Background
The prognostic value of Doppler myocardial imaging, including myocardial velocity imaging, strain, and strain rate imaging, in patients with primary (AL) amyloidosis is uncertain. The aim of this longitudinal study was to identify independent predictors of survival, comparing clinical data, hematologic and cardiac biomarkers, and standard echocardiographic and Doppler myocardial imaging measures in a cohort of patients with AL amyloidosis.
Methods
A total of 249 consecutive patients with AL amyloidosis were prospectively enrolled. The primary end point was all-cause mortality, and during a median follow-up period of 18 months, 75 patients (30%) died. Clinical and electrocardiographic data, biomarkers (brain natriuretic peptide and cardiac troponin T) and standard echocardiographic and longitudinal systolic and diastolic Doppler myocardial imaging measurements for 16 left ventricular segments were tested as potential independent predictors of survival.
Results
Age (hazard ratio [HR], 1.03; P = .03), New York Heart Association class III or IV (HR, 2.47; P = .01), the presence of pleural effusion (HR, 1.79; P = .08), brain natriuretic peptide level (HR, 1.29; P = .01), ejection time (HR, 0.99; P = .13), and peak longitudinal systolic strain of the basal anteroseptal segment (HR, 1.05; P = .02) were independent predictors in the final model.
Conclusions
Multivariate survival analysis identified independent predictors of clinical outcome in patients with AL amyloidosis: New York Heart Association class III or IV, presence of pleural effusion, brain natriuretic peptide level > 493 pg/mL, ejection time < 273 ms, and peak longitudinal systolic basal anteroseptal strain less negative than or equal to −7.5% defined a high-risk group of patients.
Primary (AL) amyloidosis is characterized by the extracellular deposition of pathologic insoluble fibrillar protein in organs and tissues. Congestive heart failure is seen in approximately 25% of patients with amyloidosis, and its development is associated with an average survival of <4 months. There is strong evidence that cardiovascular biomarkers and two-dimensional (2D) and Doppler echocardiographic findings are meaningful predictors of survival in AL amyloidosis. The prognostic value of hematologic biomarkers has also been demonstrated. Although strain rate imaging and strain imaging have been shown to be very sensitive for the detection of early cardiac dysfunction in patients with systemic amyloid, their role as predictors of survival is uncertain.
The purposes of the present observational study were twofold: (1) to identify independent predictors of all-cause mortality, considering demographic, clinical, hematologic, and cardiovascular data, including echocardiography with Doppler myocardial imaging (DMI), through a multivariate survival analysis (death from any cause was the outcome), and (2) to establish cut points for those predictors that would be useful in clinical evaluation of patients with AL amyloidosis.
Methods
Study Population
This study was approved by the institutional review board of the Mayo Clinic. Subjects with AL amyloidosis were prospectively selected from patients undergoing evaluation in the Division of Hematology at the Mayo Clinic (Rochester, MN) from January 1, 2004, through October 31, 2005, and referred for echocardiography. A cohort of 279 consecutive patients with AL amyloidosis who underwent complete echocardiographic and Doppler exams constituted the initial study population. All patients were evaluated by a hematologist and, when appropriate, by other specialists.
The diagnosis of immunoglobulin light chain amyloidosis was made by biopsy of subcutaneous fat or an involved organ that demonstrated typical Congo red birefringence under polarized light. AL amyloidosis was further confirmed by immunohistochemical typing of the amyloid. Exclusion criteria were familial or secondary amyloidosis (n = 18); senile amyloidosis (n = 7); history of moderate or greater systemic or pulmonary hypertension (n = 5); coronary artery disease detected by history of typical symptoms, as well as by regional hypokinesia or akinesia of myocardial segments by standard echocardiography, or induced by exercise echocardiography, or >50% coronary stenosis by angiography (n = 0); and history of moderate or greater degrees of valvular heart disease (n = 0).
Thirty patients were excluded from the analysis on the basis of these exclusion criteria, yielding a final study population of 249 patients. Patients with chronic atrial fibrillation (n = 10) were not excluded.
Patients with AL amyloidosis were evaluated for the extent of amyloid-related organ involvement and for dominant organ involvement, integrating standard criteria described elsewhere, along with information provided by the assessment of cardiac biomarkers and left ventricular (LV) wall thickness and diastolic function by echocardiography.
We performed a time-to-event data analysis considering patients altogether, but also selecting only those in whom laboratory and imaging assessments were performed <3 months from diagnosis, to determine the independent predictors of survival. The vital status of each patient was confirmed by review of medical records and the Social Security Death Index. The primary end point was all-cause mortality. Patients who had no records of death were censored at the time they were last known to be alive.
Biomarkers
For all patients, brain natriuretic peptide (BNP), N-terminal pro–BNP (NT-proBNP), and cardiac troponin T levels; creatinine; glomerular filtration rate; proteinuria over 24 hours; and alkaline phosphatase, albumin, and uric acid levels were collected at the time of the echocardiographic examination. Immunoglobulin free light chain and M protein measurements were collected as well. Plasma BNP, NT-proBNP, and troponin T levels were measured as previously described. To pool the data of the monoclonal kappa and lambda patients, the clonal free light chain was considered the “involved” immunoglobulin free light chain for the purpose of the analyses. Immunoglobulin clonal free light chain quantitation was carried out as previously described.
Standard Ultrasound Examination
All ultrasound examinations were performed with a commercially available echocardiographic instrument (Vivid 7; GE Vingmed Ultrasound AS, Horten, Norway). Systolic and diastolic parameters as assessed by standard echocardiography, pulsed-wave Doppler tissue imaging, and timing assessments were measured as previously described. Offline analysis of standard echocardiographic variables was performed using dedicated software (ProSolv CardioVascular Analyzer version 3.0; ProSolv CardioVascular, Indianapolis, IN). Each measurement reported is the average of measurements from 3 consecutive beats. This same method was used for patients with chronic atrial fibrillation (n = 10), because their rates were controlled such that their ventricular rhythms were nearly regular. None of the patients was tachycardic during image acquisition. The presence of pleural effusion was ascertained by chest x-ray.
Strain Imaging Data Acquisition and Analysis
DMI was used to measure longitudinal LV myocardial velocity, strain, and strain rate. Narrow-sector views for each LV wall in apical 4-chamber, long-axis, and 2-chamber standard views were acquired as previously described. Color 2D digital data from 3 cardiac cycles were analyzed offline using dedicated software (EchoPAC; GE Vingmed Ultrasound AS). Sample volumes were placed on basal, middle, and apical LV segments of the anterolateral, inferoseptal, inferolateral, anteroseptal, inferior, and anterior walls to assess longitudinal DMI.
Longitudinal peak values were determined for systolic myocardial velocity (sMV), systolic strain rate (sSR), and systolic strain (sS). Analysis was performed considering all 16 LV segments individually (data not shown) and also combining them in clusters according to LV levels and according to LV walls. DMI values (sMV, sSR, and sS) were averaged for the 6 basal segments (basal mean), for the 6 middle segments (middle mean), and for the 4 apical segments (apical mean). Values of sSR and sS were averaged for the anterolateral, inferoseptal, inferolateral, anteroseptal, inferior, and anterior walls. Longitudinal early (dMV-E and dSR-E) and late (dMV-A and dSR-A) diastolic peaks were also measured from myocardial velocity imaging and strain rate imaging for each of the 16 LV segments. Clusters of single segments, by level and by wall, were determined using the same strategy used for the systolic measurements.
Statistical Analysis
Statistical analyses were performed using commercially available software (SAS version 8.2; SAS Institute Inc, Cary, NC). The natural logs of BNP, NT-proBNP, creatinine, 24-hour proteinuria, alkaline phosphatase, and clonal free light chain were used in the construction of univariate and multivariate survival models to reduce skewness. Covariate information (clinical data, electrocardiographic findings, biomarkers, and echocardiographic variables) was collected at the time of the echocardiographic examinations in all enrolled patients. Observations with missing values for any of the covariates were excluded from the analysis.
Many of the patients in the study were first referred to the Mayo Clinic some months after their initial diagnoses, a condition known as left truncation. For these cases, a naive computation of the survival curve for time from diagnosis will be biased upward. Left truncation was appropriately taken into account for the survival curves and Cox regression analysis by not considering a subject to be in the risk set until the time of echocardiography. Univariate and multivariate analyses of the time to events were performed using Cox proportional-hazard models with demographic, clinical, electrocardiographic, standard echocardiographic, and DMI variables as independent variables. Multivariate models were created for each group of variables sequentially (clinical and electrocardiographic findings, biomarkers, standard echocardiographic findings, systolic DMI measures, and diastolic DMI measures) using a bootstrap approach. Candidate variables for the bootstrap selection were those with univariate P values < .20. One thousand bootstrap samples of size n = 249 were selected with replacement from the data set. For each bootstrap sample, a stepwise model selection technique with a P value for selection and retention of .05 was applied, and the number of times each candidate variable appeared in the final model was tallied. Those variables that appeared in the greatest percentage of models were then placed into the model together. This was repeated for each group of variables sequentially, with those variables from the previous group being retained in subsequent models even if they lost significance as additional variables were added. Nested model tests were performed to see if there was an incremental gain in information at each model step.
The final model was constructed with clinical, electrocardiographic, biomarker, and echocardiographic variables in addition to DMI measures, and it was intended to see if DMI measures added information to the model. We used the survival C-statistic, a measure of concordance between observed and predicted survival from Cox proportional-hazards models, to evaluate the discriminatory ability of the model. This measure is similar to the area under the curve for binary end points, but the survival C-statistic is interpreted as the probability of correctly ordering event times using covariate risk score. Standard errors for C-statistics were obtained from 1000 bootstrap samples and used to construct 95% confidence intervals (CIs). A confirmatory analysis was then performed by fitting the final multivariate model obtained in the overall analysis on the subgroup of patients who had diagnoses of AL amyloidosis within 3 months of echocardiography (n =106).
Cut-point values for independent predictors in the final model were determined by the method of Contal and O’Quigley using log-rank statistics. This method essentially calculates all possible splits and finds the one that maximizes the log-rank statistic. Those authors also proposed a test statistic that allows one to obtain a P value that adjusts for this maximization. This method of finding cut points has been shown to be similar to approaches based on smoothed residuals from the Cox model, but it is less subjective. Cox regression analysis adjusted for the delayed entry was then performed between the groups defined by the cut point for each predictor.
We created a risk stratification score, initially including all 8 of the independent predictors of survival that were defined by the final model. Those 6 independent predictors that are easiest to measure were then subselected to build a scoring system that is more practical from a clinical standpoint. Continuous variables were dichotomized at their optimal cut points, and a Cox model was fit with only dichotomous variables. The parameter estimates from this model were then used as a basis to assign a score to each variable. The distribution of the sum of these scores for each patient was then examined, and patients were classified into 4 groups on the basis of their total scores. A Kaplan-Meier curve was then drawn to show the survival within each risk group.
We also wanted to determine whether DMI measures provide information on risk stratification even in the absence of abnormalities in independent or generally accepted risk factors, including clinical, standard echocardiographic, and biochemical markers. Therefore, we excluded from the analysis all those patients with altered values in such independent or commonly accepted predictors of mortality and compared the most significant DMI measurements obtained by Cox regression models in our population between survivors and nonsurvivors at the end of follow-up.
Intraobserver and Interobserver Variability
To examine intraobserver variability (repeatability), a sample of 10 echocardiographic examinations were randomly selected for masked review by the same investigator. To examine interobserver variability, a second investigator blinded to the clinical information and to the results of the first investigator examined 10 randomly selected echocardiographic exams. The number of the digital echocardiographic clip was provided to the repeat observer for the assessment of either intraobserver or interobserver variability. Intraclass correlation coefficients (ICCs) for the same observer and different observers were calculated using previously described formulas for single segments and for the global mean of each DMI modality. Data are presented as mean ± SD, as medians with interquartile ranges (for biomarkers), or as counts and percentages. A difference was considered statistically significant when the P value was <.05. In the multivariate model, a variable was considered a significant predictor of survival when the P value was <.10.
Results
Baseline demographic and clinical variables as well as cardiac and hematologic biomarker levels are detailed in Table 1 . Standard 2-dimensional and Doppler echocardiographic results are detailed in Table 2 . Longitudinal systolic and diastolic DMI measurements for our study population are displayed in Tables 3 and 4 , respectively.
Variable | Value | Univariate HR ( P ) |
---|---|---|
Age (y) | 63 ± 10 | 1.03 (.005) |
Men | 155 (62%) | 1.80 (.02) |
Diagnosis to evaluation (mo) | 6 (24) | |
Heart rate (beats/min) | 76 ± 13 | 1.02 (.01) |
NYHA class III/IV | 50 (20%) | 3.00 (<.0001) |
Pericardial effusion | 81 (33%) | 2.00 (.004) |
Pleural effusion ∗ | 30 (13%) | 3.50 (<.0001) |
Low voltage on electrocardiography | 53 (22%) | NS |
Atrial fibrillation | 10 (4%) | NS |
Pacemaker/implantable cardioverter-defibrillator | 13 (5%) | NS |
Stem cell transplantation | 129 (52%) | 0.50 (.009) |
Biomarkers | ||
BNP (pg/mL) † | 188.5 (413) | 1.30 (.004) |
NT-proBNP (pg/mL) | 909 (3160) | 1.30 (.001) |
Troponin T (ng/mL) | 0.01 (0.04) | 3.20 (<.0001) |
Creatinine (mg/dL) | 1.2 (0.7) | NS |
Glomerular filtration rate (mL/min/m 2 ) | 62 (40.5) | NS |
24-h proteinuria (mg/24 h) | 1.518 (5.8225) | NS |
Alkaline phosphatase (U/L) | 92 (55.5) | NS |
Albumin (g/dL) | 3 (1.07) | NS |
Uric acid (mg/dL) | 6.4 (2.9) | NS |
Involved free light chains (mg/dL) | 12.4 (24.915) | NS |
κ/λ | 0.2 (1.4) | NS |
M spike (serum) | 0.4 (0.9) | NS |
∗ There were 13 missing values for pleural effusion.
Variable | Value | Univariate HR ( P ) |
---|---|---|
LV thickness (mm) | 13 ± 3 | 1.10 (.0007) |
LV mass (g/m 2 ) | 118 ± 42 | NS |
LV end-systolic diameter (mm) | 28 ± 5 | NS |
LV end-diastolic diameter (mm) | 45 ± 6 | 0.90 (.003) |
Left atrial volume (cm 3 /m 2 ) | 39 ± 13 | NS |
Ejection fraction (%) | 62 ± 10 | NS |
Stroke volume (mL) | 80 ± 23 | 0.90 (.01) |
Cardiac index (L/m 2 /min) | 3.09 ± 0.7 | NS |
E-wave velocity (m/s) | 0.8 ± 0.2 | NS |
E deceleration time (ms) | 197 ± 50 | 0.90 (.03) |
A-wave velocity (m/s) | 0.7 ± 0.3 | NS |
E/A ratio | 1.5 ± 1.1 | 1.70 (.007) |
Pulmonary vein S-wave velocity (m/s) | 0.6 ± 0.6 | 0.20 (.007) |
Pulmonary vein D-wave velocity (m/s) | 0.6 ± 0.6 | NS |
Pulmonary vein A reversal velocity (m/s) | 0.3 ± 0.3 | 0.004 (.003) |
Mitral A–pulmonary vein A reversal | 28 ± 55 | NS |
Medial mitral annular S′ velocity (m/s) | 0.06 ± 0.01 | 0.80 (.004) |
Medial mitral annular E′ velocity (m/s) | 0.06 ± 0.02 | 0.70 (.004) |
Medial mitral annular A′ velocity (m/s) | 0.08 ± 0.02 | 0.80 (.0002) |
E/E′ ratio | 16.0 ± 8.1 | 1.05 (.0007) |
Diastolic dysfunction | ||
Grade 1/1a | 69 (32%) | |
Grade 2 | 67 (31%) | |
Grade 3/4 | 50 (23%) | 1.60 (<.0001) |
Left index of myocardial performance | 0.52 ± 0.2 | NS |
Ejection time (ms) ∗ | 282.81 ± 41.1 | 0.90 (.0004) |
Variable | Value | Univariate HR ( P ) |
---|---|---|
Myocardial velocity imaging (cm/s) | ||
Lateral | 3.25 ± 1.4 | NS |
Inferoseptal | 3.51 ± 1.2 | 0.70 (.002) |
Posterior | 4.29 ± 1.7 | NS |
Anteroseptal | 3.56 ± 1.2 | NS |
Inferior | 3.79 ± 1.3 | NS |
Anterior | 2.73 ± 1.2 | NS |
Basal mean | 4.75 ± 1.3 | 0.80 (.01) |
Middle mean | 3.26 ± 1 | NS |
Apical mean | 1.85 ± 0.8 | NS |
Global average | 3.47 ± 1 | 0.80 (.04) |
Strain rate imaging (1/s) | ||
Lateral | −0.78 ± 0.4 | NS |
Inferoseptal | −0.96 ± 0.4 | 4.00 (<.0001) |
Posterior | −0.81 ± 0.6 | NS |
Anteroseptal | −0.98 ± 2.9 | 3.40 (<.0001) |
Inferior | −0.95 ± 0.4 | 2.30 (.005) |
Anterior | −0.82 ± 0.3 | 2.30 (.03) |
Basal mean | −0.81 ± 0.5 | 3.90 (<.0001) |
Middle mean | −0.92 ± 0.7 | 3.50(.001) |
Apical mean | −0.93 ± 0.3 | NS |
Global average | −0.88 ± 0.4 | 5.60 (<.0001) |
Strain imaging (%) | ||
Lateral | −10.96 ± 5.3 | 1.10 (.008) |
Inferoseptal | −14.35 ± 6.5 | 1.10 (<.0001) |
Posterior | −11.62 ± 7 | NS |
Anteroseptal | −12.52 ± 6.1 | 1.10 (<.0001) |
Inferior | −13.57 ± 6.3 | 1.10 (<.0001) |
Anterior | −12 ± 4.7 | 1.10 (<.0001) |
Basal mean | −11.06 ± 6 | 1.10 (<.0001) |
Middle mean | −13.65 ± 5.3 | 1.10 (<.0001) |
Apical mean | −13.7 ± 4.9 | 1.10 (.0002) |
Global average | −12.71 ± 4.8 | 1.10 (<.0001) |
Variable | Value | Univariate HR ( P ) |
---|---|---|
Myocardial velocity imaging (cm/s) | ||
Lateral E | 3.99 ± 1.7 | 0.80 (.01) |
Lateral A | 3.71 ± 2 | 0.90 (.03) |
Inferoseptal E | 3.78 ± 1.6 | 0.70 (<.0001) |
Inferoseptal A | 4.3 ± 2 | 0.80 (.0002) |
Posterior E | 4.8 ± 2.3 | 0.80 (.0006) |
Posterior A | 5.01 ± 2.7 | 0.80 (.0009) |
Anteroseptal E | 3.8 ± 1.6 | 0.70 (.0003) |
Anteroseptal A | 4.28 ± 2.1 | 0.80 (.009) |
Inferior E | 3.69 ± 1.7 | 0.80 (.02) |
Inferior A | 4.61 ± 2.3 | 0.80 (.0002) |
Anterior E | 3.42 ± 1.6 | NS |
Anterior A | 3.06 ± 1.8 | 0.90 (.04) |
Basal mean E | 4.89 ± 2 | 0.80 (<.0001) |
Basal mean A | 5.46 ± 2.5 | 0.80 (.005) |
Middle mean E | 3.85 ± 1.6 | 0.70 (.002) |
Middle mean A | 4.02 ± 2 | 0.80 (.005) |
Apex mean E | 2.2 ± 0.9 | NS |
Apex mean A | 1.9 ± 1 | 0.70 (.02) |
Global average E | 3.81 ± 1.5 | 0.70 (<.0001) |
Global average A | 4.05 ± 1.8 | 0.80 (.001) |
Strain rate imaging 1/s | ||
Lateral E | 1.07 ± 0.5 | 0.50 (.007) |
Lateral A | 0.96 ± 0.6 | 0.60 (.04) |
Inferoseptal E | 1.06 ± 0.5 | 0.50 (.003) |
Inferoseptal A | 1.14 ± 0.6 | 0.50 (.008) |
Posterior E | 0.93 ± 0.8 | NS |
Posterior A | 0.89 ± 0.7 | 0.70 (.03) |
Anteroseptal E | 0.95 ± 0.5 | NS |
Anteroseptal A | 1.12 ± 0.6 | NS |
Inferior E | 0.88 ± 0.5 | 0.50 (.02) |
Inferior A | 1.07 ± 0.6 | 0.60 (.02) |
Anterior E | 1.10 ± 0.5 | NS |
Anterior A | 1.06 ± 0.6 | NS |
Basal mean E | 0.84 ± 0.5 | NS |
Basal mean A | 0.97 ± 0.5 | 0.50 (.0019) |
Middle mean E | 1.07 ± 0.5 | 0.50 (.007) |
Middle mean A | 1.03 ± 0.5 | 0.50 (.007) |
Apex mean E | 1.12 ± 0.5 | NS |
Apex mean A | 1.08 ± 0.6 | NS |
Global average E | 0.98 ± 0.4 | 0.40 (.009) |
Global average A | 0.99 ± 0.5 | 0.40 (.001) |
The mean age was 63 years, and 155 patients (68%) were men. The heart was the major organ involved in 93 patients (37%), and myocardial biopsy was performed in 37 patients (15%). The kidney was the major organ involved in 113 patients (45%), and within these two groups, 62 patients (24%) had major and equal involvement of both the heart and the kidney. The peripheral nervous system and gastrointestinal tract were primarily involved in 13 (5%) and 12 (5%) patients, respectively. High-dose chemotherapy (melphalan plus prednisone) followed by peripheral blood stem cell transplantation (PBSCT) was performed in 129 patients (52%).
Survival Analysis
Overall Analysis
The median follow-up time for censored patients was 21 months (interquartile range, 11-35 months), and the median follow-up time for patients who died was 5 months (interquartile range, 0.6-28 months). There were 75 deaths (30%) among these patients. Survival curves are shown in Figure 1 . Hazard ratios [HRs] by univariate analysis are reported in Tables 1 to 4 . The results of univariate survival analysis are not discussed further.
![](https://clinicalpub.com/wp-content/uploads/2023/09/banner1.png)