Objectives To define whether the high sensitivity cardiac troponin T (hs-cTnT) assay in patients with immunoglobulin light chain amyloidosis (AL) improves risk prediction.
Background Cardiac involvement is the major cause of death in patients with AL amyloidosis. Risk stratification is facilitated by cardiac biomarkers such as cardiac troponin T (cTnT) and N-terminal pro B-type natriuretic peptide (NT-proBNP).
Methods Stored serum from patients with newly diagnosed AL was used to measure hs-cTnT, cTnT, and NT-proBNP. Survival modelling was performed.
Results The direct numeric result from hs-cTnT measurement cannot merely be substituted for a cTnT measurement in the Mayo AL staging system. The performance of the receiver operator curve derived an hs-cTnT cut-point of 54 ng/L which improves on the value of 35 ng/L validated with the prior iteration of the assay. An alternate staging option using hs-cTnT alone—using the two thresholds 14 ng/L and 54 ng/L—performs as well as either the original Mayo AL staging system or other systems incorporating hs-cTnT. On multivariate analysis, an hs-cTnT alone staging system was independent of period of diagnosis, type of therapy, and NT-proBNP value, the last of which dropped out of the model. Alternate models were explored, but none performed better than the original system or the new hs-cTnT system. Thus, hs-cTnT can be used alone for the staging of disease prognosis.
Conclusions A survival model based on hs-cTnT improves the prognostic staging of patients with AL amyloidosis, relegating NT-proBNP to a measure of cardiac response.
Statistics from Altmetric.com
Cardiac involvement is the major cause of death in patients with immunoglobulin light chain (AL) amyloidosis. Detection of cardiac involvement and risk stratification has been facilitated by soluble cardiac biomarkers such as N-terminal pro B-type natriuretic peptide (NT-proBNP) and cardiac troponin T (cTnT), and there are treatment options available that improve prognosis even when cardiac involvement is present1–,5 Based on our previous work, a threshold of cTnT of 0.035 μg/L (35 ng/L) has been included in the internationally accepted amyloidosis staging system.6 ,7 Recently, a new high sensitive cTnT (hs-cTnT) assay has been developed.8 Assay values correlate closely above a value of 0.1 μg/L (100 ng/L), but below that value they are very different, making extrapolation between the two assays difficult. Thus, it is not clear if the value (0.035 μg/L) previously used for upstaging patients6 ,7 will still provide optimal information with this assay. In addition, the four- to fivefold increase in sensitivity over the fourth generation cTnT assay might permit earlier detection of amyloid heart disease and improve prognostic models still further. Accordingly, we evaluated its performance in 224 patients with AL amyloidosis to determine how to incorporate these values and to make sure that they can be extrapolated into the current AL amyloidosis staging system, and to probe whether the increased sensitivity provided will further improve the predictive accuracy of the staging system.
Approval for the study was obtained from the Mayo Clinic Institutional Review Board, conforming to the Declaration of Helsinki. Sera stored at −20°C from 224 newly diagnosed AL patients who were seen at the Mayo Clinic between 19 June 1989 and 11 July 2006 were used. One hundred and fifty-seven patients had died by the time of this report. Median follow-up for surviving patients was 46.6 months (range 17.8–60 months). The fourth generation cTnT, the hs-cTnT, and NT-proBNP assays were run on the E170 Modular analyser (Roche Diagnostics, Penzberg, Germany). The 99th centile upper reference limit is >14 ng/L for hs-cTnT with a limit of detection of 5 ng/L and a 10% coefficient of variation (CV) value of 13 ng/L. Comparable metrics for the fourth generation assay are a 99th% value at the limit of detection of the assay of <0.01 μg/L and a 10% CV value of >0.035 μg/L.8 ,9
All statistical analyses were performed using JMP software (SAS, Carey, North Carolina, USA). Receiver operator curve (ROC) estimates for thresholds that best predict for 100 day mortality were generated and used to test different staging models. Previously reported cTnT,6 ,7 hs-cTnT,10 ,11 and NT-proBNP6 ,7 cut points were explored. Fisher's exact test and Kruskal-Wallace test were used to determine differences between nominal and continuous variables. Correlation coefficients were estimated using Spearmann rank. Survival was calculated according the method of Kaplan and Meier. Differences in survival were estimated according to the log-rank statistic.
Two hundred and twenty-four patients were studied. The median (range) values for hs-cTnT, cTnT, and NT-proBNP were 38 ng/L (0–754), 0.017 μg/L (0.005–0.90), and 1230 ng/L (0–32 226), respectively. Figure 1A demonstrates the correlations between the troponin assays at various cut-off values and between the cTnT assays and NT-proBNP. The correlation between cTnT and hs-cTnT was excellent (R=0.97, figure 1A), but deteriorated at hs-cTnT values under 77 ng/L (figure 1B). If the hs-cTnT results were substituted for the cTnT result in the Mayo amyloid staging system, which relies on a cTnT threshold of 0.035 μg/L and an NT-proBNP threshold of 332 ng/L, 31 patients (14%) would be upstaged, given the lack of concordance at these low levels: three patients upstaged from I to II; and 28 upstaged from II to III. No patient would be down-staged.
To address the discordances observed at these levels, ROC analyses were employed to define alternate optimal cut-points. Using 100 day mortality as the end point, the respective optimal cut-off values were as follows: cTnT, 0.047 μg/L; hs-cTnT, 54 ng/L; and NT-proBNP, 2279 ng/L. Patient risk of death is shown graphically at these cut-points in figure 2A–C. The relative risk (RR) of death were: hs-cTnT RR 2.6, 95% CI 1.9 to 3.5; cTnT RR 3.0, 95% CI 2.2 to 4.2; and NT-proBNP RR 2.4, 95% CI 1.7 to 3.3.
Singly and as a log-transformed continuous variable, hs-cTnT was a significant risk factor for death (RR 2.94, 95% CI 2.06 to 4.2). Analyses were done to determine how the hs-cTnT could perform alone. The breakpoints chosen were the >99th centile reference limit (14 ng/L) and the ROC derived threshold of 54 ng/L. As shown in table 1, there were substantial clinical differences between these three groups of patients. With increasing hs-cTnT, there was significantly increased patient age, higher soluble cardiac biomarkers (cTnT, cardiac troponin I (cTnI), and NT-proBNP), worse left ventricular EF, increased left ventricular mass, worse deceleration time, higher serum creatinine, uric acid and β-2 microglobulin values, less likelihood of receiving high dose chemotherapy with peripheral blood stem cell transplant as therapy, and a later period of diagnosis—2004–2006. This model resulted in significantly different median overall survival rates of 71, 43, and 6 months, for the low, intermediate and high values, respectively (figure 3A), which remained significant on multivariate analysis (table 2)—including the time period of diagnosis and whether or not patients had autologous stem cell transplantation (ASCT). In this model, NT-proBNP no longer significantly added to the predictive accuracy of the model.
To determine whether the currently accepted Mayo staging system using NT-proBNP and cTnT could be enhanced, and how best to incorporate hs-cTnT, additional modelling was performed. Each of the models shown in figure 4 produces relative risks of death between 1.8 and 2.0 per level. The previously published cTnT/NT-proBNP model (figure 4A) performs as well as the other models which were generated, incorporating different cut-points based on ROC analyses using either hs-cTnT (figure 4B,D) or cTnT (figures 4C,E). Figure 4D is an example of combining the ROC derived hs-cTnT threshold of 54 ng/L and the ROC derived NT-proBNP threshold of 2279 ng/L. Figure 4B is an example of combining the ROC derived hs-cTnT threshold of 54 ng/L and the previously reported NT-proBNP threshold of 332 ng/L. Although those models incorporating the ROC derived NT-proBNP cut-point of 2279 ng/L were the least well balanced in terms of patients per group, with a disproportionate number of patients falling into stage I (figure 4D,E), their risk ratios were comparable to the other models—that is, NT-proBNP no longer added to the prediction of mortality.
Figure 5A,E illustrate the relative distribution of patients into risk categories based on various combinations of cTnT (or hs-cTnT) and NT-proBNP thresholds, focusing on the cut-offs defined in the original staging system6 ,7 and in the ROC analyses. By combining the ROC analysis derived hs-cTnT threshold of 54 ng/L with the Mayo standard NT-proBNP of 332 ng/L, there was greater patient overlap with the original staging system than there was using an hs-cTnT value of 35 ng/L (figure 5A,B). The best patient overlap, however, was seen when the Mayo standard NT-proBNP 332 ng/L was combined with ROC derived values of cTnT and hs-cTnT (figure 5D); only 20 (9%) patients had discordant stage placement. Most12 of the discordance was between stages II and III. By using the ROC derived NT-proBNP value of 2279 ng/L combined with a cTnT of 0.035 μg/L or an hs-cTnT of 54 ng/L (figure 5C), there were only 10 discordant patients (5%) with discordant placement between stages II and III.
The performance of the optimally determined hs-cTnT threshold was compared with those from the published literature,6 ,7 including the original Mayo cTnT 0.035 μg/L and the recently published Pavia value of hs-cTnT 77 ng/L by Cox modelling.10 By univariate analysis, the following RRs (95% CIs) for death were found: cTnT 0.035 μg/L, 3.2 (2.3 to 4.3); hs-cTnT 35 ng/L, 2.6 (1.9 to 3.6); cTnT, 0.054 μg/L, 3.1 (2.2 to 4.3); hs-cTnT, 54 ng/L, 2.6 (1.9 to 3.5); cTnT 0.077 μg/L, 3.3 (2.2 to 4.7); and hs-cTnT 77 ng/L, 2.5 (1.7 to 3.5). When cTnT and hs-cTnT were entered into the Cox model as dichotomous variables simultaneously, there was a trend for inferior performance with hs-cTnT compared to the cTnT: hs-cTnT 35 ng/L, 1.5 (0.9 to 2.4) and cTnT 0.035 μg/L, 2.4 (1.4 to 4.0); hs-cTnT 54 ng/L, 1.5 (0.9 to 2.5) and cTnT 0.054 μg/L, 2.1 (1.3 to 3.8); hs-cTnT 77 ng/L, 1.1 (0.6 to 2.1) and cTnT 0.077 μg/L, 2.9 (1.5 to 5.8).
These data provide important new insights into the use of soluble cardiac biomarkers, particularly hs-cTnT, for risk stratification in patients with AL light chain amyloidosis. They suggest that the use of the hs-cTnT assay will simplify prognostication by eliminating the role of the natriuretic peptide NT-proBNP. Such a model would use two thresholds, one at 14 ng/L and one at 54 ng/L, and would predict for more than a threefold risk of death, independent of other factors including NT-proBNP, diagnosis before or after 2004, and type of therapy. Such an approach would simplify the algorithm without diminishing its prognostic value and avoid the difficulties of NT-proBNP, which like all natriuretic peptides manifests substantial biological variation such that it requires pronounced changes to be sure that meaningful changes in values have occurred.13
Surprisingly, exploiting the lowest values of the more sensitive hs-cTnT assay did not further improve the model, either with or without inclusion of NT-proBNP. It may be that other causes of mortality occur before cardiac disease becomes manifest. Alternatively, our cohort, which is large for a group of patients with amyloidosis, may still be underpowered to detect differences with only modest elevations in these patients, since in large population studies—albeit populations without amyloidosis—these values are notably prognostic.14–,17 It also should be clear, as suggested from the assay validation studies, that the lower values used for prognostic evaluation are sufficiently different between the hs-cTnT assay and the standard cTnT to preclude mere substitution of one test result for another. In order to continue to use the current Mayo staging system with hs-cTnT results, substituting a value of 54 ng/L for the previous value of 0.035 μg/L appears optimal.
The use of soluble cardiac biomarkers has been a major advance in the field of AL amyloidosis in terms of stratifying patients’ risk of early death, treatment related mortality, long term survival,6 ,7 ,10–,12 ,18–,24 and the response to therapy.3 ,10 ,22 ,25 Our first study used troponin as the sole soluble cardiac biomarker12 and then later the combination of troponin (both T and I) and NT-proBNP was incorporated.6 ,7 This system has been substantiated by others,3 ,21 and others have also used combinations of BNP.26 ,27 Although these soluble cardiac biomarkers allow for better and/or more objective standardisation than parameters such as performance status and echocardiographic measurements, a major challenge has been access to these tests in the laboratories at different institutions and the evolution of these assays. The largest market share in the USA of troponin assays is cTnI, which in our system has never performed as well as cTnT.12 These data are similar to a recent evaluation by the group from Pavia with the hs-cTnT assay.10 Much of Europe uses cTnT, which is now being replaced with hs-cTnT. It is only a matter of time before this change occurs in the USA as well. In addition, many cTn assays are being released that manifest better sensitivity than their predecessors. These assays, as with cTnT, may not be totally harmonised to the assays that preceded them, and thus will require revalidation to define the optimal values for use in patients with amyloidosis.
Having a widely accepted and available uniform stratification system is essential, as already illustrated by the adaptation of the Mayo system in at least three ongoing treatment clinical trials (NCT01277016; NCT01318902; NCT01078454), and the simplicity of such a system is paramount. The Pavia amyloid group10 has explored different thresholds of hs-TnT and NT-proBNP, most notably an NT-proBNP of 332 ng/L and 2736 ng/L, and an hs-TnT of 14 ng/L and 77 ng/L. Their respective HRs (95% CI) for the 171 patients studied were 4.1 (1.6 to 10.1), 2.6 (1.6 to 4.3), 2.7 (1.4 to 5.3), and 4.3 (2.6 to 7.2), which are similar to our results. Kristen et al11 also evaluated the performance of hs-cTnT in a recent study of 163 patients. They found that hs-cTnT was measurable, but not necessarily elevated, in 86% of amyloid patients with cTnT <0.01 μg/L. However, the ability to quantitate these previously unmeasurable levels did not translate into different outcomes either in isolation or in combination with NT-proBNP in their patient population. This appears to be the case in our series as well (data not shown). They too found that an hs-cTnT value of 50 ng/L—as compared to our 54 ng/L—was a valuable prognostic breakpoint, but in their population, the original Mayo staging system did not demonstrate separation between stage I and II patients after a mean follow-up of 22.3 months. Surprisingly, survival in stage II patients worsened when patients from stage I were moved into stage II, based on the hs-cTnT results using a cut-point of 50 ng/L hs-cTnT. Although the reason for this paradoxical effect is unclear, their data support our calculated threshold value for hs-cTnT of 54 ng/L, which corresponds reasonably well to a cTnT of 0.035 μg/L in our dataset. Of note, the ROC derived hs-cTnT threshold derived from Mayo Clinic patients is intermediate between that of the Heidelberg group (50 ng/L) and that of the Pavia group (77 ng/L), demonstrating that the ‘best threshold’ may be dependent on the population studied and the end points selected.
In summary, the use of hs-cTnT will simplify algorithms for prognostication in patients with AL amyloidosis, but will not provide additional prognostic information despite its increased sensitivity. Due to an imperfect correlation between cTnT and hs-cTnT at the lowest values, they cannot be used interchangeably. These data should be confirmed by other groups, and further movement towards consensus criteria should continue, especially with the use of the higher sensitivity cTnI assays that are presently available and being developed.28 –30 In aggregate these works demonstrate the advances being made in the use of cardiac markers for prognostication and therapeutic response, and should translate into improved management strategies for patients with AL.
What is already known about this study?
It is known that elevated values of cardiac troponin T (cTnT) and N-terminal pro B-type natriuretic peptide (NT-proBNP) provide important information concerning the prognosis of patients with AL amyloidosis.
What does the present study add?
The present study incorporates a high sensitivity troponin assay (hs-cTnT) into the algorithm and assesses whether lower cut-offs are necessary and whether hs-cTnT improves prognostic determinations.
How might this impact clinical practice?
The present data indicate that with the hs-cTnT assay, NT-proBNP is no longer necessary to provide robust prognostic information concerning the prognosis of patients with AL amyloidosis.
We would like to thank Tara Phelps for her maintenance of the peripheral blood sample bank, Kate Johanns for maintenance of the clinical database, and Ms Muriel Finkel of the Amyloid Support Group (http://www.amyloidosissupport.com) for her support to our patients.
Presented in part in abstract form at the 52nd annual meeting of the American Society of Hematology, San Diego, California, USA, December 2010.
Contributors All authors contributed to the concept, the revisions, and final approval. AD and ASJ were responsible for the design, AD, AKS, MG, and ASJ participated in the analysis, and AD and ASJ were responsible for the drafting of the manuscript.
Funding Supported in part by CA62242 (AD, RAK, SVR and PRP.) and CA91561 (AD) from the National Cancer Institute, the Robert A Kyle Hematologic Malignancies Fund, The JABBS Foundation, The Predolin Foundation, and Mr Howard Weitzman's 2011 Gala in Support of Amyloid Research and Awareness.
Competing interests ASJ has or presently consults for most of the major diagnostic companies including the company whose assay was used in this report.
Patient consent Obtained.
Ethics approval Mayo Clinic Institutional Review Board.
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.