[Skip to Content]
Sign In
Individual Sign In
Create an Account
Institutional Sign In
OpenAthens Shibboleth
[Skip to Content Landing]
Figure 1.
B-Type Natriuretic Peptide (BNP), Troponin I (TnI), Retinol-Binding Protein 4 (RBP4), and Transthyretin (TTR) Serum Biomarkers
B-Type Natriuretic Peptide (BNP), Troponin I (TnI), Retinol-Binding Protein 4 (RBP4), and Transthyretin (TTR) Serum Biomarkers

Note that the TnI and BNP concentrations (C and D) are both well above normal limits and inconsistently increased in transthyretin cardiac amyloidosis (ATTR) caused by substitution of valine for isoleucine at codon 122 (V122I), whereas the RBP4 and TTR concentrations (A and B) are lower. Horizontal bars reflect median values; boxes, 25th and 75th percentiles; and error bars, interquartile ranges.

Figure 2.
Receiver Operating Characteristic Curves
Receiver Operating Characteristic Curves

A, Retinol-binding protein 4 (RBP4) concentration in the diagnosis of transthyretin cardiac amyloidosis (ATTR) caused by substitution of valine for isoleucine at codon 122 (V122I). B, Full clinical model in the diagnosis of ATTR V122I cardiac amyloidosis in the development cohort. Variables included RBP4 concentration, left ventricular ejection fraction (LVEF), interventricular septal diameter (IVSD), transthyretin (TTR) concentration, QRS voltage, and the presence of grade 3 diastolic dysfunction (G3DD). C, Parsimonious model calculated by our sensitivity analysis in the diagnosis of ATTR V122I cardiac amyloidosis in the development cohort. Variables included RBP4 concentration, LVEF, IVSD, and QRS voltage. D, Full clinical model of the diagnosis of ATTR V122I cardiac amyloidosis in the validation cohort. Variables included RBP4 concentration, LVEF, IVSD, TTR concentration, QRS voltage, and the presence of G3DD. E, Parsimonious model calculated by our sensitivity analysis of the diagnosis of ATTR V122I cardiac amyloidosis in the validation cohort. Variables included RBP4 concentration, LVEF, IVSD, and QRS voltage. AUC indicates area under the curve.

Table 1.  
Baseline Differences Between Patients With ATTR V122I Amyloidosis and Nonamyloid Heart Failurea
Baseline Differences Between Patients With ATTR V122I Amyloidosis and Nonamyloid Heart Failurea
Table 2.  
Echocardiographic and Electrocardiographic Differences Between Patients With ATTR V122I Amyloidosis and Nonamyloid Heart Failurea
Echocardiographic and Electrocardiographic Differences Between Patients With ATTR V122I Amyloidosis and Nonamyloid Heart Failurea
1.
Ruberg  FL, Maurer  MS, Judge  DP,  et al.  Prospective evaluation of the morbidity and mortality of wild-type and V122I mutant transthyretin amyloid cardiomyopathy: the Transthyretin Amyloidosis Cardiac Study (TRACS).  Am Heart J. 2012;164(2):222-228.e1.PubMedGoogle ScholarCrossref
2.
Jacobson  DR, Pastore  RD, Yaghoubian  R,  et al.  Variant-sequence transthyretin (isoleucine 122) in late-onset cardiac amyloidosis in black Americans.  N Engl J Med. 1997;336(7):466-473.PubMedGoogle ScholarCrossref
3.
Jacobson  DR, Alexander  AA, Tagoe  C, Buxbaum  JN.  Prevalence of the amyloidogenic transthyretin (TTR) V122I allele in 14 333 African-Americans.  Amyloid. 2015;22(3):171-174.PubMedGoogle ScholarCrossref
4.
Yamashita  T, Hamidi Asl  K, Yazaki  M, Benson  MD.  A prospective evaluation of the transthyretin Ile122 allele frequency in an African-American population.  Amyloid. 2005;12(2):127-130.PubMedGoogle ScholarCrossref
5.
Connors  LH, Prokaeva  T, Lim  A,  et al.  Cardiac amyloidosis in African Americans: comparison of clinical and laboratory features of transthyretin V122I amyloidosis and immunoglobulin light chain amyloidosis.  Am Heart J. 2009;158(4):607-614.PubMedGoogle ScholarCrossref
6.
Damy  T, Costes  B, Hagège  AA,  et al.  Prevalence and clinical phenotype of hereditary transthyretin amyloid cardiomyopathy in patients with increased left ventricular wall thickness.  Eur Heart J. 2016;37(23):1826-1834.PubMedGoogle ScholarCrossref
7.
Ruberg  FL, Berk  JL.  Transthyretin (TTR) cardiac amyloidosis.  Circulation. 2012;126(10):1286-1300.PubMedGoogle ScholarCrossref
8.
Berk  JL, Suhr  OB, Obici  L,  et al; Diflunisal Trial Consortium.  Repurposing diflunisal for familial amyloid polyneuropathy: a randomized clinical trial.  JAMA. 2013;310(24):2658-2667.PubMedGoogle ScholarCrossref
9.
Maurer  MS, Grogan  DR, Judge  DP,  et al.  Tafamidis in transthyretin amyloid cardiomyopathy: effects on transthyretin stabilization and clinical outcomes.  Circ Heart Fail. 2015;8(3):519-526.PubMedGoogle ScholarCrossref
10.
Coelho  T, Adams  D, Silva  A,  et al.  Safety and efficacy of RNAi therapy for transthyretin amyloidosis.  N Engl J Med. 2013;369(9):819-829.PubMedGoogle ScholarCrossref
11.
Fontana  M, Pica  S, Reant  P,  et al.  Prognostic value of late gadolinium enhancement cardiovascular magnetic resonance in cardiac amyloidosis.  Circulation. 2015;132(16):1570-1579.PubMedGoogle ScholarCrossref
12.
Gillmore  JD, Maurer  MS, Falk  RH,  et al.  Nonbiopsy diagnosis of cardiac transthyretin amyloidosis.  Circulation. 2016;133(24):2404-2412.PubMedGoogle ScholarCrossref
13.
Raghu  P, Sivakumar  B.  Interactions amongst plasma retinol-binding protein, transthyretin and their ligands: implications in vitamin A homeostasis and transthyretin amyloidosis.  Biochim Biophys Acta. 2004;1703(1):1-9.PubMedGoogle ScholarCrossref
14.
Wei  S, Episkopou  V, Piantedosi  R,  et al.  Studies on the metabolism of retinol and retinol-binding protein in transthyretin-deficient mice produced by homologous recombination.  J Biol Chem. 1995;270(2):866-870.PubMedGoogle ScholarCrossref
15.
Berni  R, Malpeli  G, Folli  C, Murrell  JR, Liepnieks  JJ, Benson  MD.  The Ile-84—>Ser amino acid substitution in transthyretin interferes with the interaction with plasma retinol-binding protein.  J Biol Chem. 1994;269(38):23395-23398.PubMedGoogle Scholar
16.
Robin  X, Turck  N, Hainard  A,  et al.  pROC: an open-source package for R and S+ to analyze and compare ROC curves.  BMC Bioinformatics. 2011;12:77.PubMedGoogle ScholarCrossref
17.
Zhang  Z.  Multiple imputation with multivariate imputation by chained equation (MICE) package.  Ann Transl Med. 2016;4(2):30.PubMedGoogle Scholar
18.
Friedman  J, Hastie  T, Tibshirani  R.  Regularization paths for generalized linear models via coordinate descent.  J Stat Softw. 2010;33(1):1-22.PubMedGoogle ScholarCrossref
19.
Hosmer  DW, Hosmer  T, Le Cessie  S, Lemeshow  S.  A comparison of goodness-of-fit tests for the logistic regression model.  Stat Med. 1997;16(9):965-980.PubMedGoogle ScholarCrossref
20.
Team  RCR.  A Language and Environment for Statistical Computing. Vienna, Austria: R Foundation for Statistical Computing; 2015.
21.
Tendler  A, Helmke  S, Teruya  S, Alvarez  J, Maurer  MS.  The myocardial contraction fraction is superior to ejection fraction in predicting survival in patients with AL cardiac amyloidosis.  Amyloid. 2015;22(1):61-66.PubMedGoogle ScholarCrossref
22.
Collins  GS, Reitsma  JB, Altman  DG, Moons  KG.  Transparent Reporting of a multivariable prediction model for Individual Prognosis or Diagnosis (TRIPOD): the TRIPOD statement.  Ann Intern Med. 2015;162(1):55-63.PubMedGoogle ScholarCrossref
23.
Quarta  CC, Buxbaum  JN, Shah  AM,  et al.  The amyloidogenic V122I transthyretin variant in elderly black Americans.  N Engl J Med. 2015;372(1):21-29.PubMedGoogle ScholarCrossref
24.
Buxbaum  J, Alexander  A, Koziol  J, Tagoe  C, Fox  E, Kitzman  D.  Significance of the amyloidogenic transthyretin Val 122 Ile allele in African Americans in the Arteriosclerosis Risk in Communities (ARIC) and Cardiovascular Health (CHS) Studies.  Am Heart J. 2010;159(5):864-870.PubMedGoogle ScholarCrossref
25.
Hyung  SJ, Deroo  S, Robinson  CV.  Retinol and retinol-binding protein stabilize transthyretin via formation of retinol transport complex.  ACS Chem Biol. 2010;5(12):1137-1146.PubMedGoogle ScholarCrossref
26.
Kyle  RA, Linos  A, Beard  CM,  et al.  Incidence and natural history of primary systemic amyloidosis in Olmsted County, Minnesota, 1950 through 1989.  Blood. 1992;79(7):1817-1822.PubMedGoogle Scholar
27.
Connors  LH, Sam  F, Skinner  M,  et al.  Heart failure resulting from age-related cardiac amyloid disease associated with wild-type transthyretin: a prospective, observational cohort study.  Circulation. 2016;133(3):282-290.PubMedGoogle Scholar
Original Investigation
March 2017

Identification of Transthyretin Cardiac Amyloidosis Using Serum Retinol-Binding Protein 4 and a Clinical Prediction Model

Author Affiliations
  • 1Department of Medicine, Boston University School of Medicine, Boston Medical Center, Boston, Massachusetts
  • 2Amyloidosis Center, Boston University School of Medicine, Boston Medical Center, Boston, Massachusetts
  • 3Department of Pathology and Laboratory Medicine, Boston University School of Medicine, Boston Medical Center, Boston, Massachusetts
  • 4Department of Biostatistics, Boston University School of Public Health, Boston, Massachusetts
  • 5Hematology/Oncology Section, Medical Service, Veterans Affairs Boston Healthcare System, Boston, Massachusetts
  • 6Section of Cardiovascular Medicine, Boston University School of Medicine, Boston Medical Center, Boston, Massachusetts
JAMA Cardiol. 2017;2(3):305-313. doi:10.1001/jamacardio.2016.5864
Key Points

Question  Is retinol-binding protein 4 useful in the diagnosis of transthyretin cardiac amyloidosis?

Findings  In this case-control study that included 34 patients with transthyretin cardiac amyloidosis and 77 control participants, retinal-binding protein 4 was significantly associated with the disease independently of tested confounders. A prediction model composed of retinal-binding protein 4, transthyretin, and echocardiographic and electrocardiographic characteristics had excellent discrimination for transthyretin cardiac amyloidosis.

Meaning  A simple prediction rule using retinal-binding protein 4 concentration and readily available clinical characteristics could be a powerful tool for the diagnosis of transthyretin cardiac amyloidosis.

Abstract

Importance  Transthyretin cardiac amyloidosis (ATTR) is an underrecognized cause of heart failure (HF) in older individuals, owing in part to difficulty in diagnosis. ATTR can result from substitution of valine for isoleucine at codon 122 of the transthyretin (TTR) gene (V122I), present in 3.43% of African American individuals.

Objective  To examine whether serum retinol-binding protein 4 (RBP4), an endogenous TTR ligand, could be used as a diagnostic test for ATTR V122I amyloidosis.

Design, Setting, and Participants  In this combined prospective and retrospective cohort study performed at a tertiary care referral center, 50 African American patients 60 years or older with nonamyloid HF and cardiac wall thickening prospectively genotyped from September 1, 2014, through December 31, 2015, and a comparator cohort of 25 patients with biopsy-proven ATTR V122I amyloidosis recruited from September 1, 2009, through November 31, 2014, comprised the development cohort. Twenty-seven African American patients and 9 patients with ATTR V122I amyloidosis comprised the validation cohort.

Main Outcomes and Measures  Circulating RBP4, TTR, B-type natriuretic peptide (BNP), and troponin I (TnI) concentrations and electrocardiographic, echocardiographic, and clinical characteristics were assessed in all patients. Receiver operating characteristic (ROC) analysis was performed to identify optimal thresholds for ATTR V122I amyloidosis identification. A clinical prediction rule was developed using penalized logistic regression, evaluated using ROC analysis and validated in an independent cohort of cases and controls.

Results  Age, sex, and BNP and TnI concentrations were similar between the 25 patients with ATTR V122I amyloidosis (mean [SD] age, 72.2 [7.4] years; 18 male [72%]) and the 50 controls (mean [SD] age, 69.2 [5.7] years; 31 male [62%]). Serum RBP4 concentration was lower in patients with ATTR V122I amyloidosis compared with nonamyloid controls (31.5 vs 49.4 µg/mL, P < .001), and the difference persisted after controlling for potential confounding variables. Left ventricular ejection fraction was lower in patients with ATTR V122I amyloidosis (mean [SD], 40% [14%] vs 57% [14%], P < .001), whereas interventricular septal diameter was higher (mean [SD], 16 [3] vs 14 [2] mm, P < .001). The ROC analysis identified RBP4 as a sensitive identifier of ATTR V122I amyloidosis (area under the curve [AUC] = 0.78; 95% CI, 0.67-0.88). A clinical prediction algorithm composed of RBP4, TTR, left ventricular ejection fraction, interventricular septal diameter, mean limb lead QRS voltage, and grade 3 diastolic dysfunction yielded excellent discriminatory capacity for ATTR V122I amyloidosis (AUC = 0.97; 95% CI, 0.93-1.00), whereas a 4-parameter model, including RBP4 concentration, retained excellent discrimination (AUC = 0.92; 95% CI, 0.86-0.99). The models maintained excellent discrimination in the validation cohort.

Conclusions and Relevance  A prediction model using circulating RBP4 concentration and readily available clinical parameters accurately discriminated ATTR V122I amyloidosis from nonamyloid HF in a case-matched cohort. This clinical algorithm may be useful for identification of ATTR V122I amyloidosis in elderly African American patients with HF.

Introduction

Transthyretin cardiac amyloidosis (ATTR) results from misaggregation and myocardial deposition of wild-type (ATTRwt) or mutated (ATTRm) transthyretin (TTR) protein, increasing ventricular wall thickness, inducing heart failure (HF), and ultimately causing death.1,2 Several single amino acid substitutions have been associated with ATTRm cardiac amyloidosis. Most notably, substitution of valine for isoleucine at codon 122 of the transthyretin (TTR) gene (V122I) (OMIM 176300) constitutes one of the most prevalent familial TTR amyloidoses in the United States. The V122I mutation is present in 3.43% of African American individuals3 but generally absent in white individuals.4 ATTRm V122I amyloidosis carries significant morbidity and mortality.1,5 Unfortunately, identification generally occurs after the development of symptoms, and a large number of cases are misdiagnosed or undiagnosed.6,7 Recent advances in pharmacologic therapy for ATTR have led to the development of targeted treatments with promising results in clinical trials.8-10 Therefore, there is a pressing need for better testing to permit recognition of disease and to differentiate ATTR V122I amyloidosis from other causes of HF in at-risk patients.

Currently, definitive diagnosis of ATTR cardiomyopathy requires Congo red staining of a cardiac biopsy specimen, biochemical or immunochemical demonstration of TTR as the pathologic protein, and TTR genotyping. Although cardiac magnetic resonance imaging can identify cardiac amyloidosis,11 it does not provide conclusive information regarding amyloid disease type. Bone-avid technetium Tc 99m pyrophosphate and 3,3-diphosphono-1,2-propanodicarboxylic acid scintigraphy (DPD) techniques are rapidly emerging as tools for cardiac ATTR identification12; however, these tests require experience to perform and interpret and, although increasing in use, have not yet achieved widespread adaptation.

Normally, TTR circulates as a homotetramer complexed to retinol-binding protein 4 (RBP4), a 21-kDa protein integral to the transport of all-trans retinol (vitamin A). When bound to TTR, the RBP4-retinol complex stabilizes the TTR tetramer, inhibiting monomer dissociation and amyloid fibril formation.13 In addition, RBP4 binding to the TTR tetramer prevents glomerular filtration of low-molecular-weight RBP4, effectively increasing its circulating levels.14 It is widely held that mutations encode destabilizing amino acid replacements that induce TTR tetramer disassociation, resulting in fewer circulating RBP4-TTR complexes and increased urinary excretion of RBP4.15 Alternatively, it is conceivable that decreased plasma concentration of RBP4 could result in lower TTR binding, thereby promoting tetramer destabilization. These observations suggest that RBP4 concentration may indicate the degree of TTR misfolding and therefore amyloid fibril formation. We sought to determine the utility of circulating RBP4 as a tool to identify ATTR V122I amyloidosis in a cohort of elderly African American patients with HF. We further aimed to develop a clinical prediction model for ATTR V122I amyloidosis based on clinical, echocardiographic, and electrocardiographic characteristics, along with RBP4 and other cardiac biomarkers.

Methods
Study Design

We conducted a case-control study of self-reported African American individuals with cardiomyopathy attributable to biopsy-proven ATTR V122I amyloidosis or nonamyloid causes. The study was divided into 2 parts: the development and validation cohorts. For the development cohort, we prospectively recruited and genotyped a population of elderly African American patients with cardiomyopathy (n = 52) from November 17, 2014, through December 29, 2015. Inclusion criteria for the prospective cohort were age of 60 years or older, African American race, an International Statistical Classification of Diseases and Health-Related Problems, Ninth or Tenth Revision diagnosis of HF, and interventricular septal diameter (IVSD) of 12 mm or larger as determined by echocardiography. Those who proved to have a normal TTR genotype (n = 50) comprised the nonamyloid control cohort. Those who proved to have ATTR V122I amyloidosis (n = 2, demonstrated to be phenotype positive) were ultimately included in the validation cohort (see below). For the patients in the development cohort, patient serum samples and clinical data from 25 patients with biopsy-proven ATTR V122I amyloidosis collected from September 14, 2009, through November 10, 2014, were obtained retrospectively from the Boston University Amyloidosis Center archive and repository. Light chain amyloidosis was excluded by plasma cell dyscrasia marker analysis (serum free light chain assays, serum and urine immunofixation electrophoreses) or amyloid typing of a cardiac biopsy specimen (36% of the total ATTR cohort). Serum samples from both cohorts were tested for TTR and RBP4 concentrations; prospectively collected samples were analyzed for the TTR genotype. Height, weight, and blood pressure; medical history, including hypertension and diabetes; electrocardiographic and echocardiographic characteristics; hematocrit; and creatinine, lipid, B-type natriuretic peptide (BNP), and troponin I (TnI) levels were obtained from the clinical medical record. For the validation cohort, we prospectively recruited an additional 27 elderly African American patients with nonamyloid cardiomyopathy from December 30, 2015, through August 29, 2016 (using the same enrollment criteria as the initial part of the study) and identified an additional patient with ATTR V122I amyloidosis (total of 3 prospectively identified patients with ATTR V122I amyloidosis, 2 from the initial study period and 1 during the validation study recruitment). To enrich the validation cohort with additional ATTR V122I amyloidosis cases, we included 6 additional patients with biopsy-proven ATTR V122I amyloidosis seen at the Boston University Amyloidosis Center from July 15, 2014, through July 11, 2016. The total validation group was composed of 9 patients with ATTR V122I amyloidosis and 27 controls (eMaterial and eFigure 1 in the Supplement). The study was approved by the Boston University Medical Campus Institutional Review Board. All patients provided written consent before inclusion in the study, and data were deidentified.

Serum Protein Quantification and Identification of TTR

Blood samples were obtained at the initial visit to our clinic and stored at −80°C until required for analyses. The RBP4 concentrations were determined from each sample via commercially available enzyme-linked immunosorbent assays (R&D Systems). Testing was performed in triplicate and mean values reported. Precision of RBP4 concentration measurements was reflected in the intra-assay (5.7%-8.1%) and interassay (5.8%-8.6%) precision coefficients of variation. The TTR concentrations were determined in the clinical laboratory of the Pathology Department of Boston Medical Center by immunoturbidity assay (Abbott Laboratories). Determination of the TTR genotype was performed at the Boston University Amyloidosis Center via a 2-step process that involved serum screening for a TTR mutant protein by isoelectric focusing as previously described5 and identification of the TTR mutation by direct nucleotide sequencing of amplified genomic DNA on samples where a mutant was detected. Polymerase chain reaction sequencing of all 4 exons was performed for completeness to identify multiple mutations, as opposed to targeted polymerase chain reaction for exon 4 that contained the V122I polymorphism. The patients identified as V122I carriers from the prospective analysis were subsequently referred to the Boston University Amyloidosis Center for confirmation of cardiac amyloid phenotype by advanced imaging (cardiac magnetic resonance imaging or technetium Tc 99m pyrophosphate scintigraphy) and/or by endomyocardial biopsy. In the prospective arm, technicians (G.G.C., C.T.-A.) and investigators (M.A., C.M.K., M.P.V., D.R.J., J.L.B., L.H.C., F.L.R.) analyzing the samples for biomarkers and performing medical record review of clinical, echocardiographic, and electrocardiographic characteristics were masked to the results of genotyping for V122I during data collection.

Statistical Analysis

The Mann-Whitney Wilcoxon test or the 2-tailed, unpaired t test for continuous variables (depending on the distribution of data) and χ2 tests for categorical variables were used in pairwise comparisons between participants with ATTR V122I amyloidosis and controls with nonamyloid HF. Associations between continuous variables were assessed using the Spearman correlations. To assess whether RBP4 was independently associated with ATTR V122I amyloidosis, logistic regression was performed while controlling for potential confounders, including age, sex, body mass index (BMI) (calculated as weight in kilograms divided by height in meters squared), history of diabetes, cardiac biomarkers, and echocardiographic characteristics. The ability of RBP4 to discriminate between ATTR V122I amyloidosis and nonamyloid cardiomyopathy was assessed by receiver operating characteristic (ROC) analysis and calculation of the sensitivity and specificity of different RBP4 thresholds for the diagnosis of ATTR V122I amyloidosis by the ROC thresholds.16

For the development of an assessment tool that maximized the diagnostic potential of noninvasive characteristics, multivariable logistic regression models were created using independent variables preselected as having a potentially useful diagnostic role based on a prior study5 on ATTR V122I amyloidosis and on biological data. For 37 patients who had missing values in at least one of the predictor variables, multiple imputation was performed using predictive mean matching to fill in the missing values.17 Least absolute shrinkage and selection operator (LASSO) regression was used to select the predictors of ATTR V122I amyloidosis that provided the optimal diagnostic value while minimizing overfitting.18 The LASSO regression shrinkage parameter was determined using 10-fold cross-validation within each imputed data set. Details regarding the logistic regression analysis for the prediction rule for ATTR V122I amyloidosis can be found in the eMaterial in the Supplement. The goodness of fit of the LASSO regression prediction rule to the patient data was estimated by the Hosmer-Lemeshow test19 and by drawing calibration plots. To visualize the results of the model and its diagnostic potential, we plotted the ROC curve from the predicted probabilities from the rule and the ATTR V122I amyloidosis diagnosis. The prediction rule was subsequently tested on the validation cohort to assess its discrimination and goodness of fit. R statistical software, version 2.3.220 was used in all statistical analyses. Multiple imputation with predictive mean matching was performed using the mice R package,17 whereas LASSO regression was performed with the glmnet package18 and the ROC analysis with the pROC package.16 Statistical significance was set at P < .05 for all comparisons.

Results
Baseline Patient Characteristics

Of the 75 patients in the development cohort, 25 had ATTR V122I amyloidosis (mean [SD] age, 72.2 [7.4] years; 18 men [72%]), and 50 had nonamyloid HF (mean [SD] age, 69.2 [5.7] years; 31 men [62%]). Of the 52 patients enrolled in the prospective portion of the development data set, 2 were found to have ATTR V122I amyloidosis, amounting to a 3.8% calculated prevalence of the disease, as previously reported.3 Mean values of different clinical, biomarker, and echocardiographic characteristics at study entry for patients with ATTR V122I amyloidosis and nonamyloid HF are listed in Table 1. Age and sex did not differ significantly between the groups. Patients with ATTR V122I amyloidosis had significantly lower BMI (mean [SD], 29.2 [6.1] vs 34.4 [9.0]; P = .006) and lower systolic blood pressure (mean [SD], 117 [18] mm Hg vs 138 [20] mm Hg; P < .001) at the time of enrollment, whereas diastolic blood pressure did not differ significantly (mean [SD], 76 [9] mm Hg vs 77 [11] mm Hg; P = .80). These differences persisted in models adjusting for age and sex.

Patients with ATTR V122I amyloidosis had significantly lower levels of RBP4 (mean [SD], 31.47 [9.09] µg/mL vs 49.36 [20.59] µg/mL; P < .001), TTR (mean [SD], 16.47 [5.51] mg/dL vs 23.98 [6.85] mg/dL; P < .001), and BNP (mean [SD], 897 [657] ng/L vs 1028 [2706] ng/L; P = .004) and higher levels of TnI (mean [SD], 0.274 [0.251] µg/mL vs 0.133 [0.373] µg/mL; P < .001) and hematocrit (mean [SD], 40.5% [6.6%] vs 37.1% [6.0%]; P = .02 [a complete list of conversion to SI units for these analytes is given in the footnotes to Table 1]) compared with nonamyloid controls (Figure 1). The 2 groups did not differ in the levels of hemoglobin A1c, triglycerides, total cholesterol, or serum creatinine. Moreover, the ratio of RBP4 to TTR was not different between the 2 groups (mean [SD], 1.9 [0.5] vs 2.1 [0.7]; P = .30). In models that adjusted for age and sex, the differences in RBP4 and TTR remained statistically significant, whereas BNP and TnI were not found to be significantly different.

Patients with ATTR V122I amyloidosis had higher myocardial wall thickness as measured by IVSD (mean [SD], 16 [3] mm vs 14 [2] mm; P < .001) and lower left ventricular ejection fraction (LVEF) (mean [SD], 40% [14%] vs 57% [14%]; P < .001) compared with controls. Although present, these differences are well within the expected spectrum of wall thickness and systolic function and clinically insufficient to differentiate the 2 populations. The 2 populations did not differ in left ventricular mass index (mean [SD], 167.7 [42.7] g/m vs 158.0 [52.8] g/m; P = .20). Of note, the myocardial contraction fraction21 was significantly lower in patients with ATTR V122I amyloidosis compared with controls (mean [SD], 23% [13%] vs 49% [13%]; P < .001). At electrocardiography, the heart rate was higher in patients with ATTR V122I amyloidosis (mean [SD], 78/min [11/min] vs 70/min [15/min]; P = .003), whereas the mean limb lead QRS voltage (mean [SD], 5 [2] mm vs 8 [3] mm; P < .001) and QRS voltage to left ventricular mass ratio (mean [SD], 0.02 [0.01] vs 0.04 [0.02]; P < .001) were lower. All these findings persisted in age- and sex-adjusted models (Table 2).

RBP4 Concentration and ATTR V122I Cardiac Amyloidosis Identification

Because the RBP4 concentration was significantly lower in patients with ATTR V122I amyloidosis, we performed multivariable logistic regression analysis to assess whether this association was affected by confounding variables. The RBP4 concentration remained significantly different when controlling for age, sex, BMI, history of diabetes, creatinine level, nutritional status (albumin, modified BMI), hematocrit, cardiac biomarkers, and echocardiographic characteristics. Not surprisingly, RBP4 and TTR levels were highly correlated (Spearman ρ = 0.75, P < .001), and consequently, the association between RBP4 and ATTR V122I amyloidosis was not statistically significant when both RBP4 and TTR were included in the same logistic regression model.

The ROC analysis identified RBP4 as a potentially sensitive diagnostic test for ATTR V122I amyloidosis cardiomyopathy (area under the curve [AUC] = 0.78; 95% CI, 0.67-0.88) (Figure 2A). A threshold of 43 µg/mL maximized the Youden index, providing a sensitivity of 62% (95% CI, 47%-77%) and a specificity of 88% (95% CI, 76%-100%) for ATTR V122I amyloidosis. For screening, a cutoff value of 50 µg/mL achieved perfect sensitivity (100%; 95% CI, 100%-100%) but low specificity (38%; 95% CI, 23%-51%) for ATTR V122I amyloidosis. This threshold also yielded a negative predictive value of 100% in this total population with 33% prevalence of the disease. Of importance, 38% of the nonamyloid control cohort had RBP4 values above this threshold.

Diagnostic Model for ATTR V122I

To improve diagnostic performance, we developed a model for ATTR V122I identification by using several predetermined clinical factors, circulating biomarkers, and echocardiographic characteristics. In addition to RBP4 and TTR concentrations, we selected 6 variables available to clinicians by standard-of-care testing, including LVEF, IVSD, presence of grade 3 diastolic dysfunction, and mean limb lead QRS voltage, which significantly improved the final prediction rule. All other clinical characteristics (age, sex), circulating biomarkers (hemoglobin A1c, triglycerides, total cholesterol, BNP, and TnI), and other electrocardiographic characteristics did not add significant diagnostic value and were excluded from the final model. Of note, our diagnostic model development is compliant with the TRIPOD statement for transparent reporting22 (eTable in the Supplement).

The following prediction rule was developed as a result of this statistical modeling:

Linear Predictor = {0.315 - [0.010 × RBP4 (micrograms/milliliter)] - [0.022 × LVEF (percentage)] + [0.097 × IVSD (millimeters)] - [0.040 × TTR (milligrams per deciliter)] - [0.129 × Mean Limb Lead QRS Voltage (millivolts)] + 1.065 (If Grade 3 Diastolic Dysfunction Is Present)}

The predicted probability of ATTR V122I amyloidosis based on our model can be calculated as:

Probability = 1/(1+e-(Linear Predictor)).

Figure 2B shows the ROC curve for the prediction model. The ROC analysis identified the diagnostic model as an efficient discriminator of ATTR V122I with an AUC of 0.97 (95% CI, 0.93-1.00). Of note, the AUC of the model did not vary significantly among the different imputation data sets (range, 0.96-0.98). When applied to our original data set, the prediction rule did not demonstrate any statistical evidence of poor fit (Hosmer-Lemeshow statistic = 10.808, P = .21) (eFigure 2A in the Supplement). Of note, the median probability of controls not having ATTR V1221 based on the prediction model was 11.2% (interquartile range, 7.9%-15.6%), whereas the median probability of patients having ATTR V122I amyloidosis was 48.3% (interquartile range, 38.4%-68.9%).

Sensitivity Analysis

Recognizing the structural association of TTR and RBP4, we performed a sensitivity analysis using the same statistical method, excluding from the model TTR and the presence of grade 3 diastolic dysfunction because of its relative subjectivity (eMaterial in the Supplement). The resulting 4-variable model that incorporated only RBP4, LVEF, IVSD, and mean limb lead QRS voltage was as follows:

Linear Predictor = {1.992 - [0.034 × RBP4 (micrograms/milliliter)] - [0.056 × LVEF (percentage)] + [0.220 × IVSD (millimeters)] - [0.272 × Mean Limb Lead QRS Voltage (millivolts)]}

This final model maintained excellent discriminatory capacity for ATTR V122I amyloidosis with a ROC curve AUC of 0.92 (95% CI, 0.86-0.99) (Figure 2C) and no evidence of poor fit to our original data set (Hosmer-Lemeshow statistic = 7.81, P = .45) (eFigure 2B in the Supplement).

Validation

Both the 6-parameter model and the parsimonious model were evaluated for their performance in the validation cohort. The discrimination of both models remained excellent in the validation set (AUC = 0.93 [95% CI, 0.83-1.00] for the full model and 0.88 [95% CI, 0.74-1.00] for the parsimonious model) (Figure 2D and E), whereas none of the models had any evidence of poor fit to the replication cohort (full model: Hosmer-Lemeshow statistic = 3.52, P = .89; parsimonious model: Hosmer-Lemeshow statistic = 4.79, P = .78) (eFigure 3A and B in the Supplement). Of importance, one of the patients in the validation group was identified as being V122I TTR genotype positive but had no evidence of plasma cell dyscrasia or amyloid at technitium Tc 99m pyrophosphate scinitigraphy and was thus presumed to be genotype positive and amyloid phenotype negative. This individual’s probability of ATTR V122I amyloidosis by our predictive models was low (10.4% by the full model and 11.6% by the parsimonious model), providing further confidence in the validity of our predictive model. Of note, in aggregate among the development and validation cohorts, 32 patients (97%) with ATTR V122 amyloidosis and 7 controls (9%) had predicted probability greater than 25% based on the full model, suggesting that a probability threshold of 25% using this model for any given patient could be used to trigger further diagnostic testing for ATTR V122I amyloidosis.

Discussion

In elderly African American individuals, ATTR V122I amyloidosis is an underrecognized cause of HF owing to the complexities of diagnosis and the prevalence of hypertensive heart disease in the target cohort. To facilitate identification and testing of a population at risk for ATTR V122I amyloidosis, we developed a novel mathematical prediction of disease based on LVEF, RBP4 concentration, IVSD, and mean limb lead QRS voltage, yielding a likelihood of ATTR V122I amyloidosis in African American individuals 60 years or older with HF. We submitted this model to rigorous statistical testing to confirm its quality of fit, demonstrating excellent discriminative capacity for ATTR V122I amyloidosis identification (AUC>0.9), which was subsequently confirmed in an independent validation cohort. The model may therefore be useful to guide clinicians as to whether a patient who fits the at-risk population will require further diagnostic testing, such as technetium Tc 99m pyrophosphate or 3,3-diphosphono-1,2-propanodicarboxylic acid scintigraphy12 or heart biopsy. Of importance, when considered individually, the specific measures used in the model proved to be insufficient to identify ATTR V122I amyloidosis.

The demographic characteristics and cardiac measures among our study cohort were consistent with those in previous reports of patients with ATTR V122I amyloidosis. Restrictive filling (grade 3 diastolic dysfunction) was higher and mean limb lead QRS voltage was lower in patients with ATTR V122I amyloidosis. Of note, in our study, patients with ATTR V122I amyloidosis had higher TnI but lower BNP concentrations compared with patients with nonamyloid HF, but these results did not persist after controlling for age and sex. A previous study1 found that BNP and TnI concentrations tend to be higher in patients with ATTR V122I amyloidosis, but differences in BNP were not found in another report.5 Although most recruited control patients were identified during a clinically stable outpatient visit, it is possible that the cardiac biomarkers in our control population could have been skewed toward higher values because this information was collected from medical record review and would thereby be biased by the clinical indication for collection of these biomarkers in this group. Of interest, our results indicate that individuals with ATTR V122I amyloidosis had lower LVEF and greater wall thickness but did not differ in left ventricular mass index. This finding agrees with those of some prior reports1,5 of elderly patients with ATTR V122I amyloidosis but is contrary to the findings of the Atherosclerosis Risk in Communities (ARIC) study.23 This discrepancy could reflect the fact that the ARIC study included carriers of the V122I mutation without limiting the population to patients with confirmed disease, skewing the echocardiographic indicators toward the general control population. Indeed, the median age of patients from the ARIC study at inclusion was 52 years, with the echocardiographic measurements being obtained at visit 5 (approximately 25 years later). A previous report24 indicates that before the age of 65 years, there is no discernible effect of the V122I mutation in cardiac function. Similarly, in another cohort of elderly patients, the LVEF decreased by a mean of 3.2% in every 6-month interval of follow-up.1

RBP4 was significantly associated with ATTR amyloidosis cardiomyopathy independent of potentially confounding variables. To our knowledge, this association has not been previously reported in the clinical setting, but in vitro disease models have found a potential role of RBP4 in stabilizing TTR and preventing misfolding.25 We found that an RBP4 concentration greater than or equal to 50 µg/mL had 100% sensitivity (but modest specificity) and provided 100% negative predictive value in excluding ATTR V122I amyloidosis without the need for further testing (similar in the way that d-dimer concentration is used to exclude venous thromboembolism), suggesting a potential utility of RBP4 concentration as an initial screening test.

Limitations

Limitations of our study include the relatively small number of patients, although the number of patients with ATTR V122I amyloidosis in the context of worldwide prevalence estimates is reasonably sized. Because the prediction rule was developed in elderly African American individuals, we cannot generalize our findings to the general population of patients with HF, to younger populations, or to ATTRm from other amyloidogenic mutations. The results presented here are promising but preliminary. Although the prediction rule performance estimates remained excellent in a small replication cohort, external validation in a larger, independent patient population will be necessary before widespread clinical adaptation. Finally, although we definitively established the presence or absence of ATTR V122I amyloidosis in our study, a small possibility exists that our control group included a case of cardiac amyloidosis not attributable to ATTR V122I (ie, light chain amyloidosis or ATTRwt). The proportion of these individuals is expected to be small, if present at all, given that the known annual incidence of light chain cardiac amyloidosis is extremely low (approximately 1 case per 100 000 US population),26 and in a large recent cohort of ATTRwt, more than 90% of patients with ATTRwt were white males and not African American.27 Furthermore, even if a few of our controls had some other form of cardiac amyloid, this would not be expected to lead to an overestimate of our predictive model performance but would rather decrease the specificity (and subsequently the AUC) of our model by increasing the number of false-positive results.

Conclusions

We found a clinical prediction model based on RBP4 concentration and available standard-of-care clinical variables that has the potential to identify ATTR V122I amyloidosis and thereby trigger confirmatory testing. The test may have clinical utility as a first-line screening test for the disease in elderly African American individuals with HF, guiding clinical decision making. Furthermore, RBP4 concentration alone or in the context of this model may be useful to follow disease progression. Prospective validation of our prediction model in larger clinical cohorts is needed.

Back to top
Article Information

Corresponding Author: Frederick L. Ruberg, MD, Section of Cardiovascular Medicine, Boston University School of Medicine, Boston Medical Center, 88 E Newton St, C-818, Boston, MA 02118 (frruberg@bu.edu).

Accepted for Publication: December 12, 2016.

Published Online: February 8, 2017. doi:10.1001/jamacardio.2016.5864

Author Contributions: Drs Arvanitis and Ruberg had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Arvanitis, Koch, Jacobson, Berk, Connors, Ruberg.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Arvanitis, Jacobson, Ruberg.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Arvanitis, LaValley.

Obtained funding: Ruberg.

Administrative, technical, or material support: Koch, Chan, Torres-Arancivia, Connors, Ruberg.

Study supervision: Connors, Ruberg.

Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest and none were reported.

Funding/Support: This study was supported by grants R21AG050206 (Dr Ruberg), R01AG031804 (Dr Connors), and 1U54TR001012 (Dr Ruberg) from the National Institutes of Health and the Young Family Amyloid Research Fund.

Role of the Funder/Sponsor: The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

References
1.
Ruberg  FL, Maurer  MS, Judge  DP,  et al.  Prospective evaluation of the morbidity and mortality of wild-type and V122I mutant transthyretin amyloid cardiomyopathy: the Transthyretin Amyloidosis Cardiac Study (TRACS).  Am Heart J. 2012;164(2):222-228.e1.PubMedGoogle ScholarCrossref
2.
Jacobson  DR, Pastore  RD, Yaghoubian  R,  et al.  Variant-sequence transthyretin (isoleucine 122) in late-onset cardiac amyloidosis in black Americans.  N Engl J Med. 1997;336(7):466-473.PubMedGoogle ScholarCrossref
3.
Jacobson  DR, Alexander  AA, Tagoe  C, Buxbaum  JN.  Prevalence of the amyloidogenic transthyretin (TTR) V122I allele in 14 333 African-Americans.  Amyloid. 2015;22(3):171-174.PubMedGoogle ScholarCrossref
4.
Yamashita  T, Hamidi Asl  K, Yazaki  M, Benson  MD.  A prospective evaluation of the transthyretin Ile122 allele frequency in an African-American population.  Amyloid. 2005;12(2):127-130.PubMedGoogle ScholarCrossref
5.
Connors  LH, Prokaeva  T, Lim  A,  et al.  Cardiac amyloidosis in African Americans: comparison of clinical and laboratory features of transthyretin V122I amyloidosis and immunoglobulin light chain amyloidosis.  Am Heart J. 2009;158(4):607-614.PubMedGoogle ScholarCrossref
6.
Damy  T, Costes  B, Hagège  AA,  et al.  Prevalence and clinical phenotype of hereditary transthyretin amyloid cardiomyopathy in patients with increased left ventricular wall thickness.  Eur Heart J. 2016;37(23):1826-1834.PubMedGoogle ScholarCrossref
7.
Ruberg  FL, Berk  JL.  Transthyretin (TTR) cardiac amyloidosis.  Circulation. 2012;126(10):1286-1300.PubMedGoogle ScholarCrossref
8.
Berk  JL, Suhr  OB, Obici  L,  et al; Diflunisal Trial Consortium.  Repurposing diflunisal for familial amyloid polyneuropathy: a randomized clinical trial.  JAMA. 2013;310(24):2658-2667.PubMedGoogle ScholarCrossref
9.
Maurer  MS, Grogan  DR, Judge  DP,  et al.  Tafamidis in transthyretin amyloid cardiomyopathy: effects on transthyretin stabilization and clinical outcomes.  Circ Heart Fail. 2015;8(3):519-526.PubMedGoogle ScholarCrossref
10.
Coelho  T, Adams  D, Silva  A,  et al.  Safety and efficacy of RNAi therapy for transthyretin amyloidosis.  N Engl J Med. 2013;369(9):819-829.PubMedGoogle ScholarCrossref
11.
Fontana  M, Pica  S, Reant  P,  et al.  Prognostic value of late gadolinium enhancement cardiovascular magnetic resonance in cardiac amyloidosis.  Circulation. 2015;132(16):1570-1579.PubMedGoogle ScholarCrossref
12.
Gillmore  JD, Maurer  MS, Falk  RH,  et al.  Nonbiopsy diagnosis of cardiac transthyretin amyloidosis.  Circulation. 2016;133(24):2404-2412.PubMedGoogle ScholarCrossref
13.
Raghu  P, Sivakumar  B.  Interactions amongst plasma retinol-binding protein, transthyretin and their ligands: implications in vitamin A homeostasis and transthyretin amyloidosis.  Biochim Biophys Acta. 2004;1703(1):1-9.PubMedGoogle ScholarCrossref
14.
Wei  S, Episkopou  V, Piantedosi  R,  et al.  Studies on the metabolism of retinol and retinol-binding protein in transthyretin-deficient mice produced by homologous recombination.  J Biol Chem. 1995;270(2):866-870.PubMedGoogle ScholarCrossref
15.
Berni  R, Malpeli  G, Folli  C, Murrell  JR, Liepnieks  JJ, Benson  MD.  The Ile-84—>Ser amino acid substitution in transthyretin interferes with the interaction with plasma retinol-binding protein.  J Biol Chem. 1994;269(38):23395-23398.PubMedGoogle Scholar
16.
Robin  X, Turck  N, Hainard  A,  et al.  pROC: an open-source package for R and S+ to analyze and compare ROC curves.  BMC Bioinformatics. 2011;12:77.PubMedGoogle ScholarCrossref
17.
Zhang  Z.  Multiple imputation with multivariate imputation by chained equation (MICE) package.  Ann Transl Med. 2016;4(2):30.PubMedGoogle Scholar
18.
Friedman  J, Hastie  T, Tibshirani  R.  Regularization paths for generalized linear models via coordinate descent.  J Stat Softw. 2010;33(1):1-22.PubMedGoogle ScholarCrossref
19.
Hosmer  DW, Hosmer  T, Le Cessie  S, Lemeshow  S.  A comparison of goodness-of-fit tests for the logistic regression model.  Stat Med. 1997;16(9):965-980.PubMedGoogle ScholarCrossref
20.
Team  RCR.  A Language and Environment for Statistical Computing. Vienna, Austria: R Foundation for Statistical Computing; 2015.
21.
Tendler  A, Helmke  S, Teruya  S, Alvarez  J, Maurer  MS.  The myocardial contraction fraction is superior to ejection fraction in predicting survival in patients with AL cardiac amyloidosis.  Amyloid. 2015;22(1):61-66.PubMedGoogle ScholarCrossref
22.
Collins  GS, Reitsma  JB, Altman  DG, Moons  KG.  Transparent Reporting of a multivariable prediction model for Individual Prognosis or Diagnosis (TRIPOD): the TRIPOD statement.  Ann Intern Med. 2015;162(1):55-63.PubMedGoogle ScholarCrossref
23.
Quarta  CC, Buxbaum  JN, Shah  AM,  et al.  The amyloidogenic V122I transthyretin variant in elderly black Americans.  N Engl J Med. 2015;372(1):21-29.PubMedGoogle ScholarCrossref
24.
Buxbaum  J, Alexander  A, Koziol  J, Tagoe  C, Fox  E, Kitzman  D.  Significance of the amyloidogenic transthyretin Val 122 Ile allele in African Americans in the Arteriosclerosis Risk in Communities (ARIC) and Cardiovascular Health (CHS) Studies.  Am Heart J. 2010;159(5):864-870.PubMedGoogle ScholarCrossref
25.
Hyung  SJ, Deroo  S, Robinson  CV.  Retinol and retinol-binding protein stabilize transthyretin via formation of retinol transport complex.  ACS Chem Biol. 2010;5(12):1137-1146.PubMedGoogle ScholarCrossref
26.
Kyle  RA, Linos  A, Beard  CM,  et al.  Incidence and natural history of primary systemic amyloidosis in Olmsted County, Minnesota, 1950 through 1989.  Blood. 1992;79(7):1817-1822.PubMedGoogle Scholar
27.
Connors  LH, Sam  F, Skinner  M,  et al.  Heart failure resulting from age-related cardiac amyloid disease associated with wild-type transthyretin: a prospective, observational cohort study.  Circulation. 2016;133(3):282-290.PubMedGoogle Scholar
×