Objective To assess the association between baseline levels of highly sensitive cardiac troponin T (hs-cTnT) and long-term mortality in perimenopausal women of the general community using a gender specific 99th percentile reference limit.
Design Nested case control.
Setting The present study was conducted within the Eindhoven Perimenopausal Osteoporosis Study which is a large prospective cohort of 8503 community-derived women of the city of Eindhoven, The Netherlands.
Participants Cases were defined as Eindhoven Perimenopausal Osteoporosis Study participants who provided an adequate baseline blood sample and subsequently experienced death during follow-up between 1994 and 2003. In total, 123 cases were identified. For each case two matched controls were selected using age, body mass index and hypertension as matching factors. The gender specific 99th percentile reference limit determined in the 246 controls was 8.0 ng/l.
Main outcome measure All cause mortality.
Results Hs-cTnT was significantly higher in the cases: 3.0 ng/l versus 2.3 ng/l (p=0.04). After adjustment for matching and clinical risk factors, each 1 SD increase of the level of hs-cTnT was significantly associated with mortality (OR 1.3, 95% CI 1.1 to 1.7, p=0.018). With amino-terminal pro-B-type natriuretic peptide (NT-proBNP) in the multivariable model as a continuous variable the association of hs-cTnT with mortality was lost. With both hs-cTnT and NT-proBNP as dichotomous variables, the gender specific 99th percentile reference limit (8.0 ng/l) was associated with mortality, independent of NT-proBNP (OR 3.7, 95% CI 1.0 to 13.2, p=0.048).
Conclusions In this study of community-derived perimenopausal women, hs-cTnT was associated with long-term mortality, independent of clinical risk factors. With the use of easily applicable cut-off levels, the gender specific reference limit of hs-cTnT had a prognostic impact that was independent of NT-proBNP.
Statistics from Altmetric.com
Identification of individuals at risk of future fatal or non-fatal events has been the focus of many studies investigating biomarkers of inflammation1–4 and natriuretic peptides.5–7 In community studies, there is accumulating evidence that assessment of cardiac troponin holds promise as a predictor of all cause mortality and non-fatal cardiovascular events, independent of classical risk factors.
Studies that applied conventional cardiac troponin assays in subjects of the general population reported two important findings. First, the small proportion of subjects in whom troponin could be detected had baseline characteristics correlating with a high cardiovascular risk profile.8 ,9 Second, follow-up studies showed that a detectable troponin was associated with a higher risk of a first coronary heart disease event10 ,11 and all cause mortality.10 ,12
Recently, clinical data of patients with stable angina,13–16 diabetes,17 and of subjects derived from the general population18–20 showed that highly sensitive cardiac troponin (hs-cTnT) was related to the extent of coronary artery disease14 ,15 and an unfavourable prognosis.13 ,16–20
With the introduction of highly sensitive assays, troponin can be detected in a fair amount of healthy subjects, across a wide range of concentrations that are below the detection limit of the former conventional assays. This prompted the assessment of the biological variation of these new assays.21 Subsequently, it was found that in apparently healthy subjects the median concentration of hs-cTnT was lower in women than in men, as were the 99th percentiles.13 ,22–25
Despite this gender difference, information on women is merely derived from subgroup analyses of community studies with mixed populations of men and women.18 ,20 As of to date, there is only one report on an entirely female cohort, health professionals from the USA.26 This study described an independent association between hs-cTnT and cardiovascular mortality in individuals with diabetes, but this could not be demonstrated among the women without diabetes. Notably, the limited data on women never addressed the issue of gender specific reference values.
In view of the above, we performed a nested case-control study in a prospective cohort of perimenopausal women from a European community to study the association between hs-cTnT and all cause mortality during long-term follow-up using a gender specific 99th percentile reference limit.
We conducted a nested case-control study within the Eindhoven Perimenopausal Osteoporosis Study (EPOS). The initial studies from this cohort focused on bone mineral density and thyroid disease.27
EPOS is a large community-based prospective cohort study of perimenopausal women, born between 1941 and 1947, inhabitants of the city of Eindhoven, The Netherlands. Inclusion occurred between September 1994 and August 1995.
A total of 8503 women received a health questionnaire to assess incident disease. In case the questionnaire was not returned individuals received a second and if necessary a third invitation. The overall response rate was 81% (n=6846), and 6648 women agreed to participate in the cohort. They were invited for a baseline medical examination, and were asked to provide a blood sample at the primary care laboratory of the Diagnostic Centre Eindhoven. Blood samples were collected in tubes and immediately centrifuged and stored in aliquots at −80°C until further analysis.
In total, 6311 individuals provided a baseline blood sample and 6190 also signed the informed consent for follow-up studies. Survival status was checked in 1998 and 2003 with the Central Bureau of Statistics (CBS), The Netherlands. Of all individuals, 2% was reported dead. During follow-up 8.6% moved to another city and 1.5% emigrated. Follow-up questionnaires for additional clinical information were sent in 1998 and 2003, only to participants that were still inhabitants of Eindhoven.
End-point confirmation and selection of cases and controls
The study end-point was all cause mortality. Cases were defined as EPOS participants who provided an adequate baseline blood sample and subsequently experienced death during a follow-up period of up to 9 years (follow-up between 1994/1995 and 2003). In total, there were 126 cases. For each case with a CBS confirmed mortality status two controls were selected from the present study cohort, matched for age (within 1 year), body mass index (within one unit) and hypertension (presence or absence). At the time that a subject became a case, the control subjects had to be alive—as confirmed by the CBS—and have at least the same duration of follow-up. At the time of biomarker analyses, it appeared that in three of the cases sample volume was insufficient to perform biomarker measurements. Therefore, data-analysis refers to 123 cases and 246 controls.
Measurement of biochemical variables
All testing was done on the Modular E170 (Roche Diagnostics) at the Máxima Medisch Centrum, Veldhoven, The Netherlands. Assays were performed by staff members unaware of the participants’ identities and whether they were cases or controls.
Highly sensitive assay for cardiac troponin T: At the time of laboratory analysis the highly sensitive assay that was used for assessment of troponin T was precommercial, and had only been validated in the research department of Roche Diagnostics (research lot number: DR-lot). The lower limit of detection of hs-cTnT at that time was 1 ng/l.13 The reported 99th percentile reference limits as assessed in healthy individuals and blood donors are 13.4 ng/l and 13.3 ng/l, respectively. At this concentration the interassay coefficient of variation is <10%. In men, the 99th percentile reference limit is 14.5 ng/l, and in women 10.0 ng/l. The gender specific 99th percentile reference limit as assessed in control subjects (n=246) of the present study population is 8.0 ng/l (coefficient of variation 8.9%).
Conventional assay for cardiac troponin T: The conventional assay used was the commercially available third generation immunoassay, with a 99th percentile reference limit of 10 ng/l. In contrast to the highly sensitive assay, cardiac troponin T is not detectable below the 99th percentile with the conventional assay. Inherently, the lower limit of detection is 10 ng/l. At a troponin concentration of 30 ng/l, the coefficient of variation is ≤10%.
With regard to other analyses, total cholesterol and creatinine were measured using spectrophotometric assays. Creatinine clearance was estimated with the Cockcroft-Gault formula. Amino-terminal pro-B-type natriuretic peptide (NT-proBNP) was measured with a sandwich chemiluminescent immunoassay. The measuring range is 5–35.000 pg/ml. The reported coefficient of variation is 3.2% at 175 pg/ml and 2.3% at 4962 pg/ml. The 80th percentile as determined in controls was 112 pg/ml.
Means and proportions of the different risk factors for cardiovascular disease at baseline were calculated for cases and controls. For continuous variables, the significance of differences between group means was assessed with Student's t test or Wilcoxon rank sum test depending on the normality of their distributions. For skewed variables log transformation was performed, whenever appropriate. In the case of dichotomous variables, the significance of differences between proportions was tested using the χ2 statistic or Fisher's Exact test whenever appropriate.
The association between hs-cTnT and all cause mortality was analysed in two ways using logistic regression. Hs-cTnT was entered as a continuous variable and its association with mortality was assessed per 1 SD increase in troponin concentration. Second, hs-cTnT was analysed as a dichotomous variable, using the gender specific 99th percentile reference limit (8.0 ng/l) as determined in controls as cut-off. Risk estimates were obtained for both analytical approaches, with forward conditional logistic regression models adjusting for matching variables (age, body mass index, hypertension) (model 1); then, additional cardiovascular risk factors (diabetes, smoking status, cholesterol, prior MI/stroke) and glomerular filtration rate were added to model 1 (model 2); and finally the log transformed NT-proBNP concentration was added to the factors of model 2 (model 3). In the case troponin was analysed as a continuous variable, NT-proBNP was added as a continuous variable. When troponin was analysed using a clinically applicable cut-off (ie, 8.0 ng/l, the gender specific 99th percentile reference limit), NT-proBNP was entered as a dichotomous variable as well. In analogy to previous reports, we will report associations based upon the 80th percentile of NT-proBNP for reasons of comparibility.5 ,6 All p values are two-tailed, and values of less than 0.05 were considered to indicate statistical significance. All CI were calculated at the 95th per cent level. Data was analysed with SPSS V.16.
Baseline characteristics are outlined in table 1, including the matching variables. The presence of previous cardiovascular disease was limited to one myocardial infarction and one stroke among the cases and three strokes among the controls.
Serum biomarker levels
In none of the subjects cardiac troponin could be detected with the conventional third generation troponin T assay. In contrast, hs-cTnT exceeded the lower limit of detection in 86% (317/369) of the women.
The mean concentration of hs-cTnT was 3.0 ng/l in cases and 2.3 ng/l in controls (p=0.04). The distribution of hs-cTnT is demonstrated in figure 1 for both cases and controls. In total, 12 subjects had a hs-cTnT concentration ≥8.0 ng/l, that is, the gender specific 99th percentile reference limit: 67% (8/12) were cases. The observed ratio of cases to controls was 8:4 instead of the expected ratio of 1:2 (unadjusted OR 4.2, 95% CI 1.1 to 17.0, p=0.02).
Importantly, of these 12 subjects with a hs-cTnT ≥ the 99th gender specific reference limit, only three had a concentration exceeding 14 ng/l. In case of a hs-cTnT concentration >8.0 ng/l (n=12), 10 also had a NT-proBNP ≥ the 80th percentile. Of these individuals 70% (7/10) were cases.
The mean concentrations of NT-proBNP were 138 pg/ml and 83 pg/ml for cases and controls, respectively (p=0.02). In total, for all individuals with a NT-proBNP value ≥ the 80th percentile, the observed ratio of cases to controls was 38:49 instead of the expected ratio of 1:2 (unadjusted OR 1.8, 95% CI 1.1 to 3.0, p=0.02).
Hs-cTnT and all cause mortality
A stepwise 1 SD increase of hs-cTnT was associated with an OR of 1.4 (95% CI 1.1 to 1.7, p=0.008) for all cause mortality after adjustment for matching variables (model 1, table 2). After additional adjustment (model 2) the association between mortality and hs-cTnT remained significant, independent of classical risk factors like diabetes, smoking status, cholesterol, prior myocardial infarction/stroke and renal function. Adjustment for NT-proBNP as a continuous variable (model 3) attenuated the association to a point it was no longer significant. After multivariable adjustment only NT-proBNP (OR for natural logarithm of NT-proBNP 1.4, 95% CI 1.1 to 1.8, p=0.009) remained significantly associated with mortality.
In the second analysis we studied troponin and NT-proBNP as dichotomous variables. For hs-cTnT the prespecified gender specific 99th percentile reference limit was used as clinical cut-off (table 3), for NT-proBNP the 80th percentile, in analogy to previous reports.5 ,6 There was a significant association between hs-cTnT and mortality in all three multivariable models, also after adjustment for NT-proBNP (OR 3.7, 95% CI 1.0 to 13.2, p=0.048).
To our knowledge, this study in a cohort of perimenopausal women from a European community is the first report on an entirely female study population that demonstrates an association between hs-cTnT and long-term all cause mortality, independent of clinical risk factors. We believe that our study adds to the only other study on a female cohort, the Women's Health Study, where this association was observed for diabetics, but not for the individuals without diabetes.26 Moreover, that study population was not a representative sample of the community, as only female healthcare professionals were studied.
We observed an association with mortality that was independent of clinical risk factors, both for hs-cTnT as a continuous variable, and as a binary variable, which is more easily applicable in clinical practice to perform risk stratification. Given the gender differences in normal values of hs-cTnT,13 ,22–25 we used a prespecified gender specific 99th percentile reference limit. This methodology merits further attention, as it may carry important consequences with regard to the proportion of subjects identified as high risk. This could contribute to further refinement of individualised risk stratification in a low risk population from the community, in general, and in women in particular.
In analogy to previous reports on mixed populations of male and female individuals,11–13 18–20 we observed an intriguing interplay between troponin, NT-proBNP and long-term outcome in this population of only women.
Conventional troponin assays
In our study, troponin was not detectable with the conventional assay. Studies that used conventional assays showed that the few subjects in whom troponin was detectable had a higher cardiovascular risk profile.8 ,9 Factors like hypertension were more prevalent in individuals with a detectable troponin. Notably, follow-up studies showed that a detectable troponin was also associated with a higher risk of a first coronary heart disease event10 ,11 and all cause mortality.10 ,12
Highly sensitive troponin assays
With the introduction of highly sensitive assays, an entirely new group of individuals, in whom troponin was formerly undetectable, can now be identified. In the present study population, troponin could be assessed in about 90% of women. Inherently, this development provides improved conditions for risk stratification. In patients with stable angina13 and diabetes,17 hs-cTnT has been associated with the population's cardiovascular risk profile. Other studies have shown that hs-cTnT is associated with the extent of coronary artery disease in patients with stable angina.15 Similar findings have been observed even in lower risk individuals with suspected coronary artery disease.14
Recently, the first studies have been published that demonstrate an independent association between hs-cTnT and mortality in the community, with study populations that consisted of a heterogeneous group of men and women.18 ,20 Observations in these studies suggested that the findings apply to the subset of women as well. Only one of these reports demonstrated that in the female subset of the cohort hs-cTnT had a prognostic impact independent of clinical risk factors.20 Yet, in the Women's Health Study, the first report on a completely female study cohort, this could only be confirmed for subjects with diabetes, but not for the case-cohort of non-diabetic women.26 To date, most of the evidence for the prognostic impact of hs-cTnT in women is derived from subgroup analyses of American community studies.18 ,20
In our population of middle aged women from a European community, hs-cTnT had prognostic impact, independent of clinical risk factors. This was observed for troponin as a continuous variable, and also as a binary variable. The latter is more easily applicable in clinical practice, and, importantly, we used the gender specific 99th percentile (8.0 ng/l) reference limit, a different approach than in previous analyses.
Population dependent reference values
With the introduction of highly sensitive assays, additional aspects have to be appreciated. First, the issue with regard to gender specific reference values deserves further study. It has been demonstrated that in apparently healthy populations the median concentration and 99th percentile of hs-cTnT is lower in women than in men.13 ,22–24 In a majority of the studies on the potential prognostic impact of hs-cTnT, analyses have been performed with reference values that were independent of gender. With this methodology, some women may have been classified as low-risk individuals based on a troponin result with a concentration that was lower than the non-gender specific cut-off level. In fact, with the use of a gender specific reference level, this concentration could have exceeded the cut-off level for women, that is, the subject would have been identified as a high-risk individual. In our study, 12 women had a hs-cTnT concentration >8 ng/l. With the use of the non-gender specific reference level (14 ng/l), only three women would be considered to have a concentration exceeding the cut-off.
Second, it should be noted that determination of reference intervals in the community will to a certain degree depend on the cardiovascular risk profile. A population of higher age, and predominantly males, with many cardiovascular risk factors, will have a different reference interval than a population of predominantly younger females with less cardiovascular risk factors.24 In our study, the 99th percentile in the controls was 8.0 ng/l, whereas the reported value by the manufacturer was 13.4 ng/l in a mixed group of healthy males and females. Inherently, the introduction of a highly sensitive assay has fuelled the discussion on ‘normal’ values, since the number of true and false positive results will increase. In addition, a recent publication on sources of variation further highlighted this topic.21 In view of this issue of variation, the apparent differences between men and women, and the impact of the cardiovascular risk profile of the population described, further studies are warranted to address the definition of ‘normal’ values.25
Troponin, NT-proBNP and prognosis
As binary variables, the association between hs-cTnT and all cause mortality was found to be independent of NT-proBNP. With the use of clinically applicable cut-off values, both biomarkers have prognostic impact, independent of clinical risk factors. After a recent recalibration of the assay, the lower limit of detection is now set at 5 ng/l.28 Therefore, especially in healthy populations, the preferred way to interpret results is to consider hs-cTnT as a dichotomous variable. However, with NT-proBNP as a continuous variable, NT-proBNP remained independently associated with mortality, but troponin lost its association. This interplay between troponin and NT-proBNP has been described in previous reports on mixed study populations of men and women. In most other studies, the association between hs-cTnT and cardiovascular events was attenuated, but remained present.13 ,18–20 These observations resulted in different interpretations. One view is that the biomarkers are non-interdependent and represent distinct pathophysiological roles, which would explain their independent predictive values. A second view is that these biomarkers represent a common pathology such as myocardial stretch and dysfunction. In line with this, hs-cTnT has recently been associated with the future risk of heart failure as well.19 A potential pathophysiological mechanism could be that repetitive episodes of microinjury result in the release of hs-cTnT, and that with increasing amounts of injury myocardial dysfunction is induced, which could be reflected through elevation of NT-proBNP. Consequently, one could argue that patients with a detectable hs-cTnT, but without increased NT-proBNP levels might have a more favourable prognosis, than those with increased levels of NT-proBNP. This hypothesis remains to be tested in a larger scale study population.
With regard to inferences to the general population, the current study population refers to the late nineties, an era in which the cardiovascular risk profile of women may have been somewhat different than nowadays (smoking habits, hypertension, obesity). The 99th percentile in our controls was 8.0 ng/l versus a reported reference limit for women by the manufacturer of 10 ng/l. Potential explanations are differences in the cardiovascular risk profiles of the individuals tested, or, possibly, degradation of troponin despite conservation at –80°C.21 On the other hand, 8.0 ng/l has been reported before as 99th percentile for females.23
Given the low risk population with a rather limited number of events, we only address all cause mortality, without subgroup analysis for cardiovascular mortality, which accounted for 20% of cases. In previous, larger studies on this subject, addressing mixed populations of men and women, the cause of death was cardiovascular in 10–40% of cases.18 ,20 Analogous to our findings, these studies also reported an association between hs-cTnT and all cause mortality. Whether hs-cTnT is associated with non-cardiovascular death remains a topic of future study.
Conclusion and implications
The present findings in a well described unique population of middle aged women from a European community suggest that hs-cTnT may prove to have additional prognostic value as opposed to risk stratification based on clinical risk factors only. Moreover, given its association with NT-proBNP, the development of heart failure and mortality, further studies on troponin in the community may help us to detect and unravel different (stages of) disease processes in asymptomatic subjects.
In view of the discussed gender difference, the impact of the cardiovascular risk profile and other sources of variation on reference values for hs-cTnT, our findings call for additional community studies that specifically address these issues to further study the prognostic potential of hs-cTnT, and how the application of the different reference values may affect future risk prediction.
We would like to thank Colette Wijnands for assistance with database management. In addition, we acknowledge Jeroen Jaspers Focks and Frank Gommans for critically revising the manuscript.
Contributors GEC, MAB, VJMP and FWAV conceived and designed the study. GEC, MAB and HLV acquired the data. GEC, MAB, GAP, VJMP and MJB analysed and interpreted the data. GEC and MAB drafted the initial manuscript. HLV, GAP, MJB, VJMP and FWAV made critical revisions of the manuscript for important intellectual content. All authors approved the final version. GEC and MAB contributed equally to this work. FWAV is guarantor.
Funding This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors. Reagents to determine highly sensitive cardiac troponin T and amino terminal pro-B-type were provided by Roche Diagnostics.
Competing interests None.
Ethics approval Not required for the present nested case-control study which was conducted within the Eindhoven Perimenopausal Osteoporosis Study (EPOS). Previously, medical ethical approval was provided for EPOS by the Medical Ethical Review Board of the Máxima Medical Centre, Veldhoven, The Netherlands. All participants in the present case-control study provided written informed consent at the beginning of EPOS to perform follow-up studies.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.