, we did not have an adequate amount of data to provide a statistically valid analysis once we subdivided the data into the various subgroups.

We are just beginning to study a group of patients in our institution who have not undergone carotid surgery, but who were first seen by the vascular laboratory with various kinds of strokelike symptoms and complaints and who have had examinations by duplex scanning. We will try to come up with some information about patients who have not undergone surgery, as you have requested, because I think it is very important and it will involve neurologists in a trial of this system.

However, many of these patients who have not been operated on have never had a CT scan, MRI scan, or angiogram. So we really do not know as much about the arterial or brain pathologic condition as we would like to. On the other hand, I am sure that that study needs to be done.

Regarding your question about diabetes, in one particular subgroup diabetes was a potent factor. I do not know why it has not reappeared. Diabetes did not appear as an important factor for secondary stroke in the Stroke

Data Bank information either. It may be that there are not enough patients with a long enough follow-up period in both of these groups for it to appear.

Dr. J. Dennis Baker (Los Angeles, Calif.). The committee on Reporting Standards studied the CHAT system before incorporating it into our recommendations. Although it seems daunting, the system is not difficult once you become familiar with it. With a reasonable size series it is possible to manage both data storage and statistical analysis with a modem microcomputer.

Also, the committee on Reporting Standards consid- ered it important to include the T (target) component. We wanted a system that would be relevant to the needs, not only of vascular surgeons, but also of neurologists and neurosurgeons, for whom brain status is important.

Dr. Bernstein. I have no argument with your desire to have more precise risk factors. I f they were known they could be incorporated into CHAT. As a first effort, we selected the risk factor panel that had been adopted by our two society committees. The Joint Council committees for both peripheral arterial disease and cerebrovascular disease had described identical risk factor panels, and they seemed like a pretty good start.

Dr. William C. Krupski (Denver, Colo.). Dr. Bern- stein, what is the power of your study? Could it be that by making so many categories you diluted the numbers and that this represents a type 2 error?

Dr. Bernstein. Yes. My plea is exactly that. You've seen as much as we can learn from 700 patients with carotid disease, and if we're to learn more, we need 7000.

chat analysis of the influence of specific risk factors on late results after carotid endarterectomy

Documents