Independent prognostic value of stroke volume index in patients with light-chain amyloidosis

Abstract

Background

Heart Involvement is the most important prognostic determinant in AL amyloidosis patients. Echocardiography is a cornerstone for the diagnosis and provides important prognostic information.

Methods and Results

We studied 754 patients with AL amyloidosis who underwent echocardiographic assessment including a Doppler derived measurement of stroke-volume (SV) within 30 days of their diagnosis to explore the prognostic role of echocardiographic variables in the context of a well-established soluble cardiac biomarker staging system. Reproducibility of SV, MCF and LV-strain was assessed in a separate, yet comparable, study cohort of 150 patients from the Pavia Amyloidosis Center.

The echocardiographic measures most predictive for overall survival were SV-index (SVI)<33 mL/min, myocardial contraction fraction <34%, and cardiac-index <2.4 L/min/m² with respective hazard ratios [HR] (95% confidence intervals) of 2.95 (2.37, 3.66), 2.36 (1.96, 2.85), and 2.32 (1.91, 2.80). For the subset who had left ventricular strain (LV-strain) performed, the prognostic cut-point was -14% [HR 2.70 (1.84, 3.96)]. Each parameter was independent of systolic blood pressure, Mayo staging system (NT-proBNP and troponin), and ejection fraction on multi-variable analysis. Simple predictive models for survival including biomarker staging along with SVI or LV-strain were generated.

Conclusions

SVI prognostic performance was similar to LV-strain in predicting survival in AL amyloidosis, independently of biomarker staging. Since SVI is routinely calculated and widely available, it could serve as the preferred echocardiographic measure to predict outcomes in AL amyloidosis patients.

Introduction

The systemic amyloidoses are disorders of protein conformation and metabolism that result in tissue deposition of insoluble fibrils, organ dysfunction, and death. In light chain amyloidosis (AL), fibrils are composed mainly by the N-terminus of a monoclonal immunoglobulin light chain¹. AL is a systemic disease, and most patients have clinical involvement of more than one organ at diagnosis. Heart involvement is present in almost two-thirds of patients with this disease, representing the most common cause of death^{2, 3}.

Cardiac amyloidosis typically causes a restrictive cardiomyopathy,⁴ in which the deposition of misfolded amyloid proteins in the myocardium, causes structural alterations whereas cardiotoxic light chains cause myocardial dysfunction^{5, 6}, eventually leading to heart failure. Despite the progression of the disease, ejection fraction (EF) often remains within the “normal” or “preserved” range, and a reduced EF is a sign of late or end stage disease^{7, 8}. More subtle echocardiographic parameters such as increased left ventricular wall thickness and decreased fractional shortening, which have been found to be independent predictors of cardiac mortality in AL amyloidosis^{7, 9}, are often overlooked. Myocardial strain imaging^10-14 and midwall fractional shortening are useful for diagnosis and prognostication of survival among patients with cardiac amyloidosis^15-18, but are not routinely performed. Myocardial strain imaging is limited by technical challenges in acquisition and vendor specific variation in calculated results. The myocardial contraction fraction (MCF), which is defined as the ratio of stroke volume (SV) to myocardial volume (MV), was proposed by King et al. as a novel index of myocardial function,¹⁹ being a volumetric measure of myocardial shortening, which differentiates myocardial performance in patients with pathological versus physiologic hypertrophy. In a recent paper of 34 patients, MCF was superior to EF in predicting overall survival among AL amyloidosis patients²⁰. N-terminal pro-natriuretic type B (NT-proBNP)²¹ and cardiac troponins (cTn)^22-24 are currently the most important markers for the assessment of prognosis in AL amyloidosis and are combined in different staging systems^25-27. Unfortunately, the general cardiologist often overlooks the prognostic implications of these variables and/or they are tests that are ordered only after an index of suspicion for AL amyloidosis has been raised. We, therefore set out to assess the potential prognostic value of widely accessible echocardiographic variables in a larger cohort of AL amyloidosis patients.

Methods

The data, analytic methods, and study materials will not be made available to other researchers for purposes of reproducing the results or replicating the procedure. The study included 754 patients who underwent hematologic assessment, as well as complete echocardiographic evaluation at Mayo Clinic, Rochester, MN, USA, within 30 days of their AL amyloidosis diagnosis. Patients were diagnosed from January 1, 1996 through February 1, 2015. The diagnosis of AL amyloidosis was predicated on the presence of organ involvement as previously defined²⁸ and a tissue biopsy specimen that stained positive with Congo red and exhibited green birefringence under polarized light. The diagnosis of AL amyloidosis was confirmed by tissue typing with immunohistochemistry, immunofluorescence, or laser-microdissection mass spectrometry according to the technique available at the time of diagnosis^{29, 30}. The Mayo Foundation institutional review board approved the study, and all patients consented to have their medical records reviewed according to institutional review board practices.

Echocardiography

Because the study covered a 19 years period, clinical echocardiography protocols varied, with more of an emphasis on two-dimensional (2D) measurements. Patients were evaluated with standard echocardiographic protocols used in the Mayo Clinic Echocardiography Laboratory at the time of diagnosis. Measurements of diastolic thickness of the inter-ventricular septum and LV posterior wall, end-systolic and end-diastolic LV diameters 2D techniques, LV-mass, EF, left atrial volume and Doppler evaluation of diastolic function and SV were carried out as previously described^{31, 32}. MCF was calculated as left ventricular SV (obtained from a Doppler derived method) divided by left ventricular myocardial volume, which is defined as LV-mass divided by the mean density of myocardium (1.05 g/mL)¹⁹. Using a previously validated technique, LV end-diastolic and end-systolic volumes were calculated from two-dimensional echo guided M-mode echocardiographic dimensions as previously reported^33-35. In addition, a subset of 238 patients had measurement of left ventricular global peak longitudinal strain (LV-strain) using speckle-tracking as previously described³⁶. Sonographers with experience in obtaining 2D-STE performed the echocardiographic studies using a Vivid 7 or E9 GE Medical System with a 2.5- to 4.0-MHz transducer in accordance with the guidelines of the American Society of Echocardiography³¹. LV-strain measurements were performed during the echocardiographic examination with dedicated vendor-specific software (EchoPAC PC version 6.0, GE Healthcare Co).

Reproducibility of SV, MCF and LV-strain was assessed in a separate, yet comparable, study cohort from the Pavia Amyloidosis Center. Briefly, echo data from 150 consecutive patients with AL amyloidosis, were analyzed twice by the same observer, allowing a 2-week time interval between two blinded measurements. This allowed assessing the intra-observer reproducibility of the different measures. A second observer repeated the same measures independently, to assess inter-observer reproducibility. Reproducibility was assessed in terms of coefficients of variation (CV) and percentage difference (PD), which is defined as the absolute difference between two paired values divided by the average of the two values, and expressed as a percentage.

Biomarkers

In all patients, blood was collected at the time of diagnosis. Creatinine, glomerular filtration rate (GFR), alkaline phosphatase, albumin and 24-hour protein loss were measured at the time of the diagnosis in all the patients. Since 2004, measurement of NT-proBNP, cardiac troponin T and free light chain were added to the routine panel. The cardiac biomarkers NT-proBNP (cutoff 1800 ng/L) and troponin T (cutoff 0.025 mcg/L), difference between amyloidogenic and non-amyloidogenic free light chain (dFLC, cutoff 180 mg/L), were combined into the Mayo 2012 cardiac staging system²⁶. Stage I patients have all markers below the cutoffs, stage II only one marker above the cutoff, stage III patients three markers above the cutoffs and stage IV all markers above the cutoffs.

Statistical analysis

Continuous variables were expressed as median values and interquartile range, and categorical variables as frequencies and percentages. Correlation among variables was summarized using Pearson's correlation coefficient. Survival was analyzed using the Kaplan-Meier method to estimate the distribution of survival as a function of the follow-up duration, while censoring those not known to be deceased at last known follow-up. The best cutoffs of all the variables were evaluated using the Contal and O'Quigley method. Using these cutoffs, sensitivity and specificity were calculated based on 2 years mortality. Hazard ratios (HR) and corresponding 95% confidence intervals (CI) were estimated using Cox proportional hazards. When analyzing continuous variables, HR are reported per 1 standard deviation in each variable to allow comparability between variables with different scales. In the subset of patients assessable for cardiac biomarkers, Harrell's survival c-Statistics were calculated to assess the discriminatory ability of the models and compare different models for predicting overall mortality³⁷. Comparisons among c-Statistics were computed using the standard errors derived from 1000 boostrap samples. Survival curves were plotted according to the Kaplan-Meier method and differences in survival were tested using the log-rank test. All analyses were performed using SAS software V.9.4 and JMP 10.0 (Cary, North Carolina, USA). All p-values were two-sided and p≤0.05 were considered to be statistically significant.

Results

The study population was composed of 754 patients with a biopsy proven AL amyloidosis diagnosis.A total of 495 (66%) of patients had a positive biopsy of the abdominal fat aspirate and 80 (10%) had a positive endomyocardial biopsy; 465 (61%) had amyloid deposits in the bone marrow biopsy. In Figure 1 we defined how the patient population was divided according to the time of diagnosis. Patient characteristics are shown in Supplemetal Table 1. Median age was 64 years, and 65% were male. A total of 591 (78%) patients were considered to have cardiac amyloidosis based on diagnostic criteria (IVS>12 mm in the absence of other cardiac causes and an elevated NT-proBNP (>332 ng/L)^{28, 38}. During a median follow-up of living patients of 40.8 months [interquartile range (IQR): 19-72 months], 480 (63%) have died. Standard two-dimensional echocardiographic parameters, SVI, LV-strain, and MCF are illustrated in Supplemental Table 2 along with their impacts on the risk of death expressed by HR. In particular, an SVI below the cutoff of 33 mL/m² was able to significantly predict overall survival [HR:2.95 (CI: 2.37, 3.66), P<0.001]. In addition, MCF <34% significantly predicted survival [HR:2.36 (1.96, 2.85)]. The measures of MCF, SVI and CI were associated with a reduced survival (Figure 2, Panel A-C). These cutoffs were able to discriminate the overall survival if the study population was divided into different cohorts according to the time of diagnosis, that had a different outcome as previously described by our group³⁹. In particular, in Supplemental Figure 1 we reported the outcome of patients in the cohort of 2005-2009 (Panel A/B) and the cohort 2010-2015 (Panel C/D) for both SVI and MCF. It is also shown the effect of circulating cardiac biomarkers, assessed by the prognostic impact of the cardiac biomarker staging system (Supplemental Figure 2). Sensitivity and specificity data of the different cutoffs are reported in Supplemental Table 3.

Figure 1.

Consort diagram of the study population.

Figure 2.

Overall survival (OS) according to echocardiographic variables (N=452).

1. OS according to MCF.
2. OS according to SVI.
3. C. OS according to CI.

In the 754 patients evaluable for MCF, SVI and CI, incremental multi-variable modeling was performed to assess the association of echocardiographic variables with risk of death, excluding LV-strain due to the limited number of patients with available strain measurements. Since MCF, SVI, and CI all contain SV as part of their calculation, separate models were generated for each of them (Table 1). The three variables (MCF, SVI, CI) were independent predictors of survival along with EF, systolic blood pressure (SBP) and type of first-line therapy (autologous stem cell transplant, ASCT). Other echocardiographic parameters, such as IVS and PW, and the presence of pericardial effusion were not independent of MCF, but IVS remained an independent predictor in the models with SVI and CI (Table 1).

Table 1.

Multi-variable proportional hazards model for risk of death using echocardiographic parameters only (n=754).

	MCF<33.6% model		SVI<33 mL/m2model		CI<2.44 model
Variable	HR(95%CI)	P	HR(95%CI)	P	HR(95%CI)	P
MCF, SVI, or CI	1.72(1.39, 2.13)	<0.001	1.90(1.47, 2.46)	<0.001	1.50(1.20, 1.87)	<0.001
EF<55%	1.73(1.40, 2.14)	<0.001	1.60(1.24, 2.06)	<0.001	1.73(1.39, 2.16)	<0.001
SBP<95 mmHg	1.44(1.16, 1.81)	0.001	1.47(1.13, 1.90)	0.004	1.42(1.14, 1.78)	0.002
IVS>12 mm	Not-included		1.44(1.07, 1.93)	0.02	1.48(1.17, 1.86)	0.001
ASCT	0.57(0.43, 0.76)	<0.001	0.50(0.35, 0.72)	<0.001	0.56(0.42, 0.74)	<0.001

Additional analysis was performed using the subset of 452 patients with available circulating cardiac biomarkers. This subgroup could be stratified with the use of the Mayo 2012 cardiac biomarker staging that identified four stages with significant different overall survival (Supplemental Figure 2). A significant worsening of the echocardiographic parameters according to the different biomarkers staging was noted (Supplemental Table 4). In particular, while the median EF in stage IV resulted still>50%, MCF, SVI and LV-strain were significantly reduced compared to the median value. MCF, SVI and CI each retained prognostic value upon addition of cardiac biomarkers staging (Mayo 2012 stage III and IV) to the models shown in Table 2. Results were similar when staging was analyzed across 4 levels rather than combined, though MCF was only of borderline significance (Supplemental Table 5). Other useful echocardiographic variables that resulted significant predictor of survival at univariate analysis were tested in a stepwise multivariate approach. In particular, IVS, PW, deceleration time and relative wall thickness were added to the analyses but none of these variables were prognostic in the models with SVI in addition to cardiac biomarker staging. To identify the model most predictive of survival we compared the three systems containing the prognostic variables (SVI, MCF and CI) using the survival c-Statistic. A baseline model of EF, SBP and biomarker staging had a c-Statistic 0.70 (95%CI 0.67, 0.74). The addition SVI—but not MCF or CI—added prognostic power to the basal model (Table 2) with respective c-Statistics of 0.72 (0.68, 0.75), P=0.010. To further validate these analyses, a second method of measuring SV, the Simpson method incorporating 2D data, was used to calculate MCF and SVI; MCF was no longer independently prognostic, but SVI was (Supplemental Table 6).

Table 2.

Multi-variable proportional hazards model for risk of death incorporating circulating cardiac biomarkers (n=452).

	MCF<33.6% model		SVI<33 mL/m2 model		CI<2.44 model
Variable	HR(95%CI)	P	HR(95%CI)	P	HR(95%CI)	P
MCF, SVI, or CI	1.54(1.11, 2.14)	0.009	1.92(1.42, 2.60)	<0.001	1.48(1.10, 1.99)	0.01
EF<55%	1.59(1.20, 2.10)	0.001	1.41(1.05, 1.88)	0.02	1.58(1.19, 2.10)	0.002
SBP<95 mmHg	1.51(1.13, 2.02)	0.006	1.40(1.04, 1.88)	0.03	1.50(1.11, 2.01)	0.007
Stage III&IV	1.83(1.30, 2.58)	<0.001	1.89(1.36, 2.63)	<0.001	2.07(1.50, 2.86)	<0.001
ASCT	0.62(0.40, 0.95)	0.03	0.61(0.40, 0.94)	0.02	0.65(0.42, 0.99)	0.05
c-Statistic	0.70(0.67, 0.74)		0.72(0.68, 0.75)		0.70(0.67, 0.74)

In order to evaluate the contribution of LV-strain, the 238 patients who had baseline LV-global peak systolic strain assessment were included. The optimal cutoff of LV-strain predicting mortality at 2 years was greater than or equal to -14% (sensitivity:86.2%; specificity:57.1%) and was able to sharply discriminate the overall survival of patients (Supplemental Figure 3, A). Of note, MCF was highly correlated with LV-strain (r:-0.85, P<0.001, Supplemental Figure 3, B). The correlation between LV-strain and SVI was less strong (r:-0.70, P<0.001, not shown). When LV-strain was added to EF, SBP and cardiac biomarker stages III and IV, LV-strain was an independent risk factor (Table 3, model A); however, the addition of SVI to model A drove LV-strain from the model (Table 3, models B and C). This analysis led us to investigate a simplified model containing circulating biomarker stage and each of the most powerful cardiac measures in the group of patients evaluable for cardiac biomarkers (Supplemental Table 7). The addition of SVI to biomarker staging system had the higher c-Statistic as compared to biomarkers alone. In the subset of patients with advanced cardiac stages (stage III and IV and cardiac stage IIIa and IIIb, according to Wechalekar et al.²⁷) SVI was able to identify a subgroup of patients with a significant better overall survival (Figure 3).

Table 3.

Multi-variable proportional hazards models for risk of death incorporating LV-strain and SVI (n=238).

	Model A		Model B		Model C
Variable	HR(95%CI)	P	HR(95% CI)	P	HR(95% CI)	P
LV-Strain≥-14%	1.95(1.04, 3.66)	0.04	1.75(0.93, 3.33)	0.08	Not-included
EF<55%	1.26(0.81, 1.97)	0.31	1.02(0.64, 1.64)	0.92	1.16(0.73, 1.84)	0.54
SBP†<95 mmHg	1.60(1.01, 2.52)	0.04	1.38(0.86, 2.20)	0.18	1.36(0.85, 2.18)	0.20
Stage III&IV	2.69(1.44, 5.01)	0.002	2.51(1.34, 4.70)	0.004	3.11(1.73, 5.59)	<0.001
SVI<33 mL/m2	Not-included		2.00(1.24, 3.23)	0.004	2.14(1.32, 3.47)	0.002
ASCT	0.34(0.12, 0.96)	0.04	0.37(0.13, 1.05)	0.06	0.36(0.13, 1.02)	0.05
c-statistic	0.75(0.70, 0.80)		0.76(0.72, 0.81)		0.75(0.70, 0.80)

Figure 3.

Overall survival according to SVI in different biomarker stagings (N=452).

1. OS according to SVI in Mayo 2012 stage III.
2. OS according to SVI in Mayo 2012 stage IV.
3. OS according to SVI in stage IIIa.
4. OS according to SVI in stage IIIb.

The intra-observer and inter-observer coefficients of variations were, respectively, 0.4% and 0.7% for LV end-diastolic diameter, 3.0% and 3.4% for septal wall thickness, 2.9% and 3.2% for posterior wall LV-thickness. As to LV-strain data, intra-observer and inter-observer percentage difference were 5%±4% [r=0.98; 95%CI limits of agreements (LOA), 2.38% to 0.99%] and 6%±4% (r=0.96; 95%CI LOA, 3.02% to 1.39%), respectively. MCF and SV showed slightly worse reproducibility. Intra-observer and inter-observer percentage difference were 8%±5% (r=0.86; 95%CI LOA, 5.22% to 2.99%) and 9%±6% (r=0.82; 95%CI LOA, 5.99% to 3.23%) for MCF and 8%±5% (r=0.84; 95%CI LOA, 5.44% to 2.89%) and 10% ± 6% (r=0.78; 95%CI LOA, 6.66% to 3.43%) for SV.

Discussion

Understanding outcomes in patients with AL is essential given the heterogeneity of disease. Echocardiography was one of the first and most important diagnostic and prognostic tools for cardiac assessment among patients with AL⁴⁰. In terms of prognosis, soluble cardiac biomarkers have largely supplanted echocardiography. We set out to assess how measures like MCF⁴¹, LV-strain⁴², SVI, and CI held up against other simple prognostic measures like troponin T, NT-proBNP²⁶ and systolic blood pressure²⁷. Using one of the largest cohorts of subjects with AL amyloidosis with echocardiographic assessment and soluble cardiac biomarkers, we found that MCF correlated quite well with LV-strain and that it was prognostic for overall survival even in the context of cardiac biomarkers. The prognostic impact of these variables was also independent of the type of therapy. The echocardiographic variables reported in our manuscript are similar to those previously reported by others^{15, 16, 20}. In addition, SVI was able to identify a subset of patients with a better prognosis, even in the context of patients with severe cardiac involvement according to the cardiac biomarker staging systems. Therefore, we observed that SVI performed as well as MCF and LV-strain in survival models leading us to believe that this simple measure—SVI—is the favored echocardiographic parameter to predict outcomes among patients with AL amyloidosis. Interestingly, a similar cutoff of SVI (35mL/m²) has been found to be a useful prognostic marker in patients with aortic stenosis with preserved EF⁴³. An advantage of SVI is that is routinely calculated as part of a standard echocardiographic study, requires little time, does not require additional calculations or analysis, is not limited by the challenges in determining left ventricular mass as is required for MCF, or those of LV-strain (2D-image-quality, sonographer expertise, and vendor and software variations). We did show, however, that LV-strain when using the same vendor and software data was slightly more reproducible than MCF and SV (that are measured by manual tracing and hence more operator-dependent).

Because the published data on MCF ⁴¹ was compelling, we speculated that SVI might be as useful and more readily available in a busy cardiology practice since MCF is defined as the ratio between SV and myocardial volume. The MCF evaluates shortening in relation to the degree of amyloid infiltration (as inferred from the myocardial chamber volume) and quantitates the effect of amyloid infiltration in the myocardium. Progressive amyloid deposition in the myocardium causes an increase in left ventricular myocardial volume concomitant with a decline in SV, which is indicated by declining MCF⁴¹. It is important to note that, in those echocardiographic laboratories in which the LV-strain is not yet (or readily) available, the measure of MCF appears to be a reasonable substitute for patients with AL amyloidosis, serving as the “poor man's LV-strain.” Similar to EF, the MCF is dependent on afterload¹⁹ but in contradistinction to EF in AL amyloid in which SV and EDV show parallel declines (and hence no change in EF), the MCF declines because increases in MV (due to amyloid deposition) are accompanied by SV reduction. Thus, in the setting of amyloid-induced LV mass increase, (that can be also viewed as amyloid-induced LV “hypertrophy”), MCF is reduced, analogous to mid-wall fractional shortening⁴⁴. Although MCF does not require specific software or a specific training, since it is calculated from conventional echocardiographic measures, it is not automatically calculated by standard package software. In addition, the measurements that are needed for the calculation of LV-mass and myocardial volume are hampered by possible errors. To limit these, we focused only on 2D-derived data, excluding from the study all the subjects that had only M-mode based measurements of LV chamber size. However, the prognostic ability of MCF resulted limited, if the SV was calculated by only 2D measurement. Another important limitation of MCF is that the measurement of myocardial volume assumes a constant myocardial density (1.05g/ml) irrespective of the underlying pathology, which presumes that there is no difference between myocardial tissue and amyloid tissue density. Previous data have shown that myocardial density varies in several conditions such as cardiac infarction⁴⁵. Moreover, a recent study with cardiac magnetic resonance documented a different degree of extracellular volume expansion in the main types of cardiac amyloidosis⁴⁶. The use of SVI as a prognostic marker could overcome this assumption and the different limitations highlighted for MCF. LV-strain is a powerful tool, with the ability to promote early diagnosis by detection of subtle abnormalities in ventricular function and recognition of the characteristic pattern of apical sparing. However, unlike SVI, LV-strain requires specific vendor-dependent software, expertise, and time and is not available in all centers. In addition, the results are not standardized across ultrasound systems leading to a recommendation that serial studies be performed on the same machine. Recognizing the role of inter- and intra-observer variability in any measurement as a potential limitation, we integrated our data with an evaluation of reproducibility on a comparable patients' cohort. The reasonable reproducibility of Doppler-derived measurement of SV reinforce the role of our new proposed approach. Indeed, we documented a high prognostic role of this measurement in a real-world data setting, encompassing several years in a referral echocardiographic laboratory. In addition, a validation analysis confirmed our results by assessing the prognostic role of 2D-derived SVI. Unfortunately, a limitation of our study is that only few patients had cardiac magnetic resonance data and 3D echocardiographic data available. Therefore, no comparison with this method was possible. In addition, in this series we did not have a correlation with EKG data. Further studies in large series are warranted to assess the prognostic impact of SVI in the setting of cardiac biomarkers.

In conclusion, SVI is a simple, powerful, independent predictors of survival in patients with AL amyloidosis. In particular, the prognostic role of SVI was independent of EF, BP, the cardiac biomarkers staging, and LV-strain. SVI performed at least as well as LV-strain for predicting survival in AL, independent of biomarker staging. Since SVI is routinely calculated and widely available, it could serve as the preferred echocardiographic measure to predict outcomes in AL amyloidosis patients.