Copeptin for the Prediction of Recurrent Cerebrovascular Events After Transient Ischemic Attack
Results From the CoRisk Study
Background and Purpose—Copeptin has been associated with recurrent cerebrovascular events after transient ischemic attack (TIA). In an independent cohort, we evaluated copeptin for the prediction of recurrent cerebrovascular events within 3 months after TIA and assessed the incremental value of copeptin compared with the ABCD2 (age, blood, clinical features of TIA, duration of symptoms, presence of diabetes mellitus) and ABCD3-I (ABCD2, dual TIA [the presence of ≥2 TIA symptoms within 7 days], imaging [the presence of abnormal findings on neuroimaging]) scores.
Methods—This prospective, multicenter cohort study was conducted at 3 tertiary Stroke Centers in Switzerland and Germany.
Results—From March 2009 through April 2011, we included 302 patients with TIA admitted within 24 hours from symptom onset. Of 28 patients with a recurrent cerebrovascular event within 3 months (stroke or TIA), 11 patients had a stroke. Although the association of copeptin with recurrent cerebrovascular events was not significant, the association with stroke alone as end point was significant. After adjusting for the ABCD2 score, a 10-fold increase in copeptin levels was associated with an odds ratio for stroke of 3.39 (95% confidence interval, 1.28–8.96; P=0.01). After addition of copeptin to the ABCD2 score, the area under the curve of the ABCD2 score improved from 0.60 (95% confidence interval, 0.46–0.74) to 0.74 (95% confidence interval, 0.60–0.88, P=0.02). In patients with MRI (n=223), the area under the curve of the ABCD3-I score increased in similar magnitude, although not significantly. Based on copeptin, 31.2% of patients were correctly reclassified across the risk categories of the ABCD2 score (net reclassification improvement; P=0.17).
Conclusions—Copeptin improved the prognostic value of the ABCD2 score for the prediction of stroke but not TIA, and it may help clinicians in refining risk stratification for patients with TIA.
The risk of stroke is increased after a transient ischemic attack (TIA), but the ability to discriminate patients at high versus low stroke risk is still suboptimal. The ABCD2 score, with higher scores indicating greater stroke risk, aims at helping primary and emergency care physicians in the triage of patients with TIA.1 Although all patients with TIA need an immediate work-up, North-American guidelines recommend that patients with an ABCD2 score ≥3 should be hospitalized given their high risk of early stroke.2 Yet, a recent meta-analysis found that the ABCD2 score improved risk stratification only to a small degree.3 To advance the ABCD2 score, the presence of ischemic injuries on MRI and carotid stenosis has been incorporated into a new score, the ABCD3-I score.4 The limited availability of urgent MRI, however, prevents the ABCD3-I score from being widely used. Consequently, there is a growing interest in accurate, rapidly measurable, and affordable blood markers, which may improve the ABCD2 score, without incorporating more comprehensive and time-consuming investigations for emergency risk stratification. A recent small single-center study showed that copeptin provided additional information beyond the ABCD2 score in predicting cerebrovascular events within 3 months after TIA.5 However, before implementing copeptin in clinical practice, these results need to be validated in several studies. This study assesses copeptin for the prediction of cerebrovascular events within 3 months after TIA by comparing copeptin to the ABCD2 and to the ABCD3-I score among patients with available MRI.
The study (ClinicalTrials.gov: NCT00878813) was conducted according to the Declaration of Helsinki and approved by the institutional review boards of the 3 participating centers. All patients or their welfare guardians provided written informed consent for the collection of data, blood samples, and analyses.
Study Design and Cohort Description
We described the design of this multicenter prospective cohort study in detail previously.6 We included patients with TIA who presented to the emergency department of University Hospitals of (1) Bern, (2) Charité Universitätsmedizin Berlin, and (3) Goethe University, Frankfurt am Main, Germany. All patients included in this study represent a novel cohort that has not been analyzed or reported on previously. TIA was defined as neurological dysfunction caused by a focal brain ischemia with symptoms lasting ≤24 hours, irrespective of whether diffusion-weighted imaging showed an ischemic lesion. We chose the time-based, instead of tissue based, definition of TIA to reflect the emergency setting, where MRI is not always available. Computed tomography or MRI was performed on admission. If MRI was performed, the presence of diffusion-weighted imaging lesions was assessed by neuroradiologists. Information such as cardiac and neurovascular ultrasound and 24-hour ECGs were collected to define TIA etiology according to the Trial of Org 10172 in Acute Stroke Treatment (TOAST) classification.7 Comorbidities were assessed on admission by the modified Charlson Comorbidity Index.8
A venipuncture was performed in the emergency room within 24 hours of symptom onset. After centrifugation at 3000g for 20 minutes at room temperature, where copeptin is stable for ≥3 days,9 plasma was aliquoted into Eppendorf tubes, they were labeled with standard freezer-safe materials, and frozen at –70°C. Copeptin plasma levels were assessed in a blinded batch analysis by a chemiluminescence sandwich immunoassay (BRAHMS Thermo Fisher Scientific). The functional assay sensitivity was <1 pmol/L (<20% interassay coefficient of variation, defined as the ratio of the SD to the mean). In 359 healthy individuals, the median concentration of copeptin was 4.2 pmol/L (95% confidence interval [CI], 4.0–4.4 pmol/L).9
Ascertainment of Outcomes
Trained stroke physicians, blinded to copeptin levels, assessed outcome 3 months after TIA with structured telephone interviews. The originally prespecified6 primary end point of the Copeptin for Risk Stratification in Acute Stroke Patients: the CoRisk study was the combined end point of TIA and stroke within 3 months after the index TIA. TIA was defined as neurological dysfunction caused by a focal brain ischemia with symptoms lasting ≤24 hours. Ischemic stroke was defined as an acute neurological deficit lasting >24 hours or leading to death. Because a reliable diagnosis of recurrent TIAs emerged as vague and difficult, we focused, in a post hoc analysis, on stroke alone as end point—a clearer and harder end point. All reports of ischemic stroke were confirmed based on definitive, signed medical notes.
Discrete variables were expressed as counts and continuous variables as medians and interquartile range. Copeptin levels were log- transformed to achieve a normal distribution. Frequency comparisons for categorical measurements were performed by the Fisher exact test. Two-group comparison of continuous baseline data was performed by the Mann–Whitney U test.
Two logistic regression models were calculated to assess the association between copeptin and stroke. The 2 logistic regression models were adjusted for the ABCD2 score and ABCD3-I score, respectively. We report odds ratios (OR) as a measure of association, 95% CI as a measure of uncertainty. For the log-transformed copeptin values, the OR corresponds to a 10-fold increase of copeptin.
With the Hosmer–Lemeshow test, we assessed goodness of fit, and with the jackknife technique the variance and bias of the point estimates (internal validity).
To assess the range of potentially useful copeptin cutoffs, we plotted a receiver operating characteristic curve (ROC) for copeptin with stroke as outcome. We present copeptin cutoffs with an optimal sensitivity and another with an optimal specificity (high sensitivity required for a marker to rule out stroke, high specificity to rule in stroke), which may be used in further validation studies. In addition, in a non–prespecified secondary analysis, we assessed the sensitivity and specificity of a copeptin cutoff (9 pmol/L) found in an independent, previous cohort of patients with TIA.5 To compare the prognostic accuracy of the ABCD2 score alone versus the ABCD2 score combined with copeptin, we calculated the areas under the ROC for the ABCD2 score alone and for the logistic model combining the ABCD2 score with copeptin. The statistical significance of the difference between the 2 areas under the curves (AUCs) was computed with the likelihood-ratio test comparing the 2 underlying logistic models as recommended for nested models.10 The same procedure was followed when comparing the ABCD3-I score alone with the combination of the ABCD3-I with copeptin.
To estimate the incremental information of copeptin levels to the ABCD2 score, we calculated the integrated discrimination improvement and net reclassification improvement (NRI), as proposed by the American Heart and Stroke Association.11 The NRI is calculated by computing the difference, among patients with stroke, between the proportions of patients moving up the ABCD2-risk categories and the proportion of patients moving down, and the corresponding difference in proportions for those without a stroke during follow-up. The NRI is thus the difference between the 2 differences described above.11 We used the following validated, published ABCD2-risk categories within 3 months: (1) 0 to 3 points: low risk (3.1%); (2) 4 to 5 points: medium risk (9.8%); and (3) 6 to 7 points: high risk (17.8%).1 All testings were two-tailed, and P values <0.05 were considered statistically significant. For statistical analysis, we used Stata Release 12 (StataCorp LP, College Station, TX, 2011).
From March 2009 through April 2011, we included 302 patients with TIA, and follow-up was available in all patients (follow-up rate: 100%). At 3 months, the mortality rate in patients without stroke was 0.7% (n=2), 9.1% in those with stroke (n=1). The median age was 69.0 years (interquartile range [IQR], 59.0–78.0), 37.1% of the patients were women. The etiology of the index TIA was (1) large-vessel disease in 27.2% of patients, (2) cardioembolic in 19.2%, (3) small-artery disease in 8.3%, (4) multiple causes in 6.6%, (5) other known causes in 2.6%, and (6) undetermined in 36.1% (Table 1). The median copeptin concentration was 6.1 pmol/L (IQR, 2.8–14.6). MRI was performed in a total of 223 patients (73.8%), and an acute ischemic lesion was identified in 53 patients (23.8% of patients with MRI on admission). Patients with an acute ischemic lesion on MRI had higher copeptin values, although this difference was not statistically significant (6.6 pmol/L; IQR, 3.4–15.8 versus 5.7; IQR, 2.7–14.7; P=0.44).
Primary End Point: Copeptin for the Prediction of Recurrent Cerebrovascular Events
Overall 28 patients (9.3%) had a recurrent cerebrovascular event (ie, TIA or stroke) within 3 months after the index TIA. We found no statistically significant difference in copeptin levels between patients with versus without recurrent cerebrovascular events (5.3 pmol/L; IQR, 1.9–29.4 versus 6.2 pmol/L; IQR, 2.8–14.2; P=0.63; Table I in the online-only Data Supplement).
Post Hoc End Point: Copeptin for the Prediction of Stroke
A total of 11 patients (3.6%) experienced a stroke within 3 months after the index TIA. Copeptin levels were >4-fold higher in patients with stroke as compared with patients without stroke (24.3 pmol/L; IQR, 8.7–63.8 versus 5.8 pmol/L; IQR, 2.8–13.7; P=0.02). Women had lower copeptin levels than men (3.4 pmol/L; IQR, 2.1–9.2 versus 8.5 pmol/L; IQR, 4.0–18.2; P<0.001). Comparing patients with versus without stroke, we found no significant difference in the frequencies across the risk categories of the ABCD2- and ABCD3-I scores (Table 1).
Logistic Regression Models
In univariate regression, a 10-fold increase in copeptin level was associated with an OR for stroke of 3.27 (95% CI, 1.26–8.46; P=0.02). After adjusting for the ABCD2 score, a 10-fold increase in copeptin level was associated with an OR for stroke of 3.39 (95% CI, 1.28–8.96; P=0.01). For instance, the adjusted odds of stroke was, on average, 3.39× higher in a patient with a copeptin level of 30 pmol/L compared with a patient with a copeptin level of 3 pmol/L who presented with the same ABCD2 score (Table 2). We had no evidence for lack of fit of the model (P=0.81). After jackknife resampling to assess the precision of the point estimates, copeptin remained significantly associated with stroke. After adjusting for the ABCD3-I score, the association between copeptin and stroke was similar in magnitude but was no longer statistically significant (OR, 3.28; 95% CI, 0.89–12.04; P=0.07). We had no evidence for lack of fit of the model (P=0.99). In this cohort, both the ABCD2 and ABCD3-I scores were not significantly associated with a subsequent stroke. The multivariate logistic regression models are reported in Table 2.
Sensitivity and Specificity (C-Statistics)
The area under the ROC of copeptin to discriminate between patients with and without stroke was 0.70 (95% CI, 0.53–0.87; Table 3). A copeptin cutoff of ≥1.88 pmol/L had a sensitivity of 100% and a specificity of 12% for stroke after TIA. In other words, none of the 36 patients (12% of the cohort) with a copeptin level <1.88 pmol/L had a stroke (rule out if copeptin <1.88 pmol/L). A copeptin cutoff of ≥53.50 pmol/L had a sensitivity of 27% and a specificity of 90%. In other words, 90% of the 291 patients without stroke had a copeptin level <53.50 pmol/L (rule in if copeptin ≥53.5 pmol/L). In a secondary, not prespecified analysis, we assessed the diagnostic accuracy of the copeptin cutoff derived from a previous, independent cohort.5 The cutoff of ≥9 pmol/L had a sensitivity of 73% (80% in the derivation cohort) and a specificity of 63% (76% in the derivation cohort) when applied to the current cohort.
After the addition of copeptin to the ABCD2 score, the AUC of the ABCD2 score improved from 0.60 (95% CI, 0.46–0.74) to 0.74 (95% CI, 0.60–0.88; P=0.02; Table 3). After the addition of copeptin to the ABCD3-I score, the AUC of the ABCD3-I score increased in similar magnitude, although not significantly (AUC from 0.61 [95% CI, 0.43–0.79] to 0.75 [95% CI, 0.62–0.89]; P=0.08).
Net Reclassification Index and Integrated Discrimination Improvement
After the addition of copeptin to the ABCD2 score, 36.4% of all patients with TIA who experienced a subsequent stroke were correctly moved to a higher risk category and 18.2% falsely to a lower risk category compared with the ABCD2 score alone. Among patients with a stroke after TIA, a net of 18.2% was thus accurately reclassified. Among patients without subsequent stroke, the addition of copeptin to the ABCD2 score correctly moved 25.4% to a lower risk category and 12.4% falsely to a higher risk category, compared with the ABCD2 score alone, leaving a net of 13.0% being accurately reclassified. Overall, copeptin allowed to correctly reclassify 31.2% of patients across the risk categories of the ABCD2 score (NRI, P=0.17). The integrated discrimination improvement showed that, on average, copeptin improved the sensitivity of the ABCD2 score by 2.1% (P=0.03) at fixed specificity, that is, without any undesirable decrease in specificity.
In this prospective multicenter study, copeptin blood levels were not associated with the primary, composite end point defined as recurrent cerebrovascular events (TIA or stroke) within 3 months after TIA. However, higher copeptin levels were significantly associated with the post hoc end point stroke within 3 months after TIA both in univariate analysis and after adjusting for the ABCD2 score. Copeptin improved the prognostic ability of the ABCD2 score for the prediction of ischemic stroke in a net of 31.2% of patients. Although not a surrogate for external validation, the association between copeptin and stroke remained statistically significant even after jackknife resampling within the current cohort (internal validation). These findings endorse the hypothesis that copeptin is a novel and accurate prognostic blood marker for risk stratification among patients with TIA evaluated in the emergency department. Copeptin seems to capture unknown risk factors beyond the ABCD2 score, which are important for an accurate risk classification. However, the exact mechanism behind these associations remains unknown. Copeptin derives from a larger peptide (preprovasopressin) along with arginine-vasopressin, both are systemically released in an equimolar ratio.9 Copeptin is more stable than arginine-vasopressin and its measurement is more reliable. Arginine-vasopressin is a major hypothalamic stress hormone, and the release of stress hormones occurs when the body is exposed to a stressor such as a TIA, which threatens homeostasis.12 We hypothesize that copeptin detects the magnitude of threat to each patient and the individual stress response to that threat. The bigger the stress response, the higher is the risk for a subsequent stroke.
To date, copeptin has been studied mainly as predictor of unfavorable outcome after stroke,13,14 and only 1 prospective study has evaluated copeptin in patients with TIA.5 In the previous study, copeptin was associated with any type of recurrent cerebrovascular event (stroke and TIA), whereas in the current study copeptin was associated with stroke, but not with the combination of stroke and TIA. Probably, the difference between the 2 studies reflects a higher inter-rater agreement for the end point stroke than for the end point TIA. The larger sample size of the present study allowed us to assess copeptin along the criteria for evaluation of novel markers by the American Heart Association.15 All of the patients presented in this study were not previously reported on, and therefore represent a novel cohort.
Few other biomarkers have been studied specifically for risk stratification in patients with TIA. In a monocenter study including 167 patients with TIA, lipoprotein-associated phospholipase A2 (LpPLA2) was associated with a composite end point consisting of stroke or death within 90 days or the identification of a high-risk stroke mechanism requiring specific early intervention.16 An independent TIA cohort (n=166) confirmed the association between LpPLA2 and yet another composite end point within the first 30 days.17 Both studies on LpPLA2, however, do not report whether the addition of LpPLA2 to the ABCD2 score improved the risk classification using the NRI or integrated discrimination improvement, and both used different composite end points instead of stroke alone. Thus, LpPLA2 is not yet validated as marker for risk stratification after TIA.
At least 3 cohort studies evaluated C-reactive protein to predict ischemic events after a TIA, and the results are controversial. One study (n=167) found no association between C-reactive protein and the composite end point of stroke, death, or the identification of a high-risk stroke mechanism within 3 months.16 Another study (n=166) found no significant association between C-reactive protein and stroke within 1 year after TIA.18 A third cohort study (n=194) showed, however, a significant association between C-reactive protein and stroke within the first 2 years from TIA.19
Among patients with TIA, copeptin seems to be the first blood marker to add incremental information beyond the ABCD2 score on the risk of subsequent stroke. These findings are relevant given the high incidence of TIA and their heterogeneous management. In the United States, there is an estimate of 1 emergency department visit for TIA per 1000 people and year,20 similar to the United Kingdom with 0.7 per 1000 people and year.21 Although all patients with TIA need urgent work-up, not all patients with TIA need to be hospitalized. Compared with a specialized outpatient evaluation (TIA-clinic), hospitalization of all patients with TIA is not cost-effective, despite increased in-hospital access to tissue-type plasminogen activator therapy.22 By improving the prognostic value of the ABCD2 score, copeptin could help clinicians in refining the risk stratification for patients with TIA.
Despite the strength of this multicenter prospective study design, our study also has limitations. First, although improvement in AUC and integrated discrimination improvement was statistically significant, the NRI was not, which may reflect limited statistical power. The relatively low event rate is probably a consequence of the immediate work-up on admission to the emergency department along with a rapid start of secondary prevention. Such an acute management reflects the current standard of care and may explain the smaller observed prognostic accuracy of the ABCD2 score compared with the original report on the ABCD2 score.1 Second, MRI, and thus the ABCD3-I score, was available in only two thirds of patients. This reflects, however, the limited availability of urgent MRI in most hospitals. We did not observe a statistically significant association between copeptin and stroke after adjustment for the ABCD3-I score (P=0.07), also most likely because of a limited statistical power while the direction and magnitude of association remained stable. However, this study aimed primarily at comparing copeptin to the ABCD2 score to assist the primary care and emergency physicians in the acute management of patients with TIA. Third, women had lower copeptin levels than men in this cohort as in the healthy population,9 so that we should, ideally, report sex-specific copeptin cutoffs. However, because of the small number of strokes, we were not able to compute reliable sex-specific cutoffs.
In conclusion, copeptin represents a novel, reliable, and promising blood marker to predict stroke after TIA, adding prognostic information beyond the ABCD2 score. Although no blood marker will ever replace the clinical examination, copeptin may, if further validated, help refining and individualizing the management of patients with TIA, potentially reducing healthcare costs while keeping low the risk of stroke after TIA.
Sources of Funding
This study was supported by the Swiss National Science Foundation; an unrestricted research grant from funds off the Federal State of Brandenburg and the European Regional Development Fund to Thermo Fisher Scientific, BRAHMS (Germany); the Fondation Leducq (France); Swisslife Jubiläumsstiftung for Medical Research; Swiss Neurological Society; Fondazione Dr Ettore Balli, Bellinzona (Switzerland); Clinical Trial Unit of the University of Bern (Switzerland); Foundation of the Inselspital Bern (Switzerland); and Foundation Pro Scientia et Arte, Bern (Switzerland). All supporting entities had no role in the collection, management, analysis, and interpretation of the data or the preparation or approval of the article.
Dr De Marchis is supported by the Swiss National Science Foundation (PBBEP3_139388); Swisslife Jubiläumsstiftung for Medical Research; Swiss Neurological Society; Fondazione Dr Ettore Balli, De Quervain research grant, University of Bern; Foundation of the Inselspital Bern; and Foundation Pro Scientia et Arte, Bern (Switzerland). Dr Audebert has served as a speaker and has received consultancy honoraria from Lundbeck A/S, Boehringer Ingelheim, Bayer Vital, BMS, Pfizer, and Takeda. Served as consultant or in the advisory board for Roche and Bayer Vital. He received funding by the German Federal Ministry of Education and Research for the Center for Stroke Research. All not related to this study. Dr Arnold received an unconditional research grant from Thermo Fisher Scientific. Dr Christ-Crain has received speaker honoraria for ThermoFisherScientific, not directly related to this study. Dr Katan received funding from the Swiss National Science Foundation (PZ00P3_142422), Fondation Leduqc, and a unconditional research grant from Thermo Fisher Scientific. Dr Seiler is an employee of Thermo Fisher Scientific. Dr Morgenthaler has been an employee of Thermo Fisher Scientific until June 2011. The other authors report no conflicts.
Guest Editor for this article was Bruce Ovbiagele, MD, MSc, MAS.
The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.114.005584/-/DC1.
- Received March 31, 2014.
- Revision received July 21, 2014.
- Accepted July 24, 2014.
- © 2014 American Heart Association, Inc.
- Easton JD,
- Saver JL,
- Albers GW,
- Alberts MJ,
- Chaturvedi S,
- Feldmann E,
- et al
- De Marchis GM,
- Katan M,
- Weck A,
- Brekenfeld C,
- Mattle HP,
- Buhl D,
- et al
- Adams HP Jr,
- Bendixen BH,
- Kappelle LJ,
- Biller J,
- Love BB,
- Gordon DL,
- et al
- Goldstein LB,
- Samsa GP,
- Matchar DB,
- Horner RD
- Morgenthaler NG,
- Struck J,
- Alonso C,
- Bergmann A
- Hlatky MA,
- Greenland P,
- Arnett DK,
- Ballantyne CM,
- Criqui MH,
- Elkind MS,
- et al
- Cucchiara BL,
- Messe SR,
- Sansing L,
- MacKenzie L,
- Taylor RA,
- Pacelli J,
- et al
- Rothwell PM,
- Coull AJ,
- Silver LE,
- Fairhead JF,
- Giles MF,
- Lovelock CE,
- et al