Age and Outcomes After Carotid Stenting and Endarterectomy
The Carotid Revascularization Endarterectomy Versus Stenting Trial
Background and Purpose—High stroke event rates among carotid artery stenting (CAS)-treated patients in the Carotid Revascularization Endarterectomy Versus Stenting Trial (CREST) lead-in registry generated an a priori hypothesis that age may modify the relative efficacy of CAS versus carotid endarterectomy (CEA). In the primary CREST report, we previously noted significant effect modification by age. Here we extend this investigation by examining the relative efficacy of the components of the primary end point, the treatment-specific impact of age, and contributors to the increasing risk in CAS-treated patients at older ages.
Methods—Among 2502 CREST patients with high-grade carotid stenosis, proportional hazards models were used to examine the impact of age on the CAS-to-CEA relative efficacy, and the impact of age on risk within CAS-treated and CEA-treated patients.
Results—Age acted as a treatment effect modifier for the primary end point (P interaction=0.02), with the efficacy of CAS and CEA approximately equal at age 70 years. For CAS, risk for the primary end point increased with age (P<0.0001) by 1.77-times (95% confidence interval, 1.38–2.28) per 10-year increment; however, there was no evidence of increased risk for CEA-treated patients (P=0.27). Stroke events were the primary contributor to the overall effect modification (P interaction=0.033), with equal risk at ≈64 years. The treatment-by-age interaction for CAS and CEA was not altered by symptomatic status (P=0.96) or by sex (P=0.45).
Conclusions—Outcomes after CAS versus CEA were related to patient age, attributable to increasing risk for stroke after CAS at older ages. Patient age should be an important consideration when choosing between the 2 procedures for treating carotid stenosis.
Patient age has been shown to influence the outcomes after carotid revascularization.1–6 The Carotid Revascularization Endarterectomy vs Stenting Trial (CREST) protocol was developed in 1997,7 when age and vascular anatomy8,9 were not yet recognized as predictors of complications of carotid artery stenting (CAS). To the contrary, it was postulated that CAS might be safer than carotid endarterectomy (CEA) in the elderly. However, during the conduct of the lead-in phase of the study, a high risk of stroke events was observed among the CAS-treated patients, and octogenarians were subsequently excluded from this portion of the trial (but were continued in the randomized phase to assess if equivalent risks were present for the CEA-treated patients).10 At this time (on the basis of lead-in data only and before unblinding of randomized data), the study investigators committed to the preplanned formal assessment of the impact of age on relative efficacy reported herein.
Materials and Methods
Study Participants and Measurement
CREST is a randomized clinical trial assessing the relative efficacy of CAS versus CEA. The study enrolled 1321 symptomatic patients and 1181 asymptomatic patients. End points were adjudicated by committees blinded to treatment assignment. Details of the study are provided elsewhere.11–13 The protocol was approved by the Institutional/Ethics Review Board of all participating sites. All patients gave written informed consent.
The focus of these analyses is to assess if the age of the patient influences the relative efficacy of CAS and CEA and, if so, what are the contributors of the effect modification. As such, the primary evaluation of efficacy was assessed on an intention-to-treat analysis using proportional hazards analysis to evaluate the potential of an age-by-treatment interaction after adjustment for symptomatic status and sex. The primary outcome of the trial was stroke, myocardial infarction (MI), or death during a periprocedural period (30 days after procedure for those receiving treatment within 30 days, or 36 days after randomization for those not receiving treatment within 30 days), or ipsilateral stroke over a follow-up period extending 4 years from randomization. Potential effect modification by age was analyzed assuming a linear effect of age (after confirming the linear assumption was reasonable). The stroke component was defined as any stroke during the periprocedural period and stroke ipsilateral to the study artery for the subsequent 4-year period. The MI component was defined as elevated enzymes plus either symptoms or electrocardiographic evidence of an event during the periprocedural period. Too few deaths occurred during the periprocedural period to permit meaningful analyses of this component on its own.
The modeling approach used the addition of interaction terms to proportional hazard models predicting the composite end point, the stroke end point, and the MI end point. A priori, P<0.10 was considered suggestive of effect modification. We also provide secondary analyses within specific age strata (younger than 65, 65–74, and 75 years or older) to provide the reader details, including: (1) the number of events in this broad age range; (2) an assessment of the linearity of the primary analysis of age-by-treatment interaction; (3) event rates within age strata for each treatment for comparisons to other studies; and (4) crude estimates of CAS-to-CEA relative efficacy for these age strata.
We also assessed if any potential age-by-treatment effect modification was consistent by symptomatic status or sex by adding higher-order interaction terms to the model. Proportional hazards models were fit separately for those CAS-treated and CEA-treated patients to describe the age-related changes in risk within each treatment contributing to their relative CAS-to-CEA efficacy differences.
Finally, to identify potential causes underlying the age–treatment interaction, we conducted a mediation analysis14 to identify if the increased risk at older ages for CAS-treated patients was attributable to an increased prevalence of risk factors (hypertension, diabetes, or dyslipidemia), differences in the characteristics of the lesion (lesion length, eccentric lesions, ulcerated lesion, or percent stenosis), or differences in the procedure (fluoroscopy time or total procedure time) by entering these factors into the model and observing the change in the estimated hazard ratio associated with age. Characteristics of the lesion were determined by the local clinic. Anatomic characteristics such as aortic arch anatomy, vessel tortuosity, and calcification known to be associated with age and CAS complications were not available for analysis. The standard error of the mediation was estimated using bootstrap techniques.
For both treatment groups in CREST, with increasing age participants were more likely to be female, white, and to have higher levels of systolic blood pressure and lower levels of diastolic blood pressure; however, they were less likely to have diabetes, dyslipidemia, or to be current smokers (Table 1). There were no significant differences between treatment groups for these factors in any age strata. Figure 1 shows the distribution of the ages for CAS and CEA.
Table 2 provides the observed number of MI, strokes, and primary end points within approximate tertiles of age strata for both the periprocedural period and for the 4-year outcome. Figure 2 provides the associated Kaplan-Meier estimates of the proportion of participants with a primary end point for each age–treatment strata, showing the similarity of time-to-event across age strata for CEA-treated patients, but the differences of time-to-event across age strata for CAS-treated patients. As previously reported for the primary end point at 4 years, there was evidence of a treatment-by-age interaction (P=0.02). The CAS-to-CEA risk increased with advancing age, from 0.60 (95% confidence interval [CI], 0.31–1.18) for patients younger than 65 years to approximate equal risk for those aged 65 to 74 years (hazard ratio, 1.08; 95% CI, 0.65–1.78), and to 1.63 (95% CI, 0.99–2.69) for those aged 75 years and older. This increasing risk was associated with increasing event rates in the CAS-treated patients (3.9% in the youngest age strata, 6.3% in the middle, and 12.7% in the oldest), whereas risk was relatively stable in the CEA-treated patients (respective rates: 6.1% youngest, 6.8% middle, and 7.4% oldest). This increasing risk was driven by the stroke end point, with a higher (P=0.033) CAS-to-CEA risk across age strata, with hazard ratios of 0.78 (95% CI, 0.37–1.62), 1.42 (95% CI, 0.78–2.60), and 2.15 (95% CI, 1.19–3.91). The increasing CAS-to-CEA risk at older ages is associated with increasing stroke event rates for those CAS-treated (3.7%, 5.1%, and 10.9%), but not in those CEA-treated (4.5%, 4.6%, and 4.9%). A similar pattern of effects (increasing CAS-to-CEA risk largely driven by increasing risk at older ages in the CAS-treated patients) was observed during the periprocedural period for both the composite and stroke end points; however, these trends failed to reach a level of statistical significance (P>0.1). Contralateral strokes occurring during the periprocedural period were a component of the composite outcome and the stroke outcome, but contralateral strokes after the periprocedural period are not part of these outcomes. For the “all stroke” end point (including contralateral strokes occurring after the periprocedural period), the treatment differences across the age spectrums are diluted (P=0.19) by the addition of the stroke events beyond the periprocedural period.
Although we urge caution in interpretation, Supplemental Table I (https://stroke.ahajournals.org) provides results similar to Table 2, stratified by symptomatic status. This Table requires stratification by both age and symptomatic status, and as such the small sample size in specific stratum could lead to misleading results. We have considered age-by-symptomatic status interactions and found none to be significant (P>0.1), and differences between symptomatic and asymptomatic patients in the relationships of risk with age could have easily occurred by chance alone. However, these data are provided for comparisons with the results of other studies that do not include both asymptomatic and symptomatic patients.
The primary analysis for this report is shown in Figure 3. Figure 3A shows the CREST primary end point as a continuous function of age (identical Figure to that shown in primary study results article,11 all other Figures and analyses are novel to this article). The risk of the 2 procedures is approximately equal at age 70 years, with CAS showing superiority in younger patients, and there is an increasing benefit for CEA in older patients. The stroke component of the composite end point as a function of age is shown in Figure 3B. The steeper slope in this Figure implies a larger magnitude of effect modification by age on the occurrence of stroke (P=0.033) than for occurrence of the primary end point. We note that unlike the composite outcome in which CAS-to-CEA risk approaches a significant advantage for CAS at younger ages, the wider CI the stroke end point implies the upper limit of the 95% CI bounds remains >1.0; however, the a priori focus of this article was on the trend of risk with differences in age (rather than differences at any specific age). The point of equal risk for CAS and CEA is at age 64 years, 6 years younger than for the primary end point. The wider 95% CI bounds imply greater uncertainty for the stroke outcome compared to the primary outcome. There was no evidence (P=0.35) of effect modification by the MI component of the primary end point (Figure 3C).
For those treated with CAS, there was a 1.77-times increase in risk of primary end point event (P<0.0001; 95% CI, 1.38–2.28) and a 1.76-times increase (95% CI, 1.35–2.31) for stroke events with each 10-year difference in age. For those treated with CEA, there was no evidence of a difference in risk across the age spectrum for either the primary end point (hazard ratio, 1.16; 95% CI, 0.89–1.50; P=0.27) or for stroke events (hazard ratio, 1.12; 95% CI, 0.82–1.54; P=0.47). Introduction of higher-order interaction terms did not suggest that the age modification of treatment effect was influenced by either symptomatic status (P=0.96) or by sex (P=0.45). The sensitivity analysis using the alternative definition of MI, including 20 biomarker-only MI, showed a nonsignificant effect modification of age (P=0.75).
Mediation analysis was performed to assess factors potentially contributing to the age-related risk differences in the CAS treatment group (Table 3), but it was not performed for those randomized to CEA because of the lack of evidence for age-related changes for those randomized to CEA. There was no evidence that the effect of age in the CAS group was mediated by differences in the prevalence of hypertension, diabetes, or dyslipidemia, or by differences in observed lesion characteristics or procedure duration (P>0.05). Although total fluoroscopy time was identified as a potential mediator (P=0.046), its effect was modest, only reducing the age hazard ratio from 1.68 to 1.62 for a 10-year difference in age.
The current analysis indicates that the age-related differential efficacy observed in CREST11 is primarily attributable to the stroke component of the primary end point. In turn, the impact of the stroke component is largely driven by an increasing risk of stroke with increasing age among CAS-treated patients, but little change in the increasing risk of stroke with increasing age among CEA-treated patients.
The point of similarity for the risk of stroke for CAS and CEA is at 64 years, compared to ≈70 years for the risk of the primary end point. The occurrence of MI after either procedure did not differ with age, suggesting CAS results in fewer MI across the entire age spectrum. However, because there were fewer MI events (N=42) than stroke events (N=122), there was lesser power to detect effect modification for MI than stroke. There was no evidence that the age-by-treatment relationships differed by symptomatic status or sex.
Our observation of an age effect modification, originally reported in the primary results article,11 was subsequently confirmed by the meta-analysis of the Stent-Protected Angioplasty vs Carotid Endarterectomy (SPACE) trial, the Endarterectomy Versus Angioplasty in Patients with Symptomatic Severe Carotid Stenosis (EVA-3S) trial, and the International Carotid Stenting Study (ICSS).15 For patients 70 years and older, the risk of events in CAS-treated patients was approximately twice that for CEA-treated patients (hazard ratio, 2.04; 95% CI, 1.48–2.82). This differential age effect in the meta-analysis was also driven by stroke because MI was not a component of the primary end point for this meta-analysis. This meta-analysis showed no differences in risk for patients younger than age 70 (hazard ratio, 1.11; 95% CI, 0.73–1.71), a finding not supported by our analysis in which younger CAS-treated patients were shown to be at lower risk. Separately, the SPACE investigators reported a risk of 0.54-times (95% CI, 0.28–1.03) less for CAS relative to CEA for those younger than aged 68, and a risk of 1.80-times (95% CI, 0.96–3.40) greater for CAS relative to CEA in those aged 68 and older.16 This age effect is also consistent with reports from the lead-in series of CREST, the Carotid Acculink/Accunet Post-Approval Trial to Uncover Unanticipated or Rare Events (CAPTURE) registry, and in ICSS.10,17,18 All of these trials used eligibility criteria similar to CREST that did not incorporate anatomic exclusion criteria for CAS now thought to be important in elderly patients.9 None of the analyses included a detailed examination of the separate effect modification by the stroke and MI components of the outcomes.
When CREST was designed, we anticipated that the less invasive CAS would be superior in older age groups compared to the more invasive CEA. Accordingly, the superior performance of CEA in older individuals and the superior performance of CAS in younger individuals were unexpected. This position was challenged by the observation of high risk among the CAS-treated patients in the CREST lead-in registry10 (the observation leading to this preplanned analysis), a finding confirmed in the analyses reported herein. Observational studies completed before CREST documented age as a predictor of stroke risk8,19,20 potentially related to the marked increase in aortic arch and carotid artery tortuosity and calcification in the elderly.9 In CREST, characteristics of the carotid lesion such as the degree of stenosis, lesion length, eccentricity, and ulceration had no detectable effect on risk of CAS at older ages. However, the degrees of arterial tortuosity or lesion calcification were not available in the data set and could contribute to the increased CAS event rates in the elderly.9,21–24 We hypothesize that the risk of embolization during CAS is increased during navigation of tortuous extracranial arteries, particularly in patients with heavily calcified vessels and “extended” type II and III aortic arches. Consistent with this hypothesis, we observed that the elderly required longer fluoroscopy time for CAS. Adjustment for this factor partially mediated the magnitude of the increased risk at older ages. Of note, the higher event rates in the elderly were not associated with increase in cardiovascular risk factors, consistent with previous reports.25
The interaction between patient selection, operator experience, and technology may be relevant to the age interaction in this analysis. Recent reports of CAS using updated patient selection criteria suggest that the age differential for CAS may be absent or blunted.26,27 These studies of CAS, also using new proximal protection devices designed to be less affected by arterial tortuosity, were notable for low event rates in the elderly.28,29 Further studies are required to confirm these findings.
Strengths of the CREST analysis include a large cohort of patients with a broad age distribution, inclusion of asymptomatic patients, and age results consistent with results from the CREST credentialing study and subsequent randomized trials. Limitations include smaller numbers of events than anticipated (because of better than expected safety for both CAS and CEA) and smaller proportions of patients at the tails of the age distribution, 161 (6.4%) aged 55 years and younger and 240 (9.6%) aged 80 years and older (Figure 1). Nonetheless, the finding that the interaction test was significant provides prima facie confirmation that there are sufficient numbers of individuals in the tails of the age distribution to describe the effect of age.
This prespecified analysis of the CREST trial demonstrates that the differential efficacy of CAS compared to CEA across the age spectrum is primarily attributable to stroke events. The pattern of lower relative risk in the CAS group at younger ages and higher relative risk at older ages is driven by increased risk for stroke at older ages for CAS. For CEA, the risk for stroke is relatively constant across the entire age spectrum. We conclude that patient age should be an important factor in selecting the treatment option for carotid stenosis. The anatomic factors that may contribute to these observations require further study.
Sources of Funding
This work was supported by the National Institute of Neurological Disorders and Stroke (NINDS) and the NIH (R01 NS 038384), and by supplemental funding from Abbott Vascular Solutions (formerly Guidant), including donations of Accunet and Acculink systems, equivalent to approximately 15% of the total study cost, to CREST centers in Canada and to CREST centers in the United States that were at Veterans Affairs sites.
G.H. is a Consultant/Advisory Board for Bayer Healthcare, member of the ARRIVE Executive Committee, received research support from Amgen and Bayer Healthcare, and is a consultant to Abbott Vascular for preparation of FDA materials. G.S.R. received royalties from Abbott Vascular and Cook. D.J.C. received research support from Boston Scientific, Abbott Vascular, Edwards Lifesciences, MedRad, Merck/Schering-Plough, Medtronic, and Eli Lilly-Daiichi Sankyo, is a consultant to Schering-Plough, Eli Lilly, Medtronic, and Cordis, and is a member of the Speakers' Bureau for Eli Lilly and The Medicines Company. H.D.A. is a member of the Speakers' Bureau/Advisory Board for Medtronic.
- Received May 9, 2011.
- Accepted August 4, 2011.
- © 2011 American Heart Association, Inc.
- McCrory DC,
- Goldstein LB,
- Samsa GP,
- Oddone EZ,
- Landsman PB,
- Moore WS,
- et al
- Rothwell PM,
- Slattery J,
- Warlow CP
- Eckstein HH,
- Ringleb P,
- Allenberg JR,
- Berger J,
- Fraedrich G,
- Hacke W,
- et al
- Mathur A,
- Roubin GS,
- Iyer SS,
- Piamsonboon C,
- Liu MW,
- Gomez CR,
- et al
- Zahn R,
- Ischinger T,
- Hochadel M,
- Zeymer U,
- Schmalz W,
- Treese N,
- et al
- White CJ
- Chaturvedi S,
- Matsumura JS,
- Gray W,
- Xu C,
- Verta P
- Chiam PT,
- Roubin GS,
- Panagopoulos G,
- Iyer SS,
- Green RM,
- Brennan C,
- et al