Implementation and Efficacy of Selective Sonographic Screening for Carotid Disease before Cardiac Surgery

Article Outline

• Abstract
• Introduction
• Materials and Methods
• Results
  • Routine Sonographic Screening (Phase I)
  • Selective Sonographic Screening (Phase II)
  • Recommendation vs. Clinical Decision for Preoperative Sonography
  • Estimated vs. Detected Frequency of Carotid Disease
• Discussion
  • Study Limitations
• References
• Copyright

Background

Preoperative carotid sonography with consecutive preventive strategies might reduce stroke risk during cardiac surgery. Since routine sonography in all patients may be unfeasible, an approach to examine preselected patients was investigated.

Methods

A prognostic model predicting carotid disease was developed using the clinical data of 1,768 routinely examined patients. It recommended 1,018 of 4,814 patients of a following collective for selective sonography. Patients recommended for preoperative sonography were compared to those selected in clinical practice.

Results

Besides the evaluated predictor variables, a history of syncope/cardiogenic shock and of pulmonary disease was associated with patient selection for sonography in clinical practice, even though both variables were not associated with severe carotid disease. In patients who underwent sonography, although this was not recommended by the prognostic model, severe carotid disease was estimated lower than what was actually detected, suggesting a change in relative relevance of predicting variables along with the change in frequencies of patients' cardiovascular characteristics.

Conclusion

Prognostic models for selective screening before cardiac surgery may require reevaluation over time, especially when baseline characteristics used for prediction have changed. Criteria used in clinical practice to select patients for screening may differ from those recommended by investigational studies.

Back to Article Outline

Introduction 

Stroke remains an often catastrophic complication of cardiac surgery with cardiopulmonary bypass (CPB), with an incidence of about 2% after coronary artery bypass grafting (CABG) or valve surgery and 7% after combined CABG and valve procedures.¹, ², ³, ⁴ Carotid artery disease is a major risk factor for perioperative stroke leading to a fourfold risk increase at CABG.² Preoperative sonographic screening for severe carotid artery disease and consecutive preventive strategies like prior carotid endarterectomy or increase of intraoperative blood pressure might reduce this risk. However, considering examination time and capacities, nonselective sonography including all patients undergoing cardiac surgery may not be feasible. Examination of preselected patients using the association between carotid artery disease and disease predictors might be an alternative.⁵, ⁶, ⁷, ⁸, ⁹

We investigated the clinical implementation and efficacy of selective sonographic screening for carotid artery disease in consecutive patients who underwent cardiac surgery with CPB. We compared criteria used for patient selection for preoperative sonography in actual clinical practice with those recommended for selection as evaluated from a preceding series of patients with routine preoperative sonography.

Back to Article Outline

Materials and Methods 

From 1996 to 1998, when routine carotid screening was performed preoperatively (phase I of our study), 2,111 consecutive patients underwent CABG and/or valve surgery with CPB. During this period, 343 patients were not examined because of emergency cardiac surgery, unavailability of the sonographic device, or organizational deficiencies. From 1999 to 2005 (phase II), 1,006 of 4,814 consecutively operated patients were assigned to preoperative carotid sonography (Fig. 1).

• 
  • View Large Image
  • Download to PowerPoint

• Fig. 1 

  Patient selection. CABG, coronary artery bypass grafting; CPB, cardiopulmonary bypass.

Carotid sonography was performed by experienced examiners using Doppler (Multidop X4; Compumedics DWL, Singen, Germany) and color duplex (SSH 380; Toshiba, Tokyo, Japan) devices. Degree of stenosis was quantified according to validated criteria and expressed as local diameter reduction in steps of 10% in accordance with the method used in the European Carotid Surgery Trial.¹⁰, ¹¹, ¹², ¹³ For the present investigation, carotid disease was categorized as moderate or less (≤70%) or severe (80-100%). Preoperative diagnostics and data assessment did not change between 1996 and 2005.

Statistical analyses were performed with the SPSS 11.0 software package (SPSS, Inc., Chicago, IL). Baseline characteristics of patients were compared by chi-squared test and Mann-Whitney U-test. Significance was set at p < 0.05. A prognostic model for severe carotid disease was developed by multivariable logistic regression analysis with forward stepwise inclusion (p_inclusion < 0.05 and p_exclusion > 0.1) of baseline characteristics as listed in Table I. The cut-off point of the estimated risk score from the regression model to assume severe carotid disease was established by constructing a receiver operating characteristic (ROC) curve. The estimated prevalence of severe carotid disease in patient cohorts was calculated by adding up individual patient probabilities for severe carotid disease as derived from the regression model. Predicted and observed numbers of carotid disease were compared by chi-squared test for goodness of fit. Multivariable logistic regression analysis was also used to evaluate variables associated with actual clinical selection for preoperative sonography.

Table I. Patients' characteristics

	Routine screening (phase I)	Selective screening (phase II)
	Sonography (n = 1,768)	All (n = 4,814)	Sonography (n = 1,006)	No sonography (n = 3,808)	p
Age (years)	63.5 ± 9.6	65.5 ± 9.4	66.8 ± 9.0	65.1 ± 9.4	<0.001
≥65 years	878 (50%)	2,833 (59%)	637 (63%)	2,196 (58%)	<0.001
Female sex	463 (26%)	1,242 (26%)	254 (25%)	988 (26%)	0.750
Risk factors
Arterial hypertension	1,032 (58%)	4,172 (87%)	891 (89%)	3,281 (86%)	<0.001
Diabetes	425 (24%)	1,647 (34%)	372 (37%)	1,275 (34%)	<0.001
Hyperlipidemia	983 (56%)	3,594 (75%)	783 (78%)	2,811 (74%)	<0.001
Body mass index (kg/m2)	27.0 ± 3.1	28.0 ± 3.9	27.7 ± 3.9	28.1 ± 3.9	<0.001
Adiposity (BMI ≥30) kg/m2)	258 (15%)	1,377 (29%)	265 (26%)	1,112 (29%)	<0.001
History/concomitant diseases
Myocardial infarction	930 (53%)	2,195 (46%)	464 (46%)	1,731 (46%)	<0.001
Syncope/cardiogenic shock	433 (25%)	978 (20%)	278 (28%)	700 (18%)	<0.001
Heart surgery	58 (3%)	162 (3%)	49 (5%)	113 (3%)	0.866
Stroke/TIA	124 (7%)	298 (6%)	195 (19%)	103 (3%)	0.227
Peripheral arterial disease	363 (21%)	685 (14%)	298 (30%)	387 (10%)	<0.001
Pulmonary disease	293 (17%)	440 (9%)	126 (13%)	314 (8%)	<0.001
Renal disease	297 (17%)	692 (14%)	191 (19%)	501 (13%)	<0.001
Neck bruit	195 (11%)		250 (25%)
Cardiac findings
Atrial fibrillation	132 (8%)	431 (9%)	97 (10%)	334 (9%)	0.063
Other arrhythmias	101 (6%)	221 (5%)	48 (5%)	173 (5%)	0.071
NYHA classification					<0.001
Class I/II	135 (8%)	527 (11%)	100 (10%)	427 (11%)
Class III	1,394 (79%)	3,794 (79%)	812 (81%)	2,982 (78%)
Class IV	239 (14%)	493 (10%)	94 (9%)	399 (11%)
Coronary artery disease	1,504 (85%)	4,185 (87%)	878 (87%)	3,307 (87%)	0.055
Left main stem stenosis	172 (10%)	816 (17%)	178 (18%)	638 (17%)	<0.001
>1 diseased coronary artery	1,416 (80%)	2,993 (62%)	638 (63%)	2,355 (62%)	<0.001
Valve disease	438 (25%)	1,282 (27%)	304 (30%)	978 (26%)	0.137
Urgent (vs. elective) surgery	609 (34%)	3,077 (64%)	626 (62%)	2,451 (64%)	<0.001

Variables are presented as numbers (%) or mean (± standard deviation). p values presented for univariate comparisons between phase I cohort and whole phase II cohort.

NYHA, New York Heart Association.

Back to Article Outline

Results 

Routine Sonographic Screening (Phase I) 

During phase I, 1,768 patients underwent duplex sonography before cardiac surgery. Their baseline characteristics are presented in Table I. Prevalence of severe carotid disease was 3.7% (n = 65). Stepwise logistic multivariable regression with carotid disease as the dependent variable and the variables from Table I as potential covariables selected neck bruit, peripheral arterial disease, and history of a cerebrovascular event as independent predictors (Table II). A cut-off point for the linear score calculated from these covariables from the logistic regression model was derived from the ROC curve (area under the curve 0.874, 95% confidence interval [CI] 0.821-0.928). Selecting the point with the highest sum of sensitivity and specificity yielded a sensitivity of 0.88, specificity of 0.71, positive predictive value of 0.10, and negative predictive value of 0.99. Due to the simple binary nature of the covariables, the estimated score in an individual patient exceeded the selected cut-off point for the assumption of severe carotid disease if the patient presented at least one of the independent predictors neck bruit, peripheral arterial disease, and history of stroke/transient ischemic attack (TIA).

Table II. Predictors for severe carotid disease in 1,768 phase I patients with routine carotid sonography before cardiac surgery: stepwise multivariable logistic regression including all variables as listed in Table I as potential covariables

Selective Sonographic Screening (Phase II) 

Baseline characteristics of the 4,814 patients admitted during phase II and univariate comparison of characteristics between phase II and phase I patients are presented in Table I. To analyze differences of baseline characteristics in more detail, the phase II cohort was subdivided according to the year of surgery (1999-2005). Variables differing significantly between the phase I and phase II cohorts (Table I) showed increasing differences between phase II subcohorts over time from 1999 to 2005 (chi-squared test for trend).

Data on the 1,006 (21%) patients who actually underwent preoperative carotid sonography during phase II and on those who did not are also presented in Table I. Severe carotid disease was detected in 144 of the examined patients (14.3%). Criteria associated with the actual decision for sonography during phase II were evaluated by stepwise logistic multivariable regression and are presented in Table III. They include all three criteria used in the screening rule supplemented by a history of syncope/cardiogenic shock and pulmonary disease.

Table III. Selection criteria for carotid sonography before cardiac surgery in 4,814 consecutive phase II patients: stepwise multivariable logistic regression including all variables as listed in Table I

Recommendation vs. Clinical Decision for Preoperative Sonography 

Of the 4,814 patients during phase II, 546 (11.3%) underwent preoperative duplex examination and 3,336 (69.3%) were not examined both as suggested by the regression model evaluated in phase I and as recommended during phase II (Table IV). However, 460 (9.6%) patients underwent preoperative sonography, although this was not recommended according to the regression model (i.e., these patients had no neck bruit, no history of stroke/TIA, and no peripheral arterial disease). On the other hand, 472 (9.8%) patients recommended for preoperative sonography according to the regression model (i.e., with neck bruit, history of stroke/TIA, peripheral arterial disease, or any combination of these variables) did not undergo this examination before cardiac surgery (Table IV).

Table IV. Recommendation vs. clinical decision for selective preoperative sonographic screening for severe carotid disease in 4,814 consecutively operated phase II patients

In the 3,796 phase II patients not recommended for preoperative carotid sonography (Table IV), selection of 460 patients for sonography was associated with a history of syncope/cardiogenic shock (adjusted odds ratio [OR] = 1.59, 95% CI 1.27-1.99, p < 0.001) and pulmonary disease (adjusted OR = 1.59, 95% CI 1.17-2.17, p = 0.004) at stepwise logistic multivariable regression including all variables from Table I. However, pulmonary disease was not and syncope/cardiogenic shock was even inversely associated with the presence of severe carotid disease (adjusted OR = 0.14, 95% CI 0.03-0.60, p = 0.009). Only adiposity (body mass index [BMI] ≥30) was associated with severe carotid disease in this subcohort (adjusted OR = 2.64, 95% CI 1.32-5.25, p = 0.006).

Of those 1,018 patients recommended for sonography by the predictive model, 472 did not undergo carotid ultrasound based on clinical decision because of valve (adjusted OR = 1.76, 95% CI 1.13-2.75, p = 0.013) and urgent surgery (adjusted OR = 1.52, 95% CI 1.10-2.11, p = 0.011) or the absence of a history of stroke/TIA (adjusted OR = 4.0, 95% CI 2.89-5.52, p < 0.001) and of syncope/cardiogenic shock (adjusted OR = 1.87, 95% CI 1.30-2.70, p = 0.001).

Estimated vs. Detected Frequency of Carotid Disease 

Severe carotid disease was estimated in 113.6 of 4,814 (2.4%) phase II patients (cumulative individual risks): 93.4 of 1,018 (9.2%) recommended and 20.3 of 3,796 (0.05%) not recommended for sonography. Estimated and detected frequencies of patients with severe carotid disease could be compared in the examined subcohorts only. In 546 phase II patients who underwent recommended preoperative sonography, severe carotid disease was estimated in 73.3 (13.4%) but detected in 107 (19.6%) (p < 0.001). In 460 phase II patients who underwent preoperative sonography although not recommended, severe carotid disease was estimated in 2.5 (0.5%) but detected in 37 (8.0%) (p < 0.001).

Back to Article Outline

Discussion 

In accordance with former investigations, we evaluated neck bruit, peripheral arterial disease, and history of a cerebrovascular event as the strongest predictors for the presence of severe carotid disease in our routinely examined patients scheduled for cardiac surgery with CPB.⁵, ⁷, ⁹

However, when carotid sonography was switched from preoperative routine to a selective procedure, criteria associated with the clinical decision for examination and strength of these associations differed notably from those suggested from the routinely examined series and the literature. Excepting neck auscultation, which in this context may rather be considered as prescreening than as a clinical covariable associated with carotid disease, history of a cerebrovascular event was the strongest criterion for patient selection. Predominance of a “cerebral dysfunction” for the clinical decision to screen a patient is corroborated by the association found between history of syncope/cardiogenic shock, i.e., transient loss of consciousness, and screening. Although an association of syncope or cardiogenic shock with carotid disease has not been described—neither in patients scheduled for cardiac surgery (see “Results”) nor in general¹⁴—its presence in medical history was a significant criterion for selection, as was its absence for nonselection.

Association of patient selection for carotid screening with comorbidity of peripheral arterial disease was lower than that with history of a cerebrovascular event, although peripheral arterial disease has been described as the strongest predictor for severe carotid disease in the majority of studies, including ours, and may serve as the most valuable marker that atherosclerosis has exceeded the coronary perfusion territory.⁷, ⁹, ¹⁵ The latter may also explain that in most studies, including ours, risk factors for atherosclerosis like advanced age, arterial hypertension, diabetes, hyperlipidemia, smoking, and adiposity did not reach prognostic independence for carotid disease in the presence of substantial rates of peripheral arterial disease, i.e., a marker already indicating presence of a generalized atherosclerotic process.⁷, ⁸, ⁹

Given the large sensitivity and specificity of the screening rule derived in phase I, the relatively low predicted prevalence of severe carotid stenosis in patients recommended for selective screening during phase II (9.2%) might be surprising. However, this reflects the low positive predictive value of the screening procedure in this situation (0.10), which is due to the imbalance of patients with and without severe carotid disease.

Because we implemented selective screening in our clinical routine, we do not know the true prevalence of carotid disease in the whole cohort of 4,814 patients who underwent cardiac surgery during selective screening. As a consequence, we cannot directly compare the efficacy of selection criteria from logistic regression with the actual clinical decision. With respect to the association between carotid disease and predictor variables found in our routinely screened series, we estimated the prevalence in the total phase II study population at 2.4%. However, according to the 144 patients with severe carotid disease already detected in 1,006 examined patients and complying with 3.0% of the whole phase II population (yielding p = 0.07 in a goodness-of-fit test), true prevalence has to be assumed to be higher than estimated. This might be explained by an overfitting in the regression model derived from phase I. Due to this, the number of observed severe carotid disease cases should be lower than expected in patients with recommendation for screening but higher in those without recommendation. As patients without recommendation dominate within the whole cohort, the observed number with severe carotid disease would be higher than the expected number, which is what we found. Selection bias might be another explanation. It seems plausible to suspect that physicians used information beyond the variables of the (necessarily restricted) regression model and that these variables also might have been beyond the recorded ones in the study. Therefore, the decision for screening without recommendation could be a bias toward a higher frequency of severe carotid disease.

Both these reasons, however, would have led to a constant deviation of observed and expected cases during the phase II study course because the regression model is constant over time and the differences between recommendation and actual screening decision were approximately the same from 1999 to 2005. In contrast, however, the deviation between the expected and observed rates of severe carotid disease in patients who underwent preoperative sonographic examination despite missing recommendation increased from <1% to 15% (p < 0.001 for trend), paralleling the continuously increasing discrepancies of patients' cardiovascular characteristics over this time. This supports the hypothesis that, apart from the change in the prevalence of baseline characteristics (i.e., potential predictors for severe carotid disease), indeed their association with carotid disease had also changed over time. However, such a change is not reflected in the binary covariables of the applied logistic model and may limit its application.

Study Limitations 

Patients with neck bruit were invariably referred to carotid duplex during selective screening. However, as neck bruit was not documented systematically in the clinical case record form but only at carotid sonography, its absence in patients not referred to carotid screening may be assumed but may not be equivalent to 0% (and therefore was not presented in Table I. This might result in an underestimation of the true prevalence of severe carotid disease in the 4,814-patient cohort. However, this may not influence the suspected change of the impact of predictor variables along with the change of baseline characteristics because this was evaluated in patients who underwent sonography including standardized documentation of neck auscultation. Smoking was not assessed systematically in our patients. However, given that pulmonary disease means chronic obstructive pulmonary disease (COPD) in the vast majority of patients in industrialized countries and COPD in these countries most commonly is caused by smoking,¹⁶ pulmonary disease as assessed in our case record form may be highly related to (severe) smoking. This also may have predisposed its significance as a selection criterion for carotid screening in the absence of neck bruit, peripheral arterial disease, and history of stroke/TIA.

Prognostic screening models evaluated for detection of severe carotid disease with high sensitivity and specificity may require reevaluation over time, especially when the cardiovascular characteristics used for prediction have changed. Criteria used in clinical practice to select patients for screening before cardiac surgery may differ from those recommended from investigational studies.

Back to Article Outline

References 

1. Almassi GH, Sommers T, Moritz TE, et al. Stroke in cardiac surgical patients: determinants and outcome. Ann Thorac Surg. 1999;68:391–397
   • View In Article
   • MEDLINE
   • CrossRef
2. Naylor AR, Mehta Z, Rothwell PM, et al. Carotid artery disease and stroke during coronary artery bypass: a critical review of the literature. Eur J Vasc Endovasc Surg. 2002;23:283–294
   • View In Article
   • Abstract
   • Full-Text PDF (170 KB)
   • CrossRef
3. Salazar JD, Wityk RJ, Grega MA, et al. Stroke after cardiac surgery: short- and long-term outcomes. Ann Thorac Surg. 2001;72:1195–1201
   • View In Article
   • MEDLINE
   • CrossRef
4. Taylor NE, O'Brien S, Edwards FH, et al. Relationship between race and mortality and morbidity after valve replacement surgery. Circulation. 2005;111:1305–1312
   • View In Article
   • CrossRef
5. Aboyans V, Lacroix P, Guilloux J, et al. A predictive model for screening cerebrovascular disease in patient undergoing coronary artery bypass grafting. Interact. Cardiovasc Thorac Surg. 2005;4:90–95
   • View In Article
   • CrossRef
6. Ascher E, Hingorani A, Yorkovich W, et al. Routine preoperative carotid duplex scanning in patients undergoing open heart surgery: is it worthwhile?. Ann Vasc Surg. 2001;15:669–678
   • View In Article
   • Abstract
   • Full-Text PDF (817 KB)
   • CrossRef
7. Durand DJ, Perler BA, Roseborough GS, et al. Mandatory versus selective preoperative carotid screening: a retrospective analysis. Ann Thorac Surg. 2004;78:159–166
   • View In Article
   • CrossRef
8. Fukuda I, Gomi S, Watanabe K, et al. Carotid and aortic screening for coronary artery bypass grafting. Ann Thorac Surg. 2000;70:2034–2039
   • View In Article
   • MEDLINE
   • CrossRef
9. Salasidis GC, Latter DA, Steinmetz OK, et al. Carotid artery duplex scanning in preoperative assessment for coronary artery revascularization: the association between peripheral vascular disease, carotid artery stenosis, and stroke. J Vasc Surg. 1995;21:154–160
   • View In Article
   • Abstract
   • Full Text
   • Full-Text PDF (789 KB)
   • CrossRef
10. European Carotid Surgery Trialists' Collaborative Group. MRC European Carotid Surgery Trial. Interim results for symptomatic patients with severe (70-99%) or with mild (0-29%) carotid stenosis. Lancet. 1991;337:1235–1243
    • View In Article
    • Abstract
    • CrossRef
11. de Bray JM, Glatt B. Quantification of atheromatous stenosis in the extracranial internal carotid artery. Cerebrovasc Dis. 1995;5:414–426
    • View In Article
    • CrossRef
12. Goertler M, Widder B, Schuetz U. Quantifying medium- and high-grade carotid artery stenosis by ultrasound. J Echogr Med Ultrason. 1996;17:235–239
    • View In Article
13. Gortler M, Niethammer R, Widder B. Differentiating subtotal carotid artery stenoses from occlusions by colour-coded duplex sonography. J Neurol. 1994;241:301–305
    • View In Article
    • MEDLINE
    • CrossRef
14. Linzer M, Yang EH, Estes NA, et al. Diagnosing syncope. Part 1: Value of history, physical examination, and electrocardiography. Clinical Efficacy Assessment Project of the American College of Physicians. Ann Intern Med. 1997;126:989–996
    • View In Article
15. Chen WH, Ho DS, Ho SL, et al. Prevalence of extracranial carotid and vertebral artery disease in Chinese patients with coronary artery disease. Stroke. 1998;29:631–634
    • View In Article
    • MEDLINE
    • CrossRef
16. Lopez AD, Shibuya K, Rao C, et al. Chronic obstructive pulmonary disease: current burden and future projections. Eur Respir J. 2006;27:397–412
    • View In Article
    • MEDLINE