Abstract
Background and purpose
Premature death after TIA or stroke is more often due to heart disease or cancer than stroke. Previous studies found blood biomarkers not usefully predictive of non-fatal stroke but possibly of all-cause death. This association might be explained by potentially treatable occult cardiac disease or cancer. We therefore aimed to validate the association of a panel of biomarkers with all-cause death, particularly cardiac death and cancer death, despite the absence of associations with risk of non-fatal vascular events.
Methods
15 biomarkers were measured in 929 consecutive patients in a population-based study (Oxford Vascular Study), recruited from 2002 and followed-up to 2013. Associations were determined by Cox-regression. Model discrimination was assessed by c-statistic and the integrated discrimination improvement (IDI).
Results
During 5560 patient-years of follow-up none of the biomarkers predicted risk of non-fatal vascular events. However, soluble tumor-necrosis-factor-α-receptor-1 (sTNFR-1), von-Willebrand-factor (vWF), heart-type-fatty-acid-binding-protein (hFABP) and N-terminal-pro-B-type-natriuretic-peptide (nt-proBNP) were independently predictive of all-cause death (n=361; adjusted HR per SD, 95%CI: hFABP-1.31,1.12-1.56,p=0.002; nt-proBNP-1.34,1.11-1.62,p=0.002; sTNFR-1-1.45,1.26-1.66,p=0.02; vWF-1.19,1.04-1.36,p=0.01). The independent contribution of the four biomarkers taken together added prognostic information and improved model discrimination (IDI=0.028,p=0.0001). Nt-proBNP was most predictive of vascular death (adjusted HR=1.80,95%CI 1.34-2.41,p<0.0001) whereas hFABP predicted cancer deaths (1.64,1.26-2.12,p=0.0002). Associations were strongest in patients without known prior cardiac disease or cancer.
Conclusions
Several biomarkers predicted death of any cause after TIA and minor stroke. Nt-proBNP and hFABP might improve patient selection for additional screening for occult cardiac disease or cancer respectively. However, our results require validation in future studies.
Keywords: stroke, TIA, mortality, biomarkers
Introduction
Prevention of recurrent stroke after TIA or ischemic stroke tends to be the main focus of initial investigations and treatment. However, premature death after TIA or stroke is more likely to occur as a consequence of heart disease or cancer than of stroke.1-3 Indeed, both occult heart disease4,5 and occult cancer6,7 are also often causally associated with incident TIA and stroke. Currently, a full diagnostic work-up for occult cardiac disease or cancer is not recommended for all patients with TIA or stroke. However, early diagnosis might allow effective treatment or even cure.
Previous studies have reported on the prognostic value of various biomarkers related to different disease pathways, such as inflammation, cardiac function or thrombosis, for cardiovascular events and death in subjects with and without pre-existent cardiovascular disease.8-11 However, work investigating the value of biomarkers for recurrent vascular events or death after TIA and stroke has been conflicting.12-14 Overall, the predictive value of most biomarkers appears to be larger for fatal vascular events and all-cause death than for non-fatal vascular events.12,15-18 This finding raises the possibility that some biomarkers might be associated with serious underlying conditions rather than directly with recurrent vascular events. Levels of most biomarkers increase with age19-21 and increased levels of pro-inflammatory mediators, such as Interleukin-6 (IL6) or tumor necrosis factor-α (TNFα), have been reported in many age-related diseases such as atherosclerosis and cancer22. However, previous studies of biomarkers in TIA and stroke did not report associations of biomarkers with cause-specific mortality.12,23
We therefore aimed to validate in a prospective population-based cohort of patients with TIA or minor ischemic stroke the association of a panel of biomarkers related to inflammation, thrombosis and cardiac or neuronal function or injury with all-cause death, particularly cardiac death and cancer death, despite the absence of associations with risk of non-fatal vascular events.
Materials and methods
Patients
The methods of the Oxford Vascular Study (OXVASC) have been reported previously.24 Briefly, OXVASC is an ongoing population-based study of the incidence and outcome of cerebrovascular, cardiovascular and peripheral vascular events. The study population comprises all 92,728 individuals, irrespective of age, registered with 100 family physicians in nine general practices in Oxfordshire, UK. TIA was defined according to the WHO definition25 as an acute loss of focal brain or monocular function with symptoms lasting less than 24 hours and which are thought to be caused by inadequate brain or ocular blood supply as a result of arterial thrombosis, low flow or embolism associated with arterial, cardiac or hematologic disease. Minor stroke was defined as NIHSS≤3 points. A detailed description of patient ascertainment and follow-up is given in the Supplementary methods. The methodology of the Oxford vascular study was approved by the Oxfordshire Research Ethics Committee. All patients provided informed consent.
Classification of deaths was blind to biomarker results and was done by review of clinical records and death certificates in the study practices. Practice-specific listings of all ICD-10 (International Classification of Diseases) death codes were also obtained from central registers. Causes of death were classified into death of vascular cause and non-vascular death. Vascular death was sub-classified into stroke-related, cardiac and other. Non-vascular death was stratified into cancer related death, other systems death (respiratory, digestive system, infections, metabolic) and death due miscellaneous causes including neurodegenerative diseases, falls, senility and accidents. Details on definitions used for classification of death are given in Supplementary Table I.
Blood sampling and processing
We analysed a panel of biomarkers related to inflammation [i.e. IL6, C-reactive protein (CRP), neutrophil-gelatinase associated lipocalin (NGAL), soluble tumor necrosis factor-α receptor-1 (sTNFR-1)], thrombosis [i.e. Thrombomodulin (TM), fibrinogen, P-Selectin, D-dimer, von Willebrand factor antigen (vWF), protein-Z (PZ)], atherogenesis [i.e. anti-phosphorylcholin (anti-PC)], cardiac function [i.e. n-terminal pro-B-type natriuretic peptide (nt-proBNP)], cardiac and/or neuronal injury [i.e. neuron-specific enolase (NSE), heart-type fatty acid binding protein (hFABP)] and neural regeneration [i.e. brain-derived neurotrophic factor (BDNF)]. Samples were taken as soon as possible after the event either whilst the patient was an inpatient or attending the OXVASC outpatient emergency TIA and stroke clinic. In a consecutive series of 80 patients biomarkers were measured also at the 1-year follow-up examination. A detailed description of assay methods is given in the Supplementary methods.
The intra- and inter-assay coefficients of variation of each assay are shown in Supplementary Table II. All assays were performed blind to study status.
Statistics
Continuous data were given as mean (SD) or median (IQR) as appropriate, categorical data were given as n (%). For analysis of categorical data the chi2-test or Fisher’s exact test and for analysis of continuous data the Mann-Whitney U test was used, as appropriate. Correlations of biomarkers or of biomarkers and age were calculated by Spearman rank correlations. For survival analysis Cox proportional hazard models were used. Model discrimination was determined by c-statistics by calculating the area under receiver operating characteristic curve (AUC) and the integrated discrimination index (IDI).26, 27 Model calibration was tested with the Hosmer-Lemeshow test with 10 risk groups. All calculations were done using SPSS 20.0 (SPSS Inc., Chicago, Illinois).
A detailed description of the statistical methods is given in the Supplementary Methods.
Results
929 consecutive eligible patients with TIA (n=436) or minor stroke (n=493) were recruited from 2002-2007 and followed-up until 2013. The median age at entry was 74 years (64-83) and 469 (51%) were female (Table 1). The median time from event to sampling was 5 days (IQR 2-12). During 5560 patient-years of follow-up (median=6.4 years; IQR=3.6-8.4), 200 recurrent events occurred (148 ischemic strokes and 52 myocardial infarctions), of whom 31 died within 30 days of the recurrent event (20 strokes and 11 myocardial infarctions). Overall, 361 (39%) patients died during follow-up; 151 (42%) of a vascular cause, 184 of a non-vascular cause and 26 of uncertain cause.
Table 1.
Baseline characteristics of all patients with TIA and minor stroke and stratified into survivors and non-survivors
| All | Survivors | Non-survivors | ||
|---|---|---|---|---|
| Variable* | n=929 | n=568 | n=361 | P-value† |
| Female | 469(51) | 286(50) | 183(51) | 0.95 |
| Age,median (IQR) | 74(64-83) | 68(59-77) | 82(75-86) | <0.0001 |
| Days to sample, median (IQR) | 5(2-12) | 5(2-12) | 5(3-12) | 0.16 |
| Hypertension | 528(57) | 304(54) | 224(62) | 0.012 |
| Diabetes | 116(12) | 72(13) | 44(12) | 0.92 |
| Previous history of CVD | 117(13) | 42(7) | 75(21) | <0.0001 |
| Previous history of MI | 91(10) | 38(7) | 53(15) | 0.0001 |
| Previous history of PAD | 69(7) | 24(4) | 45(12) | <0.0001 |
| Atrial fibrillation | 127(14) | 48(8) | 79(22) | <0.0001 |
| Current smoker | 144(16) | 93(16) | 51(14) | 0.40 |
| Hyperlipidaemia | 314(34) | 202(36) | 112(31) | 0.16 |
| Previous antiplatelet therapy‡ | 364(39) | 193(34) | 171(47) | <0.0001 |
| Previous antihypertensive therapy§ | 526(57) | 288(51) | 238(67) | <0.0001 |
| Previous Statin therapy§ | 232(25) | 141(25) | 91(25) | 0.91 |
| NIHSS index event, median (IQR) | 0(0-1) | 0(0-1) | 0(0-2) | 0.0012 |
| Aetiology of stroke | <0.0001 | |||
| Cardioembolism | 151(16) | 63(11) | 88(24) | |
| Large artery | 110(12) | 61(11) | 49(14) | |
| Small artery | 159(17) | 104(18) | 55(15) | |
| Undetermined etiology | 414(45) | 293(52) | 121(34) | |
| Unknown (incomplete evaluation) | 54(6) | 20(4) | 34(9) | |
| Multiple | 14(2) | 4(1) | 10(3) | |
| Other | 27(3) | 23(4) | 4(1) |
Numbers are n(%) unless otherwise stated
P-values are from Fisher's exact tests, Chi-squared tests or Mann-Whitney U tests,as appropriate
N=927;
n=920
CVD=cerebrovascular disease, MI=myocardial infarction, PAD=peripheral artery disease, NIHSS=NIH stroke scale
Biomarker levels are given in Table 2. Samples were analysed with a multiple immunoassay system and by using specific assays for thrombotic markers (with the exception of TM, which was included into the multiple immunoassay system), nt-proBNP and anti-PC antibodies. For analysis with the two arrays of the multiple immunoassay system 55 had insufficient sample volumes or were lost due to assay failure. Nt-proBNP was measured in 886 patients after exclusion of 43 samples with insufficient sample volume. Thrombotic markers (with the exception of TM) were measured in a maximum of 735 patients after further exclusion of cases with inadequate samples or on anticoagulant therapy. Patients with missing biomarker-levels did not differ with respect to age, sex and stroke severity from patients with established biomarker levels.
Table 2.
Levels (median and IQR) of biomarkers in all patients with TIA and minor stroke and stratified into survivors and non-survivors
| All | Survivors | Non-survivors | ||
|---|---|---|---|---|
| Biomarker | Median(IQR) | Median(IQR) | Median(IQR) | P-value* |
| Inflammatory markers(n=874) | ||||
| IL6 pg/ml | 1.52(0.77-3.51) | 1.20(0.64-2.34) | 2.45(1.27-6.48) | <0.0001 |
| CRP mg/l | 2.52(1.14-6.58) | 2.15(1.04-5.11) | 3.42(1.47-10.0) | <0.0001 |
| NGAL ng/ml | 735(522-1026) | 678(479-931) | 882(646-1172) | <0.0001 |
| STNFR-1 ng/ml | 0.81(0.60-1.11) | 0.72(0.55-0.96) | 0.97(0.74-1.37) | <0.0001 |
| Thrombotic markers or anti-atherogenic markers | ||||
| TM ng/ml(n=874) | 1.65(1.34-2.07) | 1.58(1.30-1.92) | 1.76(1.44-2.26) | <0.0001 |
| Fibrinogen g/L(n=710) | 377(314-446) | 368(310-424) | 390(321-466) | 0.0011 |
| VWF iu/ml(n=715) | 153(119-195) | 140(112-178) | 172(135-223) | <0.0001 |
| P-selectin ng/ml(n=735) | 32(17-46) | 31(17-43) | 36(19-52) | 0,0016 |
| PZ ng/ml(n=697) | 1.77(1.31-2.18) | 1.83(1.33-2.24) | 1.73(1.28-2.07) | 0.039 |
| D-dimer ng/ml(n=696) | 475(291-812) | 416(262-707) | 604(401-962) | <0.0001 |
| Anti-PC U/ml(n=682) | 42.7(25.6-74.0) | 47.1(27.7-78.2) | 39.1(23.0-65.1) | 0.0069 |
| Markers of cardiac or neuronal function/injury(n=877) | ||||
| Nt-proBNP pmol/l† | 536(244-1209) | 406(191-795) | 909(408-1807) | <0.0001 |
| hFABP ng/ml | 2.84(2.01-3.99) | 2.51(1.82-3.31) | 3.56(2.49-5.27) | <0.0001 |
| NSE ng/ml | 6.27(4.27-9.98) | 6.11(4.18-9.66) | 6.85(4.49-10.50) | 0.045 |
| BDNF pg/ml | 789(473-1314) | 779(482-1243) | 793(456-1440) | 0.69 |
Mann-Whitney U test;
n=88
Correlations between biomarkers-levels at baseline are shown in Supplementary Table III. There was good cross-correlation within the subset of inflammatory markers. In addition, fibrinogen, vWF and D-Dimer correlated modest but statistically significant with each other and the inflammatory markers. HFABP correlated well with sTNFR-1 (r=0.49) and to a lesser extent with the other inflammatory markers and nt-proBNP. Anti-PC did not correlate with any other biomarker. All biomarkers but IL6, NGAL, NSE, BDNF and P-selectin were statistically significantly correlated between baseline- and one-year (Supplementary Table IV). With the exception of P-selectin, BDNF and NSE, all biomarkers were positively correlated with age, the strongest correlation being for hFABP (r=0.48, Supplementary Table IV).
None of the tested biomarkers were predictive of recurrent non-fatal ischemic stroke or myocardial infarction (Supplementary Table V). In addition, there was no evidence of any significant predictive value for events within the first year after recruitment (data not shown).
All biomarkers except BDNF differed significantly between survivors and non-survivors (Table 2). Results were similar for univariate analysis of associations with all-cause death in a Cox model either as continuous variables or categorised into tertiles (Table 3 and Supplementary Table VI, respectively). After fully adjusting for demographics (age, sex) previous therapy (antiplatelet agents, antihypertensive agents, statin therapy) and risk factors (hypertension, diabetes, atrial fibrillation, smoking, previous stroke, previous myocardial infarction, previous peripheral artery disease, hyperlipidaemia) sTNFR-1, NGAL, CRP, IL6, vWF, nt-proBNP and hFABP remained predictive of death of any cause (model 3 in Table 3 and Supplementary Table VI, respectively). The strongest associations were detected for sTNFR-1 (per SD-Ln-unit increase: HR=1.45, 95% CI: 1.26-1.66, p<0.0001; top vs. bottom-tertile: HR=2.25, 1.65-3.06, p<0.0001), nt-proBNP (per SD-Ln-unit increase: HR=1.44, 1.21-1.72, p=0.0005; top vs. bottom-tertile: HR=1.85, 1.35-2.53, p<0.0001) and hFABP (per SD-Ln-unit increase: HR=1.37, 1.19-1.57, p=0.0002; top vs. bottom-tertile: HR=1.33, 0.96-1.84, p=0.004). Of the hemostatic biomarkers only vWF remained predictive (per SD-increase: HR=1.26, 1.12-1.42, p=0.002; top vs. bottom-tertile: HR=1.47, 1.08-1.99, p=0.487). Results were similar in a sensitivity analysis including in the Cox-regression all significant biomarkers from model 3 stratified into ten-year age bands (Supplementary Table VII). Additional analysis of significant biomarkers from model 3 in a separate model including age, sex and the CCI as a validated measure of co-morbidity in the context of stroke mortality also revealed similar results (Supplementary Table VIII). Among the biomarkers predictive of all-cause death, we only found for sTNFR-1 some association with case fatality at 30 days (OR=1.75, 95%CI: 0.98-312, p=0.06) and 90 days (OR 1.93, 1.09-3.41; p=0.023) after the recurrent event (Supplementary Table IX).
Table 3.
Univariate and multivariate associations (according to 3 models*) of each biomarker with risk of all-cause death
| Unadjusted | Model 1 | Model 2 | Model 3 | |||
|---|---|---|---|---|---|---|
| Biomarker† | HR(95% CI) | P-value‡ | HR(95% CI) | HR(95% CI) | HR(95% CI) | P-value‡ |
| Inflammatory markers | ||||||
| IL6 | 1.31(1.20-1.44) | <0.0001 | 1.20(1.09-1.32) | 1.18(1.07-1.30) | 1.16(1.05-1.29) | 0.073 |
| CRP | 1.34(1.20-1.51) | <0.0001 | 1.26(1.12-1.42) | 1.25(1.11-1.41) | 1.23(1.09-1.39) | 0.016 |
| NGAL | 1.55(1.37-1.75) | <0.0001 | 1.36(1.19-1.54) | 1.31(1.15-1.50) | 1.26(1.10-1.44) | 0.013 |
| STNFR-1 | 1.87(1.67-2.10) | <0.0001 | 1.46(1.29-1.65) | 1.47(1.29-1.67) | 1.45(1.26-1.66) | <0.0001 |
| Thrombotic or anti-atherogenic markers | ||||||
| TM | 1.19(1.09-1.29) | 0.0006 | 1.05(0.94-1.19) | 1.05(0.93-1.19) | 1.06(0.94-1.19) | 1.00 |
| Fibrinogen | 1.27(1.14-1.41) | 0.0001 | 1.15(1.02-1.29) | 1.14(1.02-1.28) | 1.11(0.99-1.25) | 1.00 |
| vWF | 1.50(1.36-1.65) | <0.0001 | 1.31(1.17-1.48) | 1.31(1.16-1.47) | 1.26(1.12-1.42) | 0.002 |
| P-selectin | 1.19(1.05-1.35) | 0.111 | 1.17(1.04-1.32) | 1.15(1.02-1.30) | 1.13(1.00-1.28) | 0.719 |
| PZ | 0.90(0.80-1.02) | 1.00 | 1.00(0.89-1.13) | 1.00(0.89-1.13) | 1.01(0.90-1.15) | 1.00 |
| DD | 1.41(1.26-1.58) | <0.0001 | 1.11(0.98-1.25) | 1.09(0.97-1.24) | 1.08(0.95-1.22) | 1.00 |
| Anti-PC | 0.83(0.74-0.94) | 0.036 | 0.93(0.83-1.05) | 0.94(0.83-1.05) | 0.93(0.82-1.04) | 1.00 |
| Markers of cardiac or neuronal function/injury | ||||||
| Nt-proBNP | 2.03(1.74-2.37) | <0.0001 | 1.45(1.23-1.70) | 1.41(1.19-1.67) | 1.44(1.21-1.72) | 0.0005 |
| hFABP | 1.91(1.72-2.13) | <0.0001 | 1.36(1.20-1.55) | 1.37(1.20-1.57) | 1.37(1.19-1.57) | 0.0002 |
| NSE | 1.13(1.02-1.25) | 0.352 | 1.12(1.01-1.25) | 1.10(0.99-1.22) | 1.10(0.99-1.23) | 1.00 |
| BDNF | 1.01(0.91-1.13) | 1.00 | 1.06(0.94-1.19) | 1.06(0.94-1.19) | 1.04(0.92-1.16) | 1.00 |
Model 1: Adjusted for age and sex; Model 2: adjusted for variables in Model 1 plus previous therapy with antiplatelet agents, statins or antihypertensive agents; Model 3: adjusted for variables in Model 2 plus hypertension, diabetes, previous myocardial infarction, previous peripheral vascular disease, previous ischemic stroke, atrial fibrillation, current smoker, hyperlipidaemia
Given per SD(Ln); Fibrinogen, vWF and PZ are given as SD due to normal distribution of levels
Bonferroni-corrected p-values given
Multivariate associations of biomarkers with the separate risks of vascular and non-vascular death are given in Supplementary Table X. After full adjustment for demographics and risk factors, sTNFR-1, NGAL, vWF, hFABP and nt-proBNP were predictive of vascular death; the strongest association was detected for nt-proBNP (adjusted HR per SD=1.80, 95% CI: 1.34-2.41, p<0.0001). With the exception of hFABP, all of these biomarkers were associated with known cardiac disease at baseline (Supplementary Table XI). Although biomarker-associations were already adjusted for previous vascular risk factors we repeated the analysis after exclusion of patients with known cardiac disease to reduce any residual confounding. Again, all biomarkers except hFABP remained significantly associated with vascular death (Supplementary Table X). Nt-proBNP, hFABP, sTNFR-1 and vWF were similarly predictive of cardiac death, whereas nt-proBNP was the strongest predictor of stroke-related death. IL6, CRP, NGAL, sTNFR-1 and hFABP were predictive of non-vascular death; the strongest associations were detected for hFABP (adjusted HR=1.50, 1.22-1.83, p<0.0001) and sTNFR-1 (1.47, 1.21-1.78, p<0.0001). HFABP (1.61, 1.23-2.10, p=0.0005) and sTNFR-1 (1.41, 1.07-1.87, p=0.014) were also predictive of cancer-related death. For hFABP and cancer death, exclusion of patients with known cancer at baseline increased the predictive value (2.00, 1.45-2.78, p<0.0001).
The inflammatory biomarker with the highest hazard ratio (sTNFR-1) along with the other significant biomarkers of model 3 [i.e. vWF (thrombotic), hFABP (cardiac and neuronal injury), nt-proBNP (cardiac function)] were selected for a conjoined model incorporating all covariates used in model 3. All four biomarkers contributed independent prognostic information for all-cause mortality (adjusted HR & 95%CI: sTNFR-1: 1.21, 1.03-1.42, p=0.023; hFABP: 1.31, 1.12-1.56, p=0.002; vWF: 1.19, 1.04-1.36, p=0.014; nt-proBNP: 1.34, 1.11-1.62, p=0.002). A stepwise model produced similar results (data not shown). In an analysis of the mortality-risk at 2, 3 and 5 years of follow-up stratified by age, sTNFR-1, vWF, hFABP and nt-proBNP were predictive of all-cause death at each time point and assumption of proportional hazards was valid for each biomarker (Table 4). For nt-proBNP the predictive value was greatest at older ages. An apparent decrease of the predictive value of sTNFR-1, vWF and hFABP in the oldest age group appeared to be artefactual due to the very small number of survivors with high levels of these biomarkers.
Table 4.
Age-adjusted risk of death risk in association with the 4 most predictive biomarkers from different pathophysiological groups (STNFR-1, VWF, hFABP, nt-proBNP) stratified by age bands after 2, 3 and 5years of follow-up
| 2 year | 3 year | 5 year | ||||||
|---|---|---|---|---|---|---|---|---|
| Biomarker* | Age band (years) | Deaths (n) | HR(95% CI) | P-value | HR(95% CI) | P-value | HR(95% CI) | P-value |
| STNFR-1 | ≤70 | 22 | 3.57(1.78-7.13) | 0.0003 | 3.24(1.75-5.99) | 0.0002 | 2.72(1.67-4.44) | <0.0001 |
| >70-80 | 52 | 2.09(1.16-3.78) | 0.015 | 2.30(1.43-3.70) | 0.0006 | 1.73(1.22-2.46) | 0.0020 | |
| >80 | 162 | 1.74(1.32-2.29) | <0.0001 | 1.58(1.25-2.01) | 0.0002 | 1.44(1.20-1.73) | 0.0001 | |
|
| ||||||||
| vWF | ≤70 | 24 | 2.85(1.61-5.03) | 0.0003 | 2.45(1.48-4.04) | 0.0005 | 1.93(1.29-2.88) | 0.0013 |
| >70-80 | 47 | 1.41(0.96-2.07) | 0.081 | 1.37(1.01-1.86) | 0.044 | 1.27(0.98-1.64) | 0.069 | |
| >80 | 128 | 1.57(1.20-2.06) | 0.0009 | 1.30(1.03-1.63) | 0.026 | 1.39(1.16-1.67) | 0.0004 | |
|
| ||||||||
| nt-proBNP | ≤70 | 23 | 1.40(0.66-2.99) | 0.38 | 1.54(0.77-3.07) | 0.22 | 0.94(0.64-1.37) | 0.75 |
| >70-80 | 52 | 1.29(0.68-2.45) | 0.43 | 1.37(0.81-2.31) | 0.24 | 1.30(0.88-1.94) | 0.19 | |
| >80 | 165 | 2.26(1.62-3.14) | <0.0001 | 2.17(1.63-2.89) | <0.0001 | 2.13(1.66-2.72) | <0.0001 | |
|
| ||||||||
| hFABP | ≤70 | 23 | 2.00(1.09-3.67) | 0.026 | 2.25(1.33-3.79) | 0.0024 | 2.13(1.39-3.25) | 0.0005 |
| >70-80 | 52 | 1.80(1.11-2.91) | 0.018 | 1.68(1.12-2.50) | 0.012 | 1.80(1.32-2.44) | 0.0002 | |
| >80 | 163 | 1.33(1.06-1.67) | 0.014 | 1.27(1.03-1.56) | 0.025 | 1.23(1.03-1.46) | 0.021 | |
Given as SD(Ln) with the exception of vWF, which is given as SD due to normal distribution of levels
The independent contribution of the four biomarkers taken together added more prognostic information than the established clinical risk factors used in model 3 (clinical risk factors: chi2=43, p=0.002; 4 biomarkers: chi2=55, p<0.0001; Supplementary Table XII). Results were similar and not qualitatively different for a model using the CCI instead of the vascular risk factors from model 3 (CCI: chi2=37, p=0.0001, four biomarkers: chi2=80, p<0.0001), or for a model including all risk factors (i.e. adding the risk factors not included into the CCI to a model containing the CCI; data not shown).
For analysis of discrimination we calculated AUC for age, the CCI and the four selected biomarkers (nt-proBNP, hFABP, sTNFR-1 and vWF) separately and in combination (Supplementary Table XIII). Addition of the four biomarkers improved discrimination of the model significantly (model containing age and the CCI plus the four biomarkers: AUC 0.844 versus 0.824 without the biomarkers; IDI=0.028, p=0.0001). Model calibration did not show any significant deviation between predicted and observed risk (p=0.82).
Discussion
In this population based cohort study, we showed that high levels of biomarkers related to inflammation (sTNFR-1, Il-6, CRP, NGAL), cardiac function (nt-proBNP) cardiac and neuronal injury (hFABP) and thrombosis (vWF) were predictive of death in patients with TIA or minor ischemic stroke independently of established risk factors. In common with previous studies, biomarkers were predictive of fatal vascular events and all-cause death but not for non-fatal events.12,14-18 An advantage of our study is that we excluded patients with major stroke making effects on mortality due to reverse causation less likely. In a multivariate model incorporating the biomarkers with the highest predictive value from different pathophysiological groups (sTNFR-1, vWF, nt-proBNP and hFABP), all four biomarkers provided independent prognostic information. Taken together, the predictive value of these biomarkers was larger than that of established clinical risk factors, highlighting the potential of combined biomarker-panels focusing on multiple pathophysiological pathways. Importantly, combination of nt-proBNP, hFABP, sTNFR-1 and vWF improved risk discrimination between survivors and non-survivors. However, these results have to be validated in future studies and do not imply information on clinical value of biomarkers for prediction of death in patients with TIA or minor stroke at the current stage.27 We therefore did not include a net reclassification analysis into the present study.
Previous studies investigating the predictive value of the natriuretic peptides BNP and nt-proBNP after stroke or TIA were limited by small sample size, short duration of follow-up, and inclusion of major strokes.23 In our study the predictive value of nt-proBNP after TIA or minor stroke clearly derived from prediction of vascular death. The predictive value increased after exclusion of patients with previous history of cardiac disease, suggesting that measurement of nt-proBNP might be a useful marker to target more systematic cardiac evaluation after TIA/minor stroke. The fact, that the prognostic value of nt-proBNP was greatest in elderly patients is in line with data deriving from patients with myocardial infarction.21 An increase of nt-proBNP-levels with age has been reported in healthy adults28 and might reflect a higher prevalence of subclinical cardiac disease and/or impaired renal function29. In our cohort, the number of patients with severe renal failure was low (1%).
HFABP has been reported to be a marker of myocardial ischemia and might serve as a potential marker of ischemic damage in stroke.30 High concentrations have been associated with mortality after acute coronary syndrome31 and in the post-acute phase after myocardial infarction32. Interestingly, FABPs might also play a role in cancerogenesis33,34 and expression of hFABP has been associated with tumor aggressiveness and development of metastasis35. In our data, the association of hFABP with cancer death mainly drove the predictive value of hFABP for all-cause death and the predictive value increased after exclusion of patients with previously known malignant disease. HFABP might be of value for identification of occult cancer in patients with TIA/minor stroke. However, these findings require replication.
In our study, the prognostic value of inflammatory biomarkers was similar for vascular and non-vascular death. Among the inflammatory markers sTNFR-1 had the strongest predictive value and was also the only marker to be associated with case fatality of recurrent vascular events. Levels of sTNFR-1 were positively correlated with age, but the association of high sTNFR-1 levels with death was detectable across all age groups. TNFα-activation - of which sTNFR-1 is a good correlate due to a longer half-life in plasma36 - has been associated previously with the phenomenon of immunosenescence.22,37 The association of sTNFR-1 and other inflammatory biomarkers with earlier death in our study might therefore reflect an association with “biological age”.
One particular strength of our study is the high rate of statin use (84%), antiplatelet therapy or oral anticoagulation (97%), and strict blood pressure control (83%)38, such that potential for confounding by variation in risk factor control was limited. Our study did, however, have some limitations. First, in the majority of patients we only had a single measurement of biomarkers after the index event. Multiple measurements of biomarkers would offer the opportunity to identify whether patients with persistently increased biomarker-concentrations are at especially increased risk of death. However, the correlation between baseline and one-year samples for the main prognostic markers (sTNFR-1, vWF, hFABP and nt-proBNP) was good. Second, we were unable to measure all biomarkers in the entire patient-cohort. Third, even though our sample size and number of deaths were substantial we could adjust only for a limited number of other predictors of death and it is possible that some other confounding factors were unmeasured. However, our primary interest is in the predictive utility of these biomarkers rather than any causal association with death. Fourth, our observation might not be specific for TIA and minor stroke and it is possible that some biomarkers might predict death in other clinical settings. Finally, although there were some subtype specific effects (the predictive value of nt-proBNP was pronounced in patients with cardioembolic etiology of the index event; data not shown) value of analysis by each etiological subtype of TIA and stroke was limited as numbers of endpoints were too small for full adjustment in each group.
Conclusion
In our study high levels of sTNFR-1, IL6, CRP, NGAL, vWF, nt-proBNP and hFABP were predictive of earlier death in patients with TIA or minor stroke independently of established demographic factors or clinical risk factors. Nt-proBNP specifically predicted vascular death and hFABP death of malignant disease with the predictive value being higher in patients without pre-existing cardiac disease or cancer. If our results could be validated in future studies, nt-proBNP and hFABP would be promising biomarkers for patient-selection for screening of underlying cardiac disease or occult cancer in TIA/minor stroke. In addition, in patients with TIA or minor stroke identification of subgroups at particular high risk of cardiac death might influence further therapeutic decisions.
Supplementary Material
Acknowledgments
We are grateful to all patients who took part in the study. We thank all primary care practices and physicians who collaborate with Oxford Vascular Study (OXVASC). We also acknowledge the use of the facilities of the Acute Vascular Imaging Centre, Oxford and the Oxford Haemophilia and Thrombosis Centre Laboratory.
Sources of Funding
The Oxford Vascular Study has been funded by the Wellcome Trust, Wolfson Foundation, UK Stroke Association, British Heart Foundation, Dunhill Medical Trust, National Institute of Health Research, Medical Research Council, and the National Institute of Health Research Oxford Biomedical Research Centre. Dr Greisenegger received funding by the European Neurological Society and the Austrian Society of Neurology.
Footnotes
Disclosures
None.
References
- 1.de Jong G, van Raak L, Kessels F, Lodder J. Stroke subtype and mortality. a follow-up study in 998 patients with a first cerebral infarct. J Clin Epidemiol. 2003;56:262–268. doi: 10.1016/s0895-4356(02)00572-3. [DOI] [PubMed] [Google Scholar]
- 2.Clark TG, Murphy MFG, Rothwell PM. Long term risks of stroke, myocardial infarction, and vascular death in ‘low risk’ patients with a non-recent transient ischemic attack. J Neurol Neurosurg Psychiatry. 2003;74:577–580. doi: 10.1136/jnnp.74.5.577. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.van Wijk I, Kappelle LJ, van Gijn J, Koudstaal PJ, Franke CL, Vermeulen M, et al. Long-term survival and vascular event risk after transient ischemic attack or minor ischemic stroke: a cohort study. Lancet. 2005;365:2098–2104. doi: 10.1016/S0140-6736(05)66734-7. [DOI] [PubMed] [Google Scholar]
- 4.Wolf PA, D'Agostino RB, Belanger AJ, Kannel WB. Probability of stroke: a risk profile from the Framingham Study. Stroke. 1991;22:312–318. doi: 10.1161/01.str.22.3.312. [DOI] [PubMed] [Google Scholar]
- 5.Hermann DM, Gronewold J, Lehmann N, Moebus S, Jöckel K-H, Bauer M, et al. Coronary artery calcification is an independent stroke predictor in the general population. Stroke. 2013;44:1008–1013. doi: 10.1161/STROKEAHA.111.678078. [DOI] [PubMed] [Google Scholar]
- 6.Taccone FS, Jeangette SM, Blecic SA. First-Ever Stroke as Initial Presentation of Systemic Cancer. J Stroke Cerebrovasc Dis. 2008;17:169–174. doi: 10.1016/j.jstrokecerebrovasdis.2008.01.007. [DOI] [PubMed] [Google Scholar]
- 7.Kwon H-M, Kang BS, Yoon B-W. Stroke as the first manifestation of concealed cancer. J Neurol Sci. 2007;258:80–83. doi: 10.1016/j.jns.2007.02.035. [DOI] [PubMed] [Google Scholar]
- 8.Zethelius B, Berglund L, Sundström J, Ingelsson E, Basu S, Larsson A, et al. Use of multiple biomarkers to improve the prediction of death from cardiovascular causes. New Engl J Med. 2008;358:2107–2116. doi: 10.1056/NEJMoa0707064. [DOI] [PubMed] [Google Scholar]
- 9.Blankenberg S, Zeller T, Saarela O, Havulinna AS, Kee F, Tunstall-Pedoe H, et al. Contribution of 30 Biomarkers to 10-Year Cardiovascular Risk Estimation in 2 Population Cohorts: The MONICA, Risk, Genetics, Archiving, and Monograph (MORGAM) Biomarker Project. Circulation. 2010;121:2388–2397. doi: 10.1161/CIRCULATIONAHA.109.901413. [DOI] [PubMed] [Google Scholar]
- 10.Battistoni A, Rubattu S, Volpe M. Circulating biomarkers with preventive, diagnostic and prognostic implications in cardiovascular diseases. Int J Cardiol. 2012;157:160–168. doi: 10.1016/j.ijcard.2011.06.066. [DOI] [PubMed] [Google Scholar]
- 11.Morange PE, Bickel C, Nicaud V, Schnabel R, Rupprecht HJ, Peetz D, et al. Hemostatic Factors and the Risk of Cardiovascular Death in Patients With Coronary Artery Disease: The AtheroGene Study. Arterioscler Thromb Vasc Biol. 2006;26:2793–2799. doi: 10.1161/01.ATV.0000249406.92992.0d. [DOI] [PubMed] [Google Scholar]
- 12.Whiteley W, Jackson C, Lewis S, Lowe GDO, Rumley A, Sandercock P, et al. Association of circulating inflammatory markers with recurrent vascular events after stroke: a prospective cohort study. Stroke. 2010;42:10–16. doi: 10.1161/STROKEAHA.110.588954. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Whiteley W, Chong WL, Sengupta A, Sandercock P. Blood Markers for the Prognosis of Ischemic Stroke: A Systematic Review. Stroke. 2009;40:e380–e389. doi: 10.1161/STROKEAHA.108.528752. [DOI] [PubMed] [Google Scholar]
- 14.Segal HC, Burgess AI, Poole DL, Mehta Z, Silver LE, Rothwell PM. Population-based study of blood biomarkers in prediction of subacute recurrent stroke. Stroke. 2014;45:2912–2917. doi: 10.1161/STROKEAHA.114.005592. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Wannamethee SG, Whincup PH, Shaper AG, Rumley A, Lennon L, Lowe G. Circulating inflammatory and hemostatic biomarkers are associated with risk of myocardial infarction and coronary death, but not angina pectoris, in older men. J Thromb Haemost. 2009;7:1605–1611. doi: 10.1111/j.1538-7836.2009.03574.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Sattar N, Murray HM, Welsh P, Blauw GJ, Buckley BM, Cobbe S, et al. Are Markers of Inflammation More Strongly Associated with Risk for Fatal Than for Nonfatal Vascular Events? PLoS Med. 2009;6:e1000099. doi: 10.1371/journal.pmed.1000099. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Schnabel RB, Yin X, Larson MG, Yamamoto JF, Fontes JD, Kathiresan S, et al. Multiple inflammatory biomarkers in relation to cardiovascular events and mortality in the community. Arterioscler Thromb Vasc Biol. 2013;33:1728–1733. doi: 10.1161/ATVBAHA.112.301174. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Elkind MS, Tai W, Coates K, Paik MC, Sacco RL. High-sensitivity C-reactive protein, lipoprotein-associated phospholipase A2, and outcome after ischemic stroke. Arch Intern Med. 2006;166:2073–2080. doi: 10.1001/archinte.166.19.2073. [DOI] [PubMed] [Google Scholar]
- 19.Harris TB, Ferrucci L, Tracy RP, Corti MC, Wacholder S, Ettinger WH, et al. Associations of elevated interleukin-6 and C-reactive protein levels with mortality in the elderly. Am J Med. 1999;106:506–512. doi: 10.1016/s0002-9343(99)00066-2. [DOI] [PubMed] [Google Scholar]
- 20.Smith A, Patterson C, Yarnell J, Rumley A, Ben-Shlomo Y, Lowe G. Which Hemostatic Markers Add to the Predictive Value of Conventional Risk Factors for Coronary Heart Disease and Ischemic Stroke?: The Caerphilly Study. Circulation. 2005;112:3080–3087. doi: 10.1161/CIRCULATIONAHA.105.557132. [DOI] [PubMed] [Google Scholar]
- 21.Lorgis L, Zeller M, Dentan G, Sicard P, Buffet P, L'Huillier I, et al. Prognostic value of N-terminal pro-brain natriuretic peptide in elderly people with acute myocardial infarction: prospective observational study. BMJ. 2009;338:b1605–b1605. doi: 10.1136/bmj.b1605. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Freund A, Orjalo AV, Desprez P-Y, Campisi J. Inflammatory networks during cellular senescence: causes and consequences. Trends Mol Med. 2010;16:238–246. doi: 10.1016/j.molmed.2010.03.003. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.García-Berrocoso T, Giralt D, Bustamante A, Etgen T, Jensen JK, Sharma JC, et al. B-type natriuretic peptides and mortality after stroke: A systematic review and meta-analysis. Neurology. 2013;81:1976–1985. doi: 10.1212/01.wnl.0000436937.32410.32. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Rothwell PM, Coull AJ, Giles MF, Howard SC, Silver LE, Bull LM, et al. Change in stroke incidence, mortality, case-fatality, severity, and risk factors in Oxfordshire, UK from 1981 to 2004 (Oxford Vascular Study) Lancet. 2004;363:1925–1933. doi: 10.1016/S0140-6736(04)16405-2. [DOI] [PubMed] [Google Scholar]
- 25.Hatano S. Experience from a multicentre stroke register: a preliminary report. Bulletin of the World Health Organisation. 1976;54:541–553. [PMC free article] [PubMed] [Google Scholar]
- 26.Pencina MJ, D'Agostino RB, D'Agostino RB, Vasan RS. Evaluating the added predictive ability of a new marker: From area under the ROC curve to reclassification and beyond. Statist Med. 2007;27:157–172. doi: 10.1002/sim.2929. [DOI] [PubMed] [Google Scholar]
- 27.Hlatky MA, Greenland P, Arnett DK, Ballantyne CM, Criqui MH, Elkind MS, et al. Criteria for Evaluation of Novel Markers of Cardiovascular Risk: A Scientific Statement From the American Heart Association. Circulation. 2009;119:2408–2416. doi: 10.1161/CIRCULATIONAHA.109.192278. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Wang TJ, Larson MG, Levy D, Leip EP, Benjamin EJ, Wilson PW, et al. Impact of age and sex on plasma natriuretic peptide levels in healthy adults. Am J Cardiol. 2002;90:254–258. doi: 10.1016/s0002-9149(02)02464-5. [DOI] [PubMed] [Google Scholar]
- 29.Daniels LB, Maisel AS. Natriuretic Peptides. J Am Coll Cardiol. 2007;50:2357–2368. doi: 10.1016/j.jacc.2007.09.021. [DOI] [PubMed] [Google Scholar]
- 30.Pelsers M, Hermens WT, Glatz JF. Fatty acid-binding proteins as plasma markers of tissue injury. Clin Chim Acta. 2005;352:15–35. doi: 10.1016/j.cccn.2004.09.001. [DOI] [PubMed] [Google Scholar]
- 31.Kilcullen N, Viswanathan K, Das R, Morrell C, Farrin A, Barth JH, et al. Heart-Type Fatty Acid-Binding Protein Predicts Long-Term Mortality After Acute Coronary Syndrome and Identifies High-Risk Patients Across the Range of Troponin Values. J Am Coll Cardiol. 2007;50:2061–2067. doi: 10.1016/j.jacc.2007.08.021. [DOI] [PubMed] [Google Scholar]
- 32.Matsumoto S, Nakatani D, Sakata Y, Suna S, Shimizu M, Usami M, et al. Elevated serum heart-type fatty acid-binding protein in the convalescent stage predicts long-term outcome in patients surviving acute myocardial infarction. Circ J. 2013;77:1026–1032. doi: 10.1253/circj.cj-12-0999. [DOI] [PubMed] [Google Scholar]
- 33.Tölle A, Suhail S, Jung M, Jung K, Stephan C. Fatty acid binding proteins (FABPs) in prostate, bladder and kidney cancer cell lines and the use of IL-FABP as survival predictor in patients with renal cell carcinoma. BMC Cancer. 2011;11:302. doi: 10.1186/1471-2407-11-302. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Ohmachi T, Inoue H, Mimori K, Tanaka F, Sasaki A, Kanda T, et al. Fatty Acid Binding Protein 6 Is Overexpressed in Colorectal Cancer. Clin Cancer Res. 2006;12:5090–5095. doi: 10.1158/1078-0432.CCR-05-2045. [DOI] [PubMed] [Google Scholar]
- 35.Hashimoto T, Kusakabe T, Sugino T, Fukuda T, Watanabe K, Sato Y, et al. Expression of Heart-Type Fatty Acid-Binding Protein in Human Gastric Carcinoma and Its Association with Tumor Aggressiveness, Metastasis and Poor Prognosis. Pathobiology. 2004;71:267–273. doi: 10.1159/000080061. [DOI] [PubMed] [Google Scholar]
- 36.Valgimigli M, Ceconi C, Malagutti P. Tumor Necrosis Factor- Receptor 1 Is a Major Predictor of Mortality and New-Onset Heart Failure in Patients With Acute Myocardial Infarction: The Cytokine-Activation and Long-Term Prognosis in Myocardial Infarction (C-ALPHA) Study. Circulation. 2005;111:863–870. doi: 10.1161/01.CIR.0000155614.35441.69. [DOI] [PubMed] [Google Scholar]
- 37.McElhaney JE, Effros RB. Immunosenescence: what does it mean to health outcomes in older adults? Curr Opin Immunol. 2009;21:418–424. doi: 10.1016/j.coi.2009.05.023. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Rothwell PM, Giles MF, Chandratheva A, Marquardt L, Geraghty O, Redgrave JN, et al. Effect of urgent treatment of transient ischemic attack and minor stroke on early recurrent stroke (EXPRESS study): a prospective population-based sequential comparison. Lancet. 2007;370:1432–1442. doi: 10.1016/S0140-6736(07)61448-2. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.