Association of Biomarker and Physiologic Indices With Mortality in Older Adults: Cardiovascular Health Study

Association of Biomarker and Physiologic Indices With Mortality in Older Adults: Cardiovascular... Abstract Background A goal of gerontology is discovering aging phenotypes that reflect biological aging distinct from disease pathogenesis. Biomarkers that strongly and independently associated with mortality and that statistically attenuated chronologic age could be used to define such a phenotype. We determined the association of a Biomarker Index (BI) with mortality and compared it with a validated Physiologic Index (PI) in older adults. Methods The indices were constructed in the Cardiovascular Health Study, mean (SD) age 74.5 (5.1) years. The BI incorporated circulating levels of new biomarkers, including insulin-like growth factor-1, insulin-like growth factor–binding protein 3, amino-terminal pro-B-type natriuretic peptide, dehydroepiandrosterone sulfate, and interleukin-6, and was built in test (N = 2,197) and validation (N = 1,124) samples. The PI included carotid intima–media thickness, pulmonary capacity, brain white matter grade, cystatin-C, and fasting glucose. Multivariable Cox proportional hazards models predicting death were calculated with 10 years of follow-up. Results In separate age-adjusted models, the hazard ratio for mortality per point of the BI was 1.30 (95% confidence interval 1.25, 1.34) and the BI attenuated age by 25%. The hazard ratio for the PI was 1.28 (1.24, 1.33; 29% age attenuation). In the same model, the hazard ratio for the BI was 1.23 (1.18, 1.28) and for the PI was 1.22 (1.17, 1.26), and age was attenuated 42.5%. Associations persisted after further adjustment. Conclusions The BI and PI were significantly and independently associated with mortality. Both attenuated the age effect on mortality substantially. The indices may be feasible phenotypes for developing interventions hoping to alter the trajectory of aging. Biomarker, Mortality, Phenotype, Longevity A central goal of gerontology is to discover aging phenotypes that reflect biological aging distinct from disease pathogenesis. Healthy aging phenotypes could serve as end points for developing interventions that alter the trajectory of aging toward a more desired pattern. Due to wider distribution and possibly incorporating multiple biological processes, composite scores have the potential for improved capability to stratify mortality risk (1). In 2008, Newman and colleagues published the Physiologic Index of Comorbidity in the Cardiovascular Health Study (CHS) (2). The Physiologic Index combined measurements including carotid intima–media thickness, pulmonary vital capacity, serum cystatin-C, brain white matter grade, and fasting glucose. In the CHS (mean age 74.5 years), only 1.7% of older adults had no evidence of disease on these five tests (2). Their 9-year mortality rate was very low at 7/1,000 person-years (2), frailty was rare (3), and they had longer leukocyte telomere length (4), demonstrating that the tests identify individuals who are exceptional. Furthermore, the Physiologic Index explained 40% of the effect of age on mortality; thus, the Physiologic Index captures aspects of biological aging for which chronologic age is a surrogate. Although a valuable phenotype, measurement of several key components of the Physiologic Index is resource intensive and costly. An index composed solely of blood-based biomarkers readily measurable in clinical laboratories would have lower expense and greater applicability. We proposed criteria to validate biomarkers of aging in epidemiologic studies (5) that are distinct from definitions of biomarkers of aging derived from laboratory science (6). The criteria include (a) demonstrating biological plausibility that a biomarker describes a basic aging process; (b) demonstrating potential for clinical applicability, using highly reproducible, highly interindividually variable, and low-cost measurements; (c) subsequently assessing associations with important aging outcomes using optimal epidemiologic study designs; and (d) testing these associations in accordance with key statistical considerations. An ideal biomarker of human aging should meet these criteria. In this study, we developed a Biomarker Index of aging using only blood-based biomarkers demonstrated to have a role in the biology of aging in laboratory and epidemiologic studies. We compared the associations of the Biomarker Index and the Physiologic Index with 10-year mortality, and the ability of each Index to explain the association of age with mortality. Methods Study Population The CHS is an ongoing community-based study of cardiovascular risk in 5,888 men and women over the age of 65 years, from four regions of the United States (7,8). The cohort was enrolled in 1989–1990 and was supplemented with added recruitment of African Americans in 1992–1993. Participants and eligible household members were identified from Medicare eligibility lists. To be eligible, participants should be 65 years and older, should not have cancer under active treatment, should not be wheelchair bound or bed bound in the home, and should not plan to move out of the area within 3 years. We used data from the 1992–1993 examination as baseline to include the supplemental African American cohort and because the brain magnetic resonance imaging scan was conducted at that time. The CHS is approved by the institutional review boards of all participating institutions, and all participants gave informed consent. Physiologic Index of Comorbidity The Physiologic Index was calculated as previously described from continuous measures of age-related dysfunction in five major organ systems, each of which predicts mortality and disability across a large continuum of risk (2). Detailed methods are provided in Supplementary Material. The systems included were the vascular (carotid intima–media thickness on ultrasound), neurologic (white matter grade on magnetic resonance imaging), renal (cystatin-C), pulmonary (forced vital capacity), and metabolic (fasting glucose). To construct the Physiologic Index, tertiles were considered for four of the five measures, with the best values classified as 0 and the worst as 2; fasting glucose was scored according to clinical cutoffs. Although the choice of cut points was arbitrary, scores of “0” generally included normal values in healthy young individuals, and values of “2” were in the range of those with chronic disease. Individual scores were summed for a total score ranging from 0 (healthy) to 10 (unhealthy). Candidate Biomarkers Candidate biomarkers for inclusion in the Biomarker Index were chosen for their documented association with aging-related processes in laboratory studies and outcomes in epidemiologic studies, though debate remains on their individual role as causative or correlative factors with aging. Candidates included insulin-like growth factor (IGF)-1, IGF-binding protein (IGFBP)-3, adiponectin, interleukin-6 (IL-6), dehydroepiandrosterone sulfate (DHEAS), and amino-terminal pro-B-type natriuretic peptide (NT-proBNP). Fasting blood samples were collected at the 1992–1993 exam using standardized protocols and quality assurance (8,9). IGF-1 and IGFBP-3 were measured after an extraction step using enzyme-linked immunosorbent assays (ELISA; Diagnostics Systems Laboratory, Webster, TX) (10). The analytic coefficient of variation (CV) was 4%–6% for IGF-1 and 3%–5% for IGFBP-3. Adiponectin was measured with an ELISA (R&D Systems, Minneapolis, MN); intra- and interassay CVs were 2.5%–4.7% and 5.8%–6.9%, respectively. IL-6 was measured by ultrasensitive ELISA (Quantikine HS Human IL-6 Immunoassay; R&D Systems); intra- and interassay CV were 2.9%–8.7% and 7.3%–9.0%, respectively. Plasma DHEAS was measured with a competitive immunoassay (Alpco Diagnostics, Windham, NH) with interassay CV of 3.8%–7.2%. NT-proBNP was measured on the Elecsys 2010 system (Roche Diagnostics, Indianapolis, IN) with a CV of 2%–5%. Mortality Deaths were ascertained through participant surveillance every 6 months from study inception. Confirmation of deaths was conducted through reviews of obituaries, medical records, death certificates, the Center for Medicare Studies health care utilization database, and the National Death Index. Contacts and proxies were also interviewed for participants unavailable for follow-up. Ascertainment of vital status was 100% complete. Potential Confounders Age, sex, race, smoking status, alcohol consumption, and diet were determined by self-report (8). The Alternative Healthy Eating Score was used to account for dietary patterns (11). Anthropometrics (weight, height, waist circumference) were measured with standard protocols, and weight and height were used to calculate body mass index in kilogram per square meter. Physical activity was based on the Modified Minnesota Leisure Time Activities questionnaire that assessed frequency and duration of 18 activities in the prior week to calculate kilocalories of energy expended. Prevalent and incident cardiovascular disease (CVD) outcomes were adjudicated by an expert panel using medical records; medication information and prevalent disease status were updated at each examination (12). Incident CVD was defined as a first occurrence of stroke, coronary heart disease, congestive heart failure, or claudication after the 1992–1993 examination among participants without a history of any component at the time of the exam. Cancer and pneumonia were defined using ICD-9 codes. Statistical Analysis CHS participants were included in the development of the Biomarker Index if they had data on all six biomarkers (N = 3,321). We derived the Biomarker Index in a randomly selected test sample of two third of the cohort (N = 2,197) and validated it in the remainder of participants (N = 1,124). In the test sample, the association of each biomarker with mortality was visualized using cubic splines. Because biomarker levels differed by gender and several associations between biomarkers and mortality were not monotonic, cut points to score each biomarker were made using sex-specific quintiles (Tables 1 and 2). Multivariable Cox proportional hazards models including age and all of the biomarkers (as indicator variables for quintiles) were run for men (N = 862) and women (N = 1,335) separately in the test sample. Only biomarkers that were individually associated with mortality at a significance level of less than or equal to .05 were included in the final Biomarker Index. Adiponectin was not associated with mortality for either men or women and IGF-1 and DHEAS were not significantly associated with mortality in men so were not included in the final models. The coefficients from the multivariable biomarker models were used to compute points for each level of each biomarker, selecting the quintile with lowest risk as the reference group. Each beta coefficient was divided by three times the coefficient for age for scaling and rounded to produce a point score. Although 5 years of age has been used in other risk scores, notably the Framingham Risk Score, 3 years of age was selected as a reasonable comparison given the older ages of our cohort. The Biomarker Index was tested in the validation sample and, when found consistent with the results in the test set, the validation and test samples were combined to derive estimates of risk associated with both indices in the full sample. Table 1. Description of Cardiovascular Health Study Cohort Characteristic  Women  Men  All  p Value  N = 2,021  N = 1,300  N = 3,321  Age, y  74.4 (5.0)  74.7 (5.2)  74.5 (5.1)  .05  Black race  310 (15.3)  158 (12.2)  468 (14.1)  .01  Current smoker  190 (9.4)  117 (9.0)  307 (9.2)  .70  Drinks alcohol  801 (39.6)  711 (54.7)  1,512 (45.5)  <.001  Physical activity, kcals  473 (150–1,110)  907 (907–1,967)  598 (187–1,417)  <.001  BMI, kg/m2  26.9 (5.2)  26.6 (3.8)  26.8 (4.7)  .07  Waist, cm  95.9 (14.6)  98.9 (10.5)  97.1 (13.2)  <.001  History of CHDa  339 (16.8)  359 (27.6)  698 (21.0)  <.001  History of CHF  91 (4.5)  66 (5.1)  157 (4.7)  .45  History of stroke  67 (3.3)  93 (7.2)  160 (4.8)  <.001  Adiponectin, mg/L  14.1 (10.2–19.6)  9.6 (7.0–13.6)  12.2 (8.5–17.5)  <.001  NT-proBNP, pg/mL  143 (76–257)  128 (60–306)  137 (69–268)  .11  DHEAS, µg/mL  0.52 (0.32–0.80)  0.79 (0.48–1.18)  0.61 (0.38–0.94)  <.001  IL-6, pg/mL  2.58 (1.75–3.91)  2.87 (2.02–4.34)  2.70 (1.85–4.08)  <.001  IGF-1, µg/L  89 (70–112.0)  103 (82–126)  94 (74–118)  <.001  IGFBP3, µg/L  3,725 (3,135–4,281)  3,349 (2,836–3,918)  3,565 (3,009–4,162)  <.001  Characteristic  Women  Men  All  p Value  N = 2,021  N = 1,300  N = 3,321  Age, y  74.4 (5.0)  74.7 (5.2)  74.5 (5.1)  .05  Black race  310 (15.3)  158 (12.2)  468 (14.1)  .01  Current smoker  190 (9.4)  117 (9.0)  307 (9.2)  .70  Drinks alcohol  801 (39.6)  711 (54.7)  1,512 (45.5)  <.001  Physical activity, kcals  473 (150–1,110)  907 (907–1,967)  598 (187–1,417)  <.001  BMI, kg/m2  26.9 (5.2)  26.6 (3.8)  26.8 (4.7)  .07  Waist, cm  95.9 (14.6)  98.9 (10.5)  97.1 (13.2)  <.001  History of CHDa  339 (16.8)  359 (27.6)  698 (21.0)  <.001  History of CHF  91 (4.5)  66 (5.1)  157 (4.7)  .45  History of stroke  67 (3.3)  93 (7.2)  160 (4.8)  <.001  Adiponectin, mg/L  14.1 (10.2–19.6)  9.6 (7.0–13.6)  12.2 (8.5–17.5)  <.001  NT-proBNP, pg/mL  143 (76–257)  128 (60–306)  137 (69–268)  .11  DHEAS, µg/mL  0.52 (0.32–0.80)  0.79 (0.48–1.18)  0.61 (0.38–0.94)  <.001  IL-6, pg/mL  2.58 (1.75–3.91)  2.87 (2.02–4.34)  2.70 (1.85–4.08)  <.001  IGF-1, µg/L  89 (70–112.0)  103 (82–126)  94 (74–118)  <.001  IGFBP3, µg/L  3,725 (3,135–4,281)  3,349 (2,836–3,918)  3,565 (3,009–4,162)  <.001  Note: BMI = body mass index; CHD = coronary heart disease; CHF = congestive heart failure; DHEAS = dehydroepiandrosterone sulfate; IGF-1 = Insulin-like growth factor-1; IGFBP3 = IGF-binding protein-3; IL-6 = interleukin-6; NT-proBNP = amino-terminal pro-B-type natriuretic peptide. Entries in table are mean (SD), N (%), or median (interquartile range). p Values comparing men and women are from t test, chi-square test, or Kruskal–Wallis test, corresponding to whether or not the mean, N, or median are reported. aDefined as angina, angioplasty, bypass surgery, or myocardial Infarction. View Large Table 1. Description of Cardiovascular Health Study Cohort Characteristic  Women  Men  All  p Value  N = 2,021  N = 1,300  N = 3,321  Age, y  74.4 (5.0)  74.7 (5.2)  74.5 (5.1)  .05  Black race  310 (15.3)  158 (12.2)  468 (14.1)  .01  Current smoker  190 (9.4)  117 (9.0)  307 (9.2)  .70  Drinks alcohol  801 (39.6)  711 (54.7)  1,512 (45.5)  <.001  Physical activity, kcals  473 (150–1,110)  907 (907–1,967)  598 (187–1,417)  <.001  BMI, kg/m2  26.9 (5.2)  26.6 (3.8)  26.8 (4.7)  .07  Waist, cm  95.9 (14.6)  98.9 (10.5)  97.1 (13.2)  <.001  History of CHDa  339 (16.8)  359 (27.6)  698 (21.0)  <.001  History of CHF  91 (4.5)  66 (5.1)  157 (4.7)  .45  History of stroke  67 (3.3)  93 (7.2)  160 (4.8)  <.001  Adiponectin, mg/L  14.1 (10.2–19.6)  9.6 (7.0–13.6)  12.2 (8.5–17.5)  <.001  NT-proBNP, pg/mL  143 (76–257)  128 (60–306)  137 (69–268)  .11  DHEAS, µg/mL  0.52 (0.32–0.80)  0.79 (0.48–1.18)  0.61 (0.38–0.94)  <.001  IL-6, pg/mL  2.58 (1.75–3.91)  2.87 (2.02–4.34)  2.70 (1.85–4.08)  <.001  IGF-1, µg/L  89 (70–112.0)  103 (82–126)  94 (74–118)  <.001  IGFBP3, µg/L  3,725 (3,135–4,281)  3,349 (2,836–3,918)  3,565 (3,009–4,162)  <.001  Characteristic  Women  Men  All  p Value  N = 2,021  N = 1,300  N = 3,321  Age, y  74.4 (5.0)  74.7 (5.2)  74.5 (5.1)  .05  Black race  310 (15.3)  158 (12.2)  468 (14.1)  .01  Current smoker  190 (9.4)  117 (9.0)  307 (9.2)  .70  Drinks alcohol  801 (39.6)  711 (54.7)  1,512 (45.5)  <.001  Physical activity, kcals  473 (150–1,110)  907 (907–1,967)  598 (187–1,417)  <.001  BMI, kg/m2  26.9 (5.2)  26.6 (3.8)  26.8 (4.7)  .07  Waist, cm  95.9 (14.6)  98.9 (10.5)  97.1 (13.2)  <.001  History of CHDa  339 (16.8)  359 (27.6)  698 (21.0)  <.001  History of CHF  91 (4.5)  66 (5.1)  157 (4.7)  .45  History of stroke  67 (3.3)  93 (7.2)  160 (4.8)  <.001  Adiponectin, mg/L  14.1 (10.2–19.6)  9.6 (7.0–13.6)  12.2 (8.5–17.5)  <.001  NT-proBNP, pg/mL  143 (76–257)  128 (60–306)  137 (69–268)  .11  DHEAS, µg/mL  0.52 (0.32–0.80)  0.79 (0.48–1.18)  0.61 (0.38–0.94)  <.001  IL-6, pg/mL  2.58 (1.75–3.91)  2.87 (2.02–4.34)  2.70 (1.85–4.08)  <.001  IGF-1, µg/L  89 (70–112.0)  103 (82–126)  94 (74–118)  <.001  IGFBP3, µg/L  3,725 (3,135–4,281)  3,349 (2,836–3,918)  3,565 (3,009–4,162)  <.001  Note: BMI = body mass index; CHD = coronary heart disease; CHF = congestive heart failure; DHEAS = dehydroepiandrosterone sulfate; IGF-1 = Insulin-like growth factor-1; IGFBP3 = IGF-binding protein-3; IL-6 = interleukin-6; NT-proBNP = amino-terminal pro-B-type natriuretic peptide. Entries in table are mean (SD), N (%), or median (interquartile range). p Values comparing men and women are from t test, chi-square test, or Kruskal–Wallis test, corresponding to whether or not the mean, N, or median are reported. aDefined as angina, angioplasty, bypass surgery, or myocardial Infarction. View Large Table 2. Biomarker Index Scoring   Women  Men  Biomarker Quintile  Biomarker Range  β Coefficient  Pointsa  Biomarker Range  β Coefficient  Pointsa  NT-proBNP, pg/mL   0  ≤76  0  0  ≤62  0  0   1  78–127  0.2173  1  63–121  0.1673  1   2  128–199  0.2324  1  122–211  0.3381  1   3  200–355  0.5158  2  212–485  0.6764  3   4  >355  0.8539  3  >485  1.141  5  IL-6, pg/mL   0  ≤1.8  0  0  ≤2.0  0  0   1  1.9–2.5  0.1333  0  2.1–2.7  0.1577  1   2  2.6–3.3  0.3954  1  2.8–3.6  0.1863  1   3  3.4–4.6  0.4857  2  3.7–5.1  0.4775  2   4  >4.6  0.7344  3  >5.1  0.5773  2  IGF-1, µg/L   0  ≤66  0.1144  0  ≤75    0   1  67–80  0.1014  0  76–94    0   2  81–96  0.1284  0  95–111    0   3  97–120  0  0  112–1,332    0   4  >120  0.3517  1  >133    0  IGFBP3, µg/L   0  ≤2,951  0.3066  1  ≤2,621  0.2073  1   1  2,952–3,436  0  0  2,622–3,100  0.2306  1   2  3,437–3,881  0.0049  0  3,101–3,495  0.2000  1   3  3,882–4,379  0.1763  1  3,496–4,061  0  0   4  >4,379  0.1287  1  >4,061  0.4038  2  DHEAS, µg/mL   0  ≤0.26  0.2547  1  ≤0.41    0   1  0.27–0.42  0.1520  1  0.42–0.61    0   2  0.43–0.61  0  0  0.62–0.87    0   3  0.62–0.90  0.0255  0  0.88–1.24    0   4  >0.90  0.2674  1  >1.24    0    Women  Men  Biomarker Quintile  Biomarker Range  β Coefficient  Pointsa  Biomarker Range  β Coefficient  Pointsa  NT-proBNP, pg/mL   0  ≤76  0  0  ≤62  0  0   1  78–127  0.2173  1  63–121  0.1673  1   2  128–199  0.2324  1  122–211  0.3381  1   3  200–355  0.5158  2  212–485  0.6764  3   4  >355  0.8539  3  >485  1.141  5  IL-6, pg/mL   0  ≤1.8  0  0  ≤2.0  0  0   1  1.9–2.5  0.1333  0  2.1–2.7  0.1577  1   2  2.6–3.3  0.3954  1  2.8–3.6  0.1863  1   3  3.4–4.6  0.4857  2  3.7–5.1  0.4775  2   4  >4.6  0.7344  3  >5.1  0.5773  2  IGF-1, µg/L   0  ≤66  0.1144  0  ≤75    0   1  67–80  0.1014  0  76–94    0   2  81–96  0.1284  0  95–111    0   3  97–120  0  0  112–1,332    0   4  >120  0.3517  1  >133    0  IGFBP3, µg/L   0  ≤2,951  0.3066  1  ≤2,621  0.2073  1   1  2,952–3,436  0  0  2,622–3,100  0.2306  1   2  3,437–3,881  0.0049  0  3,101–3,495  0.2000  1   3  3,882–4,379  0.1763  1  3,496–4,061  0  0   4  >4,379  0.1287  1  >4,061  0.4038  2  DHEAS, µg/mL   0  ≤0.26  0.2547  1  ≤0.41    0   1  0.27–0.42  0.1520  1  0.42–0.61    0   2  0.43–0.61  0  0  0.62–0.87    0   3  0.62–0.90  0.0255  0  0.88–1.24    0   4  >0.90  0.2674  1  >1.24    0  Note: DHEAS = dehydroepiandrosterone sulfate; IGF-1 = Insulin-like growth factor-1; IGFBP3 = IGF-binding protein-3; IL-6 = interleukin-6; NT-proBNP = amino-terminal pro-B-type natriuretic peptide. aPoints determined by dividing the β coefficient by three times the β coefficient for age = 0.0912. IL-6 = interleukin-6. View Large Table 2. Biomarker Index Scoring   Women  Men  Biomarker Quintile  Biomarker Range  β Coefficient  Pointsa  Biomarker Range  β Coefficient  Pointsa  NT-proBNP, pg/mL   0  ≤76  0  0  ≤62  0  0   1  78–127  0.2173  1  63–121  0.1673  1   2  128–199  0.2324  1  122–211  0.3381  1   3  200–355  0.5158  2  212–485  0.6764  3   4  >355  0.8539  3  >485  1.141  5  IL-6, pg/mL   0  ≤1.8  0  0  ≤2.0  0  0   1  1.9–2.5  0.1333  0  2.1–2.7  0.1577  1   2  2.6–3.3  0.3954  1  2.8–3.6  0.1863  1   3  3.4–4.6  0.4857  2  3.7–5.1  0.4775  2   4  >4.6  0.7344  3  >5.1  0.5773  2  IGF-1, µg/L   0  ≤66  0.1144  0  ≤75    0   1  67–80  0.1014  0  76–94    0   2  81–96  0.1284  0  95–111    0   3  97–120  0  0  112–1,332    0   4  >120  0.3517  1  >133    0  IGFBP3, µg/L   0  ≤2,951  0.3066  1  ≤2,621  0.2073  1   1  2,952–3,436  0  0  2,622–3,100  0.2306  1   2  3,437–3,881  0.0049  0  3,101–3,495  0.2000  1   3  3,882–4,379  0.1763  1  3,496–4,061  0  0   4  >4,379  0.1287  1  >4,061  0.4038  2  DHEAS, µg/mL   0  ≤0.26  0.2547  1  ≤0.41    0   1  0.27–0.42  0.1520  1  0.42–0.61    0   2  0.43–0.61  0  0  0.62–0.87    0   3  0.62–0.90  0.0255  0  0.88–1.24    0   4  >0.90  0.2674  1  >1.24    0    Women  Men  Biomarker Quintile  Biomarker Range  β Coefficient  Pointsa  Biomarker Range  β Coefficient  Pointsa  NT-proBNP, pg/mL   0  ≤76  0  0  ≤62  0  0   1  78–127  0.2173  1  63–121  0.1673  1   2  128–199  0.2324  1  122–211  0.3381  1   3  200–355  0.5158  2  212–485  0.6764  3   4  >355  0.8539  3  >485  1.141  5  IL-6, pg/mL   0  ≤1.8  0  0  ≤2.0  0  0   1  1.9–2.5  0.1333  0  2.1–2.7  0.1577  1   2  2.6–3.3  0.3954  1  2.8–3.6  0.1863  1   3  3.4–4.6  0.4857  2  3.7–5.1  0.4775  2   4  >4.6  0.7344  3  >5.1  0.5773  2  IGF-1, µg/L   0  ≤66  0.1144  0  ≤75    0   1  67–80  0.1014  0  76–94    0   2  81–96  0.1284  0  95–111    0   3  97–120  0  0  112–1,332    0   4  >120  0.3517  1  >133    0  IGFBP3, µg/L   0  ≤2,951  0.3066  1  ≤2,621  0.2073  1   1  2,952–3,436  0  0  2,622–3,100  0.2306  1   2  3,437–3,881  0.0049  0  3,101–3,495  0.2000  1   3  3,882–4,379  0.1763  1  3,496–4,061  0  0   4  >4,379  0.1287  1  >4,061  0.4038  2  DHEAS, µg/mL   0  ≤0.26  0.2547  1  ≤0.41    0   1  0.27–0.42  0.1520  1  0.42–0.61    0   2  0.43–0.61  0  0  0.62–0.87    0   3  0.62–0.90  0.0255  0  0.88–1.24    0   4  >0.90  0.2674  1  >1.24    0  Note: DHEAS = dehydroepiandrosterone sulfate; IGF-1 = Insulin-like growth factor-1; IGFBP3 = IGF-binding protein-3; IL-6 = interleukin-6; NT-proBNP = amino-terminal pro-B-type natriuretic peptide. aPoints determined by dividing the β coefficient by three times the β coefficient for age = 0.0912. IL-6 = interleukin-6. View Large The associations of the Biomarker and Physiologic Indices with mortality were calculated using Cox proportional hazards models with each index modeled linearly among participants with both indices (N = 2,515). A likelihood ratio test was used to confirm that there was no improvement over the linear model by a model based on indicator variables for categories of the Biomarker Index. The proportional hazard assumption was tested with Schoenfeld residuals, and when violated, baseline hazards were stratified by sex and follow-up was truncated at 10 years to conform to the assumption. Models were built sequentially adjusting for age, sex, race, behaviors (current smoking, diet, any alcohol consumption, and ln(kcal) physical activity), body size (body mass index and waist circumference), prevalent clinical CVD or cancer, and, finally, incident CVD, cancer, or hospitalization for pneumonia during follow-up. All analyses were performed with STATA 14.2 (College Station, TX) or R version 2.13.0 (http://www.r-project.org). Results The mean (SD) age of the participants with all biomarkers (N = 3,321) was 74.5 (5.1) years, and 14.1% were black (Table 1). Compared with men, women were less likely to drink alcohol or have a history of coronary heart disease or stroke. Women had higher levels of adiponectin and IGFBP-3; lower levels of DHEAS, IL-6, and IGF-1; and similar NT-proBNP (Table 1). Among the 3,321 participants, there were 2,536 deaths over 19.5 years, and among the 2,525 participants with both indices, there were 1,003 deaths over 10 years. Cubic spline plots of the associations of the biomarkers with mortality in the test set showed departures from linearity, although data were sparse in the extremes of the distributions, as evidenced by the wide confidence intervals. Scoring of each biomarker is shown in Table 2. When the scores for each biomarker were summed into the total Biomarker Index, there was a clear gradation of mortality across the Biomarker Index score, with lower scores associated with lower mortality and higher scores associated with higher mortality (Table 3). Table 3. Incidence Rates for Mortality by Biomarker Index Score Biomarker Index  N  Person-Years  Number of Deaths  Incidence Ratea (95% CI)  0  77  720  15  20.8 (12.6, 34.5)  1  296  2,754  53  19.2 (14.7, 25.2)  2  610  5,574  147  26.4 (22.4, 31.0)  3  718  6,144  260  42.3 (37.5, 47.8)  4  576  4,754  243  51.1 (45.1, 58.0)  5  417  3,222  208  64.6 (56.3, 73.9)  6  287  2,053  178  86.7 (74.9, 100.4)  7  194  1,232  136  110.4 (83.3, 130.6)  8  126  627  106  169.1 (140.8, 204.5)  9  20  87  19  219.3 (139.9, 343.8)  Biomarker Index  N  Person-Years  Number of Deaths  Incidence Ratea (95% CI)  0  77  720  15  20.8 (12.6, 34.5)  1  296  2,754  53  19.2 (14.7, 25.2)  2  610  5,574  147  26.4 (22.4, 31.0)  3  718  6,144  260  42.3 (37.5, 47.8)  4  576  4,754  243  51.1 (45.1, 58.0)  5  417  3,222  208  64.6 (56.3, 73.9)  6  287  2,053  178  86.7 (74.9, 100.4)  7  194  1,232  136  110.4 (83.3, 130.6)  8  126  627  106  169.1 (140.8, 204.5)  9  20  87  19  219.3 (139.9, 343.8)  Note: CI = confidence interval. Wide confidence intervals for values of 0 and 9 with few deaths. aPer thousand person-years based on 10 years of follow-up. View Large Table 3. Incidence Rates for Mortality by Biomarker Index Score Biomarker Index  N  Person-Years  Number of Deaths  Incidence Ratea (95% CI)  0  77  720  15  20.8 (12.6, 34.5)  1  296  2,754  53  19.2 (14.7, 25.2)  2  610  5,574  147  26.4 (22.4, 31.0)  3  718  6,144  260  42.3 (37.5, 47.8)  4  576  4,754  243  51.1 (45.1, 58.0)  5  417  3,222  208  64.6 (56.3, 73.9)  6  287  2,053  178  86.7 (74.9, 100.4)  7  194  1,232  136  110.4 (83.3, 130.6)  8  126  627  106  169.1 (140.8, 204.5)  9  20  87  19  219.3 (139.9, 343.8)  Biomarker Index  N  Person-Years  Number of Deaths  Incidence Ratea (95% CI)  0  77  720  15  20.8 (12.6, 34.5)  1  296  2,754  53  19.2 (14.7, 25.2)  2  610  5,574  147  26.4 (22.4, 31.0)  3  718  6,144  260  42.3 (37.5, 47.8)  4  576  4,754  243  51.1 (45.1, 58.0)  5  417  3,222  208  64.6 (56.3, 73.9)  6  287  2,053  178  86.7 (74.9, 100.4)  7  194  1,232  136  110.4 (83.3, 130.6)  8  126  627  106  169.1 (140.8, 204.5)  9  20  87  19  219.3 (139.9, 343.8)  Note: CI = confidence interval. Wide confidence intervals for values of 0 and 9 with few deaths. aPer thousand person-years based on 10 years of follow-up. View Large The Biomarker Index was similarly distributed in the test and validation samples with a slight right skew (Figure 1, chi-square p value = .75). Estimates of the age, sex, and race adjusted hazard ratios (HRs) and 95% confidence intervals of the Biomarker Index with mortality were identical in the test (1.29 [1.24, 1.33]) and validation (1.29 [1.23, 1.35]) subsets, and measures of concordance were similar (Harrell’s C-statistic = 0.699 and 0.708 in the test and validation sets, respectively). Figure 1. View largeDownload slide Distribution of Biomarker Index score in test and validation samples of the Cardiovascular Health Study. Figure 1. View largeDownload slide Distribution of Biomarker Index score in test and validation samples of the Cardiovascular Health Study. Compared with participants with data on both indices, participants with only data for the Biomarker Index were on average of similar age and gender, but were more likely African American, had a larger waist circumference, had lower physical activity, and were more likely to have coronary heart disease or congestive heart failure, as well as slightly higher NT-proBNP and IL-6 (Supplementary Table 1). Among the 2,515 participants with data on both indices, combining the test and validation subsets, the mean (SD) of the Biomarker Index and Physiologic Index was 3.6 (1.9) and 4.4 (2.1), respectively. The indices were significantly correlated (Spearman correlation coefficient = 0.35, p < .0001). In separate unadjusted Cox models with the baseline hazard stratified by sex, the estimated HRs for the two indices were nearly identical at 1.34 for the Biomarker Index and 1.36 for the Physiologic Index (Table 4) with respective measures of concordance of 0.66 and 0.67. For the unadjusted Biomarker Index (Table 4, Model 1), the C-statistic was 0.656 and area under the curve was 0.685. For the unadjusted Physiologic Index, the C-statistic was 0.669 and area under the curve was 0.714 (test of comparison p value of .02). The HRs were attenuated after adjustment for age and for each other, but remained similar and statistically significant in all adjusted models. Table 4. Hazard Ratios for Death by Age and the Two Indices, With Various Levels of Adjustment Adjustment Variables  Age (per Year)  Physiologic Index (per 1 Index Point)  Biomarker Index (per 1 Index Point)  Sex  1.11 (1.10, 1.13)  1.36 (1.31, 1.40)  1.34 (1.30, 1.38)  Sex, race, and agea  1.08 (1.07, 1.09)  1.28 (1.23, 1.32)    1.09 (1.08, 1.10)    1.27 (1.23, 1.32)  + Both indices  1.07 (1.06, 1.08)  1.22 (1.18, 1.26)  1.21 (1.17, 1.25)  + Behaviors  1.07 (1.06, 1.09)  1.20 (1.15, 1.24)  1.20 (1.16, 1.24)  + Body size  1.06 (1.05, 1.08)  1.21 (1.17, 1.26)  1.19 (1.15, 1.23)  + Prevalent clinical CVD or cancer  1.06 (1.05, 1.08)  1.21 (1.16, 1.25)  1.18 (1.13, 1.22)  + Incident CVD, cancer, or hospitalization for pneumonia  1.05 (1.03, 1.07)  1.20 (1.15, 1.25)  1.19 (1.14, 1.24)  Adjustment Variables  Age (per Year)  Physiologic Index (per 1 Index Point)  Biomarker Index (per 1 Index Point)  Sex  1.11 (1.10, 1.13)  1.36 (1.31, 1.40)  1.34 (1.30, 1.38)  Sex, race, and agea  1.08 (1.07, 1.09)  1.28 (1.23, 1.32)    1.09 (1.08, 1.10)    1.27 (1.23, 1.32)  + Both indices  1.07 (1.06, 1.08)  1.22 (1.18, 1.26)  1.21 (1.17, 1.25)  + Behaviors  1.07 (1.06, 1.09)  1.20 (1.15, 1.24)  1.20 (1.16, 1.24)  + Body size  1.06 (1.05, 1.08)  1.21 (1.17, 1.26)  1.19 (1.15, 1.23)  + Prevalent clinical CVD or cancer  1.06 (1.05, 1.08)  1.21 (1.16, 1.25)  1.18 (1.13, 1.22)  + Incident CVD, cancer, or hospitalization for pneumonia  1.05 (1.03, 1.07)  1.20 (1.15, 1.25)  1.19 (1.14, 1.24)  Note: CVD = cardiovascular disease. The + symbol in the table indicates that each model with the + includes variables included in the prior model. aFirst line is adjusted for the Physiologic Index and second line for the Biomarker Index. View Large Table 4. Hazard Ratios for Death by Age and the Two Indices, With Various Levels of Adjustment Adjustment Variables  Age (per Year)  Physiologic Index (per 1 Index Point)  Biomarker Index (per 1 Index Point)  Sex  1.11 (1.10, 1.13)  1.36 (1.31, 1.40)  1.34 (1.30, 1.38)  Sex, race, and agea  1.08 (1.07, 1.09)  1.28 (1.23, 1.32)    1.09 (1.08, 1.10)    1.27 (1.23, 1.32)  + Both indices  1.07 (1.06, 1.08)  1.22 (1.18, 1.26)  1.21 (1.17, 1.25)  + Behaviors  1.07 (1.06, 1.09)  1.20 (1.15, 1.24)  1.20 (1.16, 1.24)  + Body size  1.06 (1.05, 1.08)  1.21 (1.17, 1.26)  1.19 (1.15, 1.23)  + Prevalent clinical CVD or cancer  1.06 (1.05, 1.08)  1.21 (1.16, 1.25)  1.18 (1.13, 1.22)  + Incident CVD, cancer, or hospitalization for pneumonia  1.05 (1.03, 1.07)  1.20 (1.15, 1.25)  1.19 (1.14, 1.24)  Adjustment Variables  Age (per Year)  Physiologic Index (per 1 Index Point)  Biomarker Index (per 1 Index Point)  Sex  1.11 (1.10, 1.13)  1.36 (1.31, 1.40)  1.34 (1.30, 1.38)  Sex, race, and agea  1.08 (1.07, 1.09)  1.28 (1.23, 1.32)    1.09 (1.08, 1.10)    1.27 (1.23, 1.32)  + Both indices  1.07 (1.06, 1.08)  1.22 (1.18, 1.26)  1.21 (1.17, 1.25)  + Behaviors  1.07 (1.06, 1.09)  1.20 (1.15, 1.24)  1.20 (1.16, 1.24)  + Body size  1.06 (1.05, 1.08)  1.21 (1.17, 1.26)  1.19 (1.15, 1.23)  + Prevalent clinical CVD or cancer  1.06 (1.05, 1.08)  1.21 (1.16, 1.25)  1.18 (1.13, 1.22)  + Incident CVD, cancer, or hospitalization for pneumonia  1.05 (1.03, 1.07)  1.20 (1.15, 1.25)  1.19 (1.14, 1.24)  Note: CVD = cardiovascular disease. The + symbol in the table indicates that each model with the + includes variables included in the prior model. aFirst line is adjusted for the Physiologic Index and second line for the Biomarker Index. View Large In the model with only age, the HR per year of age was 1.11 (1.10, 1.13; Table 4). Adjusting for the Biomarker Index reduced the HR for age to 1.09 (1.08, 1.10; 20% attenuation in age beta) and adjusting for the Physiologic Index reduced the HR for age to 1.08 (1.07, 1.09; 29% attenuation in age beta). Adjusting for both indices, the HR for age was 1.07 (1.06, 1.08; 39% attenuation in age beta); the HR for the Biomarker Index was 1.21 (1.17, 1.25); and the HR for the Physiologic Index was 1.22 (1.18, 1.26). Associations remained similar with additional adjustment for health behaviors, body size, and clinical CVD, cancer, and pneumonia hospitalization (Figure 2). Figure 2. View largeDownload slide Kaplan–Meier curve of cumulative mortality over the follow-up period by score of Biomarker Index. Figure 2. View largeDownload slide Kaplan–Meier curve of cumulative mortality over the follow-up period by score of Biomarker Index. Discussion In this sample of community-dwelling older adults, we show that the relative risk of mortality associated with an index of five new circulating biomarkers is similar to that of an index using physiologic markers that require more resource intensive ascertainment but that both sets of variables remain significant predictors of mortality even in the presence of each other. The Biomarker and Physiologic Indices attenuate age substantially and independently of each other, suggesting that each captures a part of biological aging that is distinct. Previously, we showed that the Physiologic Index and variations thereof are associated with death, disability, age-related chronic disease, frailty and vigor, telomere length, and many health behaviors in the Cardiovascular Health Study and Health, Aging, and Body Composition Study (1,2,4,13). In the Long Life Family Study, a family-based study of exceptional longevity, a variation of the Physiologic Index termed the Healthy Aging Index was also moderately and significantly heritable (14). Supporting prior analyses, the present data suggest that the two indices represent distinct aspects of the aging process. Thus, they may be feasible outcomes for benchmarking interventions being tested to alter the trajectory of aging. What aspects of aging biology may be reflected by the Biomarker Index? IL-6 is a cytokine produced by immune cells, vascular endothelium, adipose tissue, and muscle. It has proinflammatory and anti-inflammatory effects, rises with age in the absence of disease, and is robustly associated with myriad health outcomes (15,16). NT-proBNP is released primarily by ventricular myocytes in response to elevated filling pressure, acting hormonally on the kidneys to relieve cardiac stress through sodium and water excretion. NT-proBNP elevations occur in a number of cardiovascular aging and disease phenotypes, including coronary artery disease and myocardial infarction, left ventricular hypertrophy, atrial fibrillation, and valvular diseases, reflecting increased myocardial stretch. Higher NT-proBNP levels also occur with decreased renal clearance of NT-proBNP and are seen with advancing age (17,18). Various studies have shown NT-proBNP to be independently associated with mortality (19,20). Thus, NT-proBNP may be a marker of undiagnosed or subclinical cardiovascular damage that occurs with aging and disease, as well as co-existing renal dysfunction. DHEAS can be thought of as a pool for sex steroid hormones. More than 90% of estrogens in postmenopausal women and 30% of androgens in men are derived from DHEAS (21). DHEAS levels peak at birth and in the third decade of life and fall by 80%–90% by age 80, which partly explains the age-associated decline in estrogen and testosterone (22–24). The cause of this age-associated decline is unknown, although we reported that in older adults in the CHS All Stars study, CVD, male gender, and black race were associated with greater DHEAS decline over 9 years (25). Epidemiological data on an association between DHEAS and mortality are equivocal (26–29). The IGF pathway was the first pathway shown to influence aging and life span in animals, conceivably through growth and development and metabolic regulation (30). Studies in model systems are robust and suggest a causal role for IGF-1 and longevity in nonhuman systems, and cross-sectional epidemiologic studies of community-dwelling older adults confirmed that they have lower IGF-1 levels compared with younger persons (31–36). Nonetheless, although the Framingham Heart Study showed that greater baseline IGF-1 was associated with lower mortality (37), no association was initially detected in five other large cohort studies (10,38–41). IGF-1 was not strongly scored in the Biomarker Index, suggesting that at these ages it may be associated with mortality only in the extremes or in the context of other biomarker levels. We acknowledge several limitations in this analysis. First, the Biomarker Index was specifically scored by weighting its components using regression coefficients from Cox models of death. This methodology was not used to score the Physiologic Index whose components were monotonically associated with mortality and were scored with tertiles summed across the components. Consequently, the Physiologic Index may have been “disadvantaged” compared with the Biomarker Index in predicting death. However, we derived the Biomarker Index scoring in a randomly selected test sample and used it to predict death in a validation sample, where it performed similarly to the test sample, suggesting it was not overcalibrated. Second, the choice of biomarkers to include in the Biomarker Index was made from convenience in that the biomarkers were available in the CHS. However, we had good scientific rationale for including each candidate biomarker in the index. Before inclusion, each biomarker’s statistical appropriateness was additionally confirmed (ie, association with mortality independent of age and sex). In the future, high-throughput technology could allow affordable prospective measurement of hundreds or thousands of biomarkers simultaneously, which may increase predictive power and suggest new mechanistic associations. The challenge of incorporating large numbers of biomarkers into clinical practice may eventually be overcome by designing targeted “aging” panels using validated biomarkers, similar to targeted genetic prediction panels already used in clinical oncology. Third, we measured each biomarker only once and levels may have changed throughout time. Previous analyses suggest it is unlikely that participants improved significantly on any component of the Physiologic Index, but there is evidence that biomarkers can have varied trajectories that themselves may predict death better than a single measurement (42,43). Fourth, residual confounding may be present if confounders were not included in our models, such as dietary habits, although adjustment for major health behaviors and clinical disease minimally altered results. Fifth, although some different associations have been detected previously in CHS for some of these biomarkers (eg, association between adiponectin and mortality varying by CVD status (44)), the aim here was to find a biomarker generally predictive of mortality irrespective of disease status. Differences in study population and analytic design may have influenced results compared with prior reports. In conclusion, an index of five circulating biomarkers can predict death in older community-dwelling adults as accurately as an index of physiologic markers, and both indices substantially and independently attenuate the effect of age on mortality. Although the Biomarker Index needs further validation, these data suggest that Biomarker Index and the Physiologic Index measure important aspects of the underlying aging process, and they may serve as end points for studies seeking to develop interventions to improve aging. Supplementary Material Supplementary data is available at The Journals of Gerontology, Series A: Biological Sciences and Medical Sciences online. Funding This research was supported by contracts HHSN268201200036C, HHSN268200800007C, N01 HC55222, N01HC85079 through N01HC85083, N01HC85086, and grant HL080295 from the National Heart, Lung, and Blood Institute (NHLBI), with additional contribution from the National Institute of Neurological Disorders and Stroke (NINDS). Additional support was provided by AG023629, AG-15928, AG-20098, and AG-027058 from the National Institute on Aging (NIA). IGF-1 measurement was supported by 1R01AG031890 to R.C.K. A full list of principal CHS investigators and institutions can be found at CHS-NHLBI.org. Conflict of Interest None reported. Acknowledgments We thank Liana del Gobbo and Dariush Mozaffarian for their prior work on dietary indices in CHS that allowed calculation of the Alternative Healthy Eating Index. References 1. Sanders JL, Boudreau RM, Penninx BW, et al.  ; Health ABC Study. Association of a modified physiologic index with mortality and incident disability: the Health, Aging, and Body Composition Study. J Gerontol A Biol Sci Med Sci . 2012; 67: 1439– 1446. doi: 10.1093/gerona/gls123 Google Scholar CrossRef Search ADS PubMed  2. Newman AB, Boudreau RM, Naydeck BL, Fried LF, Harris TB. A physiologic index of comorbidity: relationship to mortality and disability. J Gerontol A Biol Sci Med Sci . 2008; 63: 603– 609. Google Scholar CrossRef Search ADS PubMed  3. Sanders JL, Boudreau RM, Fried LP, Walston JD, Harris TB, Newman AB. Measurement of organ structure and function enhances understanding of the physiological basis of frailty: the Cardiovascular Health Study. J Am Geriatr Soc . 2011; 59: 1581– 1588. doi: 10.1111/j.1532-5415.2011.03557.x Google Scholar CrossRef Search ADS PubMed  4. Sanders JL, Fitzpatrick AL, Boudreau RM, et al.   Leukocyte telomere length is associated with noninvasively measured age-related disease: the Cardiovascular Health Study. J Gerontol A Biol Sci Med Sci . 2012; 67: 409– 416. doi: 10.1093/gerona/glr173 Google Scholar CrossRef Search ADS PubMed  5. Sanders JL, Boudreau R, Newman A. Understanding the aging process using epidemiologic approaches. In: Newman A, Cauley J, eds. Epidemiology of Aging . Dordrecht, The Netherlands: Springer; 2012: 187– 214. Google Scholar CrossRef Search ADS   6. Sprott RL. Biomarkers of aging and disease: introduction and definitions. Exp Gerontol . 2010; 45: 2– 4. doi: 10.1016/j.exger.2009.07.008 Google Scholar CrossRef Search ADS PubMed  7. Newman AB, Arnold AM, Sachs MC, et al.   Long-term function in an older cohort – the Cardiovascular Health Study All Stars study. J Am Geriatr Soc . 2009; 57: 432– 440. doi: 10.1111/j.1532-5415.2008.02152.x Google Scholar CrossRef Search ADS PubMed  8. Fried LP, Borhani NO, Enright P, et al.   The Cardiovascular Health Study: design and rationale. Ann Epidemiol . 1991; 1: 263– 276. Google Scholar CrossRef Search ADS PubMed  9. Cushman M, Cornell ES, Howard PR, Bovill EG, Tracy RP. Laboratory methods and quality assurance in the Cardiovascular Health Study. Clin Chem . 1995; 41: 264– 270. Google Scholar PubMed  10. Kaplan RC, McGinn AP, Pollak MN, et al.   Total insulinlike growth factor 1 and insulin-like growth factor binding protein levels, functional status, and mortality in older adults. J Am Geriatr Soc . 2008; 56: 652– 660. doi: 10.1111/j.1532-5415.2007.01637.x Google Scholar CrossRef Search ADS PubMed  11. Del Gobbo LC, Kalantarian S, Imamura F, et al.   Contribution of major lifestyle risk factors for incident heart failure in older adults: the Cardiovascular Health Study. JACC Heart Fail . 2015; 3: 520– 528. doi: 10.1016/j.jchf.2015.02.009 Google Scholar CrossRef Search ADS PubMed  12. Psaty BM, Kuller LH, Bild D, et al.   Methods of assessing prevalent cardiovascular disease in the Cardiovascular Health Study. Ann Epidemiol . 1995; 5: 270– 277. Google Scholar CrossRef Search ADS PubMed  13. Sanders JL, Boudreau RM, Fried LP, Walston JD, Harris TB, Newman AB. Measurement of organ structure and function enhances understanding of the physiological basis of frailty: the Cardiovascular Health Study. J Am Geriatr Soc . 2011; 59: 1581– 1588. doi: 10.1111/j.1532-5415.2011.03557.x Google Scholar CrossRef Search ADS PubMed  14. Sanders JL, Minster RL, Barmada MM, et al.   Heritability of and mortality prediction with a longevity phenotype: the healthy aging index. J Gerontol A Biol Sci Med Sci . 2014; 69: 479– 485. doi: 10.1093/gerona/glt117 Google Scholar CrossRef Search ADS PubMed  15. Singh T, Newman AB. Inflammatory markers in population studies of aging. Ageing Res Rev . 2011; 10: 319– 329. doi: 10.1016/j.arr.2010.11.002 Google Scholar CrossRef Search ADS PubMed  16. Maggio M, Guralnik JM, Longo DL, Ferrucci L. Interleukin-6 in aging and chronic disease: a magnificent pathway. J Gerontol A Biol Sci Med Sci . 2006; 61: 575– 584. Google Scholar CrossRef Search ADS PubMed  17. Redfield MM, Rodeheffer RJ, Jacobsen SJ, Mahoney DW, Bailey KR, Burnett JCJr. Plasma brain natriuretic peptide concentration: impact of age and gender. J Am Coll Cardiol . 2002; 40: 976– 982. Google Scholar CrossRef Search ADS PubMed  18. Charloux A, Brandenberger G, Piquard F, Geny B. Dysregulation of pulsatility in aging IV. Pulsatile signaling and cardiovascular aging: functions and regulation of natriuretic peptide signaling. Ageing Res Rev . 2008; 7: 151– 163. doi: 10.1016/j.arr.2007.12.001 Google Scholar CrossRef Search ADS PubMed  19. Witham MD, Gillespie ND, Hutcheon SD, Struthers AD, McMurdo ME. B-type natriuretic peptide is associated with mortality in older functionally impaired patients. J Am Geriatr Soc . 2005; 53: 1991– 1995. doi: 10.1111/j.1532-5415.2005.53555.x Google Scholar CrossRef Search ADS PubMed  20. Beleigoli AM, Boersma E, Diniz MdFH, Vidigal PG, Lima-Costa MF, Ribeiro AL. C-reactive protein and B-type natriuretic peptide yield either a non-significant or a modest incremental value to traditional risk factors in predicting long-term overall mortality in older adults. PLoS One . 2013; 8: e75809. doi: 10.1371/journal.pone.0075809 Google Scholar CrossRef Search ADS PubMed  21. Labrie F, Bélanger A, Simard J, Van L-T, Labrie C. DHEA and peripheral androgen and estrogen formation: intracinology. Ann NY Acad Sci . 1995; 774: 16– 28. Google Scholar CrossRef Search ADS PubMed  22. Longcope C. Metabolism of dehydroepiandrosterone. Ann NY Acad Sci . 1995; 774: 143– 148. Google Scholar CrossRef Search ADS PubMed  23. Orentreich N, Brind JL, Rizer RL, Vogelman JH. Age changes and sex differences in serum dehydroepiandrosterone sulfate concentrations throughout adulthood. J Clin Endocrinol Metab . 1984; 59: 551– 555. doi: 10.1210/jcem-59-3-551 Google Scholar CrossRef Search ADS PubMed  24. Vermeulen A. Dehydroepiandrosterone sulfate and aging. Ann NY Acad Sci . 1995; 774: 121– 127. Google Scholar CrossRef Search ADS PubMed  25. Sanders JL, Boudreau RM, Cappola AR, et al.   Cardiovascular disease is associated with greater incident dehydroepiandrosterone sulfate decline in the oldest old: the Cardiovascular Health Study All Stars study. J Am Geriatr Soc . 2010; 58: 421– 426. doi: 10.1111/j.1532-5415.2010.02724.x Google Scholar CrossRef Search ADS PubMed  26. Tchernof A, Labrie F. Dehydroepiandrosterone, obesity and cardiovascular disease risk: a review of human studies. Eur J Endocrinol . 2004; 151: 1– 14. Google Scholar CrossRef Search ADS PubMed  27. Maggio M, Lauretani F, Ceda GP, et al.   Relationship between low levels of anabolic hormones and 6-year mortality in older men: the Aging in the Chianti Area (InCHIANTI) study. Arch Intern Med . 2007; 167: 2249– 2254. doi: 10.1001/archinte.167.20.2249 Google Scholar CrossRef Search ADS PubMed  28. Glei DA, Goldman N. Dehydroepiandrosterone sulfate (DHEAS) and risk for mortality among older Taiwanese. Ann Epidemiol . 2006; 16: 510– 515. doi: 10.1016/j.annepidem.2005.09.010 Google Scholar CrossRef Search ADS PubMed  29. Enomoto M, Adachi H, Fukami A, et al.   Serum dehydroepiandrosterone sulfate levels predict longevity in men: 27-year follow-up study in a community-based cohort (Tanushimaru study). J Am Geriatr Soc . 2008; 56: 994– 998. doi: 10.1111/j.1532-5415.2008.01692.x Google Scholar CrossRef Search ADS PubMed  30. Kenyon C. The plasticity of aging: insights from long-lived mutants. Cell . 2005; 120: 449– 460. doi: 10.1016/j.cell.2005.02.002 Google Scholar CrossRef Search ADS PubMed  31. Landin-Wilhelmsen K, Wilhelmsen L, Lappas G, et al.   Serum insulin-like growth factor I in a random population sample of men and women: relation to age, sex, smoking habits, coffee consumption and physical activity, blood pressure and concentrations of plasma lipids, fibrinogen, parathyroid hormone and osteocalcin. Clin Endocrinol (Oxf) . 1994; 41: 351– 357. Google Scholar CrossRef Search ADS PubMed  32. Papadakis MA, Grady D, Tierney MJ, Black D, Wells L, Grunfeld C. Insulin-like growth factor 1 and functional status in healthy older men. J Am Geriatr Soc . 1995; 43: 1350– 1355. Google Scholar CrossRef Search ADS PubMed  33. Goodman-Gruen D, Barrett-Connor E. Epidemiology of insulin-like growth factor-I in elderly men and women: the Rancho Bernardo Study. Am J Epidemiol . 1997; 145: 970– 976. Google Scholar CrossRef Search ADS PubMed  34. Harris TB, Kiel D, Roubenoff R, et al.   Association of insulin-like growth factor-I with body composition, weight history, and past health behaviors in the very old: the Framingham Heart Study. J Am Geriatr Soc . 1997; 45: 133– 139. Google Scholar CrossRef Search ADS PubMed  35. O’Connor KG, Tobin JD, Harman SM, et al.   Serum levels of insulin-like growth factor-I are related to age and not to body composition in healthy women and men. J Gerontol A Biol Sci Med Sci . 1998; 53: M176– M182. Google Scholar CrossRef Search ADS PubMed  36. Boonen S, Lysens R, Verbeke G, et al.   Relationship between age-associated endocrine deficiencies and muscle function in elderly women: a cross-sectional study. Age Ageing . 1998; 27: 449– 454. Google Scholar CrossRef Search ADS PubMed  37. Roubenoff R, Parise H, Payette HA, et al.   Cytokines, insulin-like growth factor 1, sarcopenia, and mortality in very old community-dwelling men and women: the Framingham Heart Study. Am J Med . 2003; 115: 429– 435. Google Scholar CrossRef Search ADS PubMed  38. Cappola AR, Xue QL, Ferrucci L, Guralnik JM, Volpato S, Fried LP. Insulin-like growth factor I and interleukin-6 contribute synergistically to disability and mortality in older women. J Clin Endocrinol Metab . 2003; 88: 2019– 2025. doi: 10.1210/jc.2002-021694 Google Scholar CrossRef Search ADS PubMed  39. Saydah S, Graubard B, Ballard-Barbash R, Berrigan D. Insulin-like growth factors and subsequent risk of mortality in the United States. Am J Epidemiol . 2007; 166: 518– 526. doi: 10.1093/aje/kwm124 Google Scholar CrossRef Search ADS PubMed  40. Laughlin GA, Barrett-Connor E, Criqui MH, Kritz-Silverstein D. The prospective association of serum insulin-like growth factor I (IGF-I) and IGF-binding protein-1 levels with all cause and cardiovascular disease mortality in older adults: the Rancho Bernardo Study. J Clin Endocrinol Metab . 2004; 89: 114– 120. doi: 10.1210/jc.2003-030967 Google Scholar CrossRef Search ADS PubMed  41. Harrela M, Qiao Q, Koistinen R, et al.   High serum insulin-like growth factor binding protein-1 is associated with increased cardiovascular mortality in elderly men. Horm Metab Res . 2002; 34: 144– 149. doi: 10.1055/s-2002-23198 Google Scholar CrossRef Search ADS PubMed  42. Sanders JL, Guo W, O’Meara ES, et al.   Trajectories of IGF-I predict mortality in older adults: the Cardiovascular Health Study. J Gerontol A Biol Sci Med Sci . 2017. doi:10.1093/gerona/glx143. 43. Cappola AR, O’Meara ES, Guo W, Bartz TM, Fried LP, Newman AB. Trajectories of dehydroepiandrosterone sulfate predict mortality in older adults: the Cardiovascular Health Study. J Gerontol A Biol Sci Med Sci . 2009; 64: 1268– 1274. doi: 10.1093/gerona/glp129 Google Scholar CrossRef Search ADS PubMed  44. Kizer JR, Benkeser D, Arnold AM, et al.   Associations of total and high-molecular-weight adiponectin with all-cause and cardiovascular mortality in older persons: the Cardiovascular Health Study. Circulation . 2012; 126: 2951– 2961. doi: 10.1161/CIRCULATIONAHA.112.135202 Google Scholar CrossRef Search ADS PubMed  © The Author(s) 2018. Published by Oxford University Press on behalf of The Gerontological Society of America. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png The Journals of Gerontology Series A: Biomedical Sciences and Medical Sciences Oxford University Press

Loading next page...
 
/lp/ou_press/association-of-biomarker-and-physiologic-indices-with-mortality-in-Qc4Eu5H5j9
Publisher
Oxford University Press
Copyright
© The Author(s) 2018. Published by Oxford University Press on behalf of The Gerontological Society of America. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com.
ISSN
1079-5006
eISSN
1758-535X
D.O.I.
10.1093/gerona/gly075
Publisher site
See Article on Publisher Site

Abstract

Abstract Background A goal of gerontology is discovering aging phenotypes that reflect biological aging distinct from disease pathogenesis. Biomarkers that strongly and independently associated with mortality and that statistically attenuated chronologic age could be used to define such a phenotype. We determined the association of a Biomarker Index (BI) with mortality and compared it with a validated Physiologic Index (PI) in older adults. Methods The indices were constructed in the Cardiovascular Health Study, mean (SD) age 74.5 (5.1) years. The BI incorporated circulating levels of new biomarkers, including insulin-like growth factor-1, insulin-like growth factor–binding protein 3, amino-terminal pro-B-type natriuretic peptide, dehydroepiandrosterone sulfate, and interleukin-6, and was built in test (N = 2,197) and validation (N = 1,124) samples. The PI included carotid intima–media thickness, pulmonary capacity, brain white matter grade, cystatin-C, and fasting glucose. Multivariable Cox proportional hazards models predicting death were calculated with 10 years of follow-up. Results In separate age-adjusted models, the hazard ratio for mortality per point of the BI was 1.30 (95% confidence interval 1.25, 1.34) and the BI attenuated age by 25%. The hazard ratio for the PI was 1.28 (1.24, 1.33; 29% age attenuation). In the same model, the hazard ratio for the BI was 1.23 (1.18, 1.28) and for the PI was 1.22 (1.17, 1.26), and age was attenuated 42.5%. Associations persisted after further adjustment. Conclusions The BI and PI were significantly and independently associated with mortality. Both attenuated the age effect on mortality substantially. The indices may be feasible phenotypes for developing interventions hoping to alter the trajectory of aging. Biomarker, Mortality, Phenotype, Longevity A central goal of gerontology is to discover aging phenotypes that reflect biological aging distinct from disease pathogenesis. Healthy aging phenotypes could serve as end points for developing interventions that alter the trajectory of aging toward a more desired pattern. Due to wider distribution and possibly incorporating multiple biological processes, composite scores have the potential for improved capability to stratify mortality risk (1). In 2008, Newman and colleagues published the Physiologic Index of Comorbidity in the Cardiovascular Health Study (CHS) (2). The Physiologic Index combined measurements including carotid intima–media thickness, pulmonary vital capacity, serum cystatin-C, brain white matter grade, and fasting glucose. In the CHS (mean age 74.5 years), only 1.7% of older adults had no evidence of disease on these five tests (2). Their 9-year mortality rate was very low at 7/1,000 person-years (2), frailty was rare (3), and they had longer leukocyte telomere length (4), demonstrating that the tests identify individuals who are exceptional. Furthermore, the Physiologic Index explained 40% of the effect of age on mortality; thus, the Physiologic Index captures aspects of biological aging for which chronologic age is a surrogate. Although a valuable phenotype, measurement of several key components of the Physiologic Index is resource intensive and costly. An index composed solely of blood-based biomarkers readily measurable in clinical laboratories would have lower expense and greater applicability. We proposed criteria to validate biomarkers of aging in epidemiologic studies (5) that are distinct from definitions of biomarkers of aging derived from laboratory science (6). The criteria include (a) demonstrating biological plausibility that a biomarker describes a basic aging process; (b) demonstrating potential for clinical applicability, using highly reproducible, highly interindividually variable, and low-cost measurements; (c) subsequently assessing associations with important aging outcomes using optimal epidemiologic study designs; and (d) testing these associations in accordance with key statistical considerations. An ideal biomarker of human aging should meet these criteria. In this study, we developed a Biomarker Index of aging using only blood-based biomarkers demonstrated to have a role in the biology of aging in laboratory and epidemiologic studies. We compared the associations of the Biomarker Index and the Physiologic Index with 10-year mortality, and the ability of each Index to explain the association of age with mortality. Methods Study Population The CHS is an ongoing community-based study of cardiovascular risk in 5,888 men and women over the age of 65 years, from four regions of the United States (7,8). The cohort was enrolled in 1989–1990 and was supplemented with added recruitment of African Americans in 1992–1993. Participants and eligible household members were identified from Medicare eligibility lists. To be eligible, participants should be 65 years and older, should not have cancer under active treatment, should not be wheelchair bound or bed bound in the home, and should not plan to move out of the area within 3 years. We used data from the 1992–1993 examination as baseline to include the supplemental African American cohort and because the brain magnetic resonance imaging scan was conducted at that time. The CHS is approved by the institutional review boards of all participating institutions, and all participants gave informed consent. Physiologic Index of Comorbidity The Physiologic Index was calculated as previously described from continuous measures of age-related dysfunction in five major organ systems, each of which predicts mortality and disability across a large continuum of risk (2). Detailed methods are provided in Supplementary Material. The systems included were the vascular (carotid intima–media thickness on ultrasound), neurologic (white matter grade on magnetic resonance imaging), renal (cystatin-C), pulmonary (forced vital capacity), and metabolic (fasting glucose). To construct the Physiologic Index, tertiles were considered for four of the five measures, with the best values classified as 0 and the worst as 2; fasting glucose was scored according to clinical cutoffs. Although the choice of cut points was arbitrary, scores of “0” generally included normal values in healthy young individuals, and values of “2” were in the range of those with chronic disease. Individual scores were summed for a total score ranging from 0 (healthy) to 10 (unhealthy). Candidate Biomarkers Candidate biomarkers for inclusion in the Biomarker Index were chosen for their documented association with aging-related processes in laboratory studies and outcomes in epidemiologic studies, though debate remains on their individual role as causative or correlative factors with aging. Candidates included insulin-like growth factor (IGF)-1, IGF-binding protein (IGFBP)-3, adiponectin, interleukin-6 (IL-6), dehydroepiandrosterone sulfate (DHEAS), and amino-terminal pro-B-type natriuretic peptide (NT-proBNP). Fasting blood samples were collected at the 1992–1993 exam using standardized protocols and quality assurance (8,9). IGF-1 and IGFBP-3 were measured after an extraction step using enzyme-linked immunosorbent assays (ELISA; Diagnostics Systems Laboratory, Webster, TX) (10). The analytic coefficient of variation (CV) was 4%–6% for IGF-1 and 3%–5% for IGFBP-3. Adiponectin was measured with an ELISA (R&D Systems, Minneapolis, MN); intra- and interassay CVs were 2.5%–4.7% and 5.8%–6.9%, respectively. IL-6 was measured by ultrasensitive ELISA (Quantikine HS Human IL-6 Immunoassay; R&D Systems); intra- and interassay CV were 2.9%–8.7% and 7.3%–9.0%, respectively. Plasma DHEAS was measured with a competitive immunoassay (Alpco Diagnostics, Windham, NH) with interassay CV of 3.8%–7.2%. NT-proBNP was measured on the Elecsys 2010 system (Roche Diagnostics, Indianapolis, IN) with a CV of 2%–5%. Mortality Deaths were ascertained through participant surveillance every 6 months from study inception. Confirmation of deaths was conducted through reviews of obituaries, medical records, death certificates, the Center for Medicare Studies health care utilization database, and the National Death Index. Contacts and proxies were also interviewed for participants unavailable for follow-up. Ascertainment of vital status was 100% complete. Potential Confounders Age, sex, race, smoking status, alcohol consumption, and diet were determined by self-report (8). The Alternative Healthy Eating Score was used to account for dietary patterns (11). Anthropometrics (weight, height, waist circumference) were measured with standard protocols, and weight and height were used to calculate body mass index in kilogram per square meter. Physical activity was based on the Modified Minnesota Leisure Time Activities questionnaire that assessed frequency and duration of 18 activities in the prior week to calculate kilocalories of energy expended. Prevalent and incident cardiovascular disease (CVD) outcomes were adjudicated by an expert panel using medical records; medication information and prevalent disease status were updated at each examination (12). Incident CVD was defined as a first occurrence of stroke, coronary heart disease, congestive heart failure, or claudication after the 1992–1993 examination among participants without a history of any component at the time of the exam. Cancer and pneumonia were defined using ICD-9 codes. Statistical Analysis CHS participants were included in the development of the Biomarker Index if they had data on all six biomarkers (N = 3,321). We derived the Biomarker Index in a randomly selected test sample of two third of the cohort (N = 2,197) and validated it in the remainder of participants (N = 1,124). In the test sample, the association of each biomarker with mortality was visualized using cubic splines. Because biomarker levels differed by gender and several associations between biomarkers and mortality were not monotonic, cut points to score each biomarker were made using sex-specific quintiles (Tables 1 and 2). Multivariable Cox proportional hazards models including age and all of the biomarkers (as indicator variables for quintiles) were run for men (N = 862) and women (N = 1,335) separately in the test sample. Only biomarkers that were individually associated with mortality at a significance level of less than or equal to .05 were included in the final Biomarker Index. Adiponectin was not associated with mortality for either men or women and IGF-1 and DHEAS were not significantly associated with mortality in men so were not included in the final models. The coefficients from the multivariable biomarker models were used to compute points for each level of each biomarker, selecting the quintile with lowest risk as the reference group. Each beta coefficient was divided by three times the coefficient for age for scaling and rounded to produce a point score. Although 5 years of age has been used in other risk scores, notably the Framingham Risk Score, 3 years of age was selected as a reasonable comparison given the older ages of our cohort. The Biomarker Index was tested in the validation sample and, when found consistent with the results in the test set, the validation and test samples were combined to derive estimates of risk associated with both indices in the full sample. Table 1. Description of Cardiovascular Health Study Cohort Characteristic  Women  Men  All  p Value  N = 2,021  N = 1,300  N = 3,321  Age, y  74.4 (5.0)  74.7 (5.2)  74.5 (5.1)  .05  Black race  310 (15.3)  158 (12.2)  468 (14.1)  .01  Current smoker  190 (9.4)  117 (9.0)  307 (9.2)  .70  Drinks alcohol  801 (39.6)  711 (54.7)  1,512 (45.5)  <.001  Physical activity, kcals  473 (150–1,110)  907 (907–1,967)  598 (187–1,417)  <.001  BMI, kg/m2  26.9 (5.2)  26.6 (3.8)  26.8 (4.7)  .07  Waist, cm  95.9 (14.6)  98.9 (10.5)  97.1 (13.2)  <.001  History of CHDa  339 (16.8)  359 (27.6)  698 (21.0)  <.001  History of CHF  91 (4.5)  66 (5.1)  157 (4.7)  .45  History of stroke  67 (3.3)  93 (7.2)  160 (4.8)  <.001  Adiponectin, mg/L  14.1 (10.2–19.6)  9.6 (7.0–13.6)  12.2 (8.5–17.5)  <.001  NT-proBNP, pg/mL  143 (76–257)  128 (60–306)  137 (69–268)  .11  DHEAS, µg/mL  0.52 (0.32–0.80)  0.79 (0.48–1.18)  0.61 (0.38–0.94)  <.001  IL-6, pg/mL  2.58 (1.75–3.91)  2.87 (2.02–4.34)  2.70 (1.85–4.08)  <.001  IGF-1, µg/L  89 (70–112.0)  103 (82–126)  94 (74–118)  <.001  IGFBP3, µg/L  3,725 (3,135–4,281)  3,349 (2,836–3,918)  3,565 (3,009–4,162)  <.001  Characteristic  Women  Men  All  p Value  N = 2,021  N = 1,300  N = 3,321  Age, y  74.4 (5.0)  74.7 (5.2)  74.5 (5.1)  .05  Black race  310 (15.3)  158 (12.2)  468 (14.1)  .01  Current smoker  190 (9.4)  117 (9.0)  307 (9.2)  .70  Drinks alcohol  801 (39.6)  711 (54.7)  1,512 (45.5)  <.001  Physical activity, kcals  473 (150–1,110)  907 (907–1,967)  598 (187–1,417)  <.001  BMI, kg/m2  26.9 (5.2)  26.6 (3.8)  26.8 (4.7)  .07  Waist, cm  95.9 (14.6)  98.9 (10.5)  97.1 (13.2)  <.001  History of CHDa  339 (16.8)  359 (27.6)  698 (21.0)  <.001  History of CHF  91 (4.5)  66 (5.1)  157 (4.7)  .45  History of stroke  67 (3.3)  93 (7.2)  160 (4.8)  <.001  Adiponectin, mg/L  14.1 (10.2–19.6)  9.6 (7.0–13.6)  12.2 (8.5–17.5)  <.001  NT-proBNP, pg/mL  143 (76–257)  128 (60–306)  137 (69–268)  .11  DHEAS, µg/mL  0.52 (0.32–0.80)  0.79 (0.48–1.18)  0.61 (0.38–0.94)  <.001  IL-6, pg/mL  2.58 (1.75–3.91)  2.87 (2.02–4.34)  2.70 (1.85–4.08)  <.001  IGF-1, µg/L  89 (70–112.0)  103 (82–126)  94 (74–118)  <.001  IGFBP3, µg/L  3,725 (3,135–4,281)  3,349 (2,836–3,918)  3,565 (3,009–4,162)  <.001  Note: BMI = body mass index; CHD = coronary heart disease; CHF = congestive heart failure; DHEAS = dehydroepiandrosterone sulfate; IGF-1 = Insulin-like growth factor-1; IGFBP3 = IGF-binding protein-3; IL-6 = interleukin-6; NT-proBNP = amino-terminal pro-B-type natriuretic peptide. Entries in table are mean (SD), N (%), or median (interquartile range). p Values comparing men and women are from t test, chi-square test, or Kruskal–Wallis test, corresponding to whether or not the mean, N, or median are reported. aDefined as angina, angioplasty, bypass surgery, or myocardial Infarction. View Large Table 1. Description of Cardiovascular Health Study Cohort Characteristic  Women  Men  All  p Value  N = 2,021  N = 1,300  N = 3,321  Age, y  74.4 (5.0)  74.7 (5.2)  74.5 (5.1)  .05  Black race  310 (15.3)  158 (12.2)  468 (14.1)  .01  Current smoker  190 (9.4)  117 (9.0)  307 (9.2)  .70  Drinks alcohol  801 (39.6)  711 (54.7)  1,512 (45.5)  <.001  Physical activity, kcals  473 (150–1,110)  907 (907–1,967)  598 (187–1,417)  <.001  BMI, kg/m2  26.9 (5.2)  26.6 (3.8)  26.8 (4.7)  .07  Waist, cm  95.9 (14.6)  98.9 (10.5)  97.1 (13.2)  <.001  History of CHDa  339 (16.8)  359 (27.6)  698 (21.0)  <.001  History of CHF  91 (4.5)  66 (5.1)  157 (4.7)  .45  History of stroke  67 (3.3)  93 (7.2)  160 (4.8)  <.001  Adiponectin, mg/L  14.1 (10.2–19.6)  9.6 (7.0–13.6)  12.2 (8.5–17.5)  <.001  NT-proBNP, pg/mL  143 (76–257)  128 (60–306)  137 (69–268)  .11  DHEAS, µg/mL  0.52 (0.32–0.80)  0.79 (0.48–1.18)  0.61 (0.38–0.94)  <.001  IL-6, pg/mL  2.58 (1.75–3.91)  2.87 (2.02–4.34)  2.70 (1.85–4.08)  <.001  IGF-1, µg/L  89 (70–112.0)  103 (82–126)  94 (74–118)  <.001  IGFBP3, µg/L  3,725 (3,135–4,281)  3,349 (2,836–3,918)  3,565 (3,009–4,162)  <.001  Characteristic  Women  Men  All  p Value  N = 2,021  N = 1,300  N = 3,321  Age, y  74.4 (5.0)  74.7 (5.2)  74.5 (5.1)  .05  Black race  310 (15.3)  158 (12.2)  468 (14.1)  .01  Current smoker  190 (9.4)  117 (9.0)  307 (9.2)  .70  Drinks alcohol  801 (39.6)  711 (54.7)  1,512 (45.5)  <.001  Physical activity, kcals  473 (150–1,110)  907 (907–1,967)  598 (187–1,417)  <.001  BMI, kg/m2  26.9 (5.2)  26.6 (3.8)  26.8 (4.7)  .07  Waist, cm  95.9 (14.6)  98.9 (10.5)  97.1 (13.2)  <.001  History of CHDa  339 (16.8)  359 (27.6)  698 (21.0)  <.001  History of CHF  91 (4.5)  66 (5.1)  157 (4.7)  .45  History of stroke  67 (3.3)  93 (7.2)  160 (4.8)  <.001  Adiponectin, mg/L  14.1 (10.2–19.6)  9.6 (7.0–13.6)  12.2 (8.5–17.5)  <.001  NT-proBNP, pg/mL  143 (76–257)  128 (60–306)  137 (69–268)  .11  DHEAS, µg/mL  0.52 (0.32–0.80)  0.79 (0.48–1.18)  0.61 (0.38–0.94)  <.001  IL-6, pg/mL  2.58 (1.75–3.91)  2.87 (2.02–4.34)  2.70 (1.85–4.08)  <.001  IGF-1, µg/L  89 (70–112.0)  103 (82–126)  94 (74–118)  <.001  IGFBP3, µg/L  3,725 (3,135–4,281)  3,349 (2,836–3,918)  3,565 (3,009–4,162)  <.001  Note: BMI = body mass index; CHD = coronary heart disease; CHF = congestive heart failure; DHEAS = dehydroepiandrosterone sulfate; IGF-1 = Insulin-like growth factor-1; IGFBP3 = IGF-binding protein-3; IL-6 = interleukin-6; NT-proBNP = amino-terminal pro-B-type natriuretic peptide. Entries in table are mean (SD), N (%), or median (interquartile range). p Values comparing men and women are from t test, chi-square test, or Kruskal–Wallis test, corresponding to whether or not the mean, N, or median are reported. aDefined as angina, angioplasty, bypass surgery, or myocardial Infarction. View Large Table 2. Biomarker Index Scoring   Women  Men  Biomarker Quintile  Biomarker Range  β Coefficient  Pointsa  Biomarker Range  β Coefficient  Pointsa  NT-proBNP, pg/mL   0  ≤76  0  0  ≤62  0  0   1  78–127  0.2173  1  63–121  0.1673  1   2  128–199  0.2324  1  122–211  0.3381  1   3  200–355  0.5158  2  212–485  0.6764  3   4  >355  0.8539  3  >485  1.141  5  IL-6, pg/mL   0  ≤1.8  0  0  ≤2.0  0  0   1  1.9–2.5  0.1333  0  2.1–2.7  0.1577  1   2  2.6–3.3  0.3954  1  2.8–3.6  0.1863  1   3  3.4–4.6  0.4857  2  3.7–5.1  0.4775  2   4  >4.6  0.7344  3  >5.1  0.5773  2  IGF-1, µg/L   0  ≤66  0.1144  0  ≤75    0   1  67–80  0.1014  0  76–94    0   2  81–96  0.1284  0  95–111    0   3  97–120  0  0  112–1,332    0   4  >120  0.3517  1  >133    0  IGFBP3, µg/L   0  ≤2,951  0.3066  1  ≤2,621  0.2073  1   1  2,952–3,436  0  0  2,622–3,100  0.2306  1   2  3,437–3,881  0.0049  0  3,101–3,495  0.2000  1   3  3,882–4,379  0.1763  1  3,496–4,061  0  0   4  >4,379  0.1287  1  >4,061  0.4038  2  DHEAS, µg/mL   0  ≤0.26  0.2547  1  ≤0.41    0   1  0.27–0.42  0.1520  1  0.42–0.61    0   2  0.43–0.61  0  0  0.62–0.87    0   3  0.62–0.90  0.0255  0  0.88–1.24    0   4  >0.90  0.2674  1  >1.24    0    Women  Men  Biomarker Quintile  Biomarker Range  β Coefficient  Pointsa  Biomarker Range  β Coefficient  Pointsa  NT-proBNP, pg/mL   0  ≤76  0  0  ≤62  0  0   1  78–127  0.2173  1  63–121  0.1673  1   2  128–199  0.2324  1  122–211  0.3381  1   3  200–355  0.5158  2  212–485  0.6764  3   4  >355  0.8539  3  >485  1.141  5  IL-6, pg/mL   0  ≤1.8  0  0  ≤2.0  0  0   1  1.9–2.5  0.1333  0  2.1–2.7  0.1577  1   2  2.6–3.3  0.3954  1  2.8–3.6  0.1863  1   3  3.4–4.6  0.4857  2  3.7–5.1  0.4775  2   4  >4.6  0.7344  3  >5.1  0.5773  2  IGF-1, µg/L   0  ≤66  0.1144  0  ≤75    0   1  67–80  0.1014  0  76–94    0   2  81–96  0.1284  0  95–111    0   3  97–120  0  0  112–1,332    0   4  >120  0.3517  1  >133    0  IGFBP3, µg/L   0  ≤2,951  0.3066  1  ≤2,621  0.2073  1   1  2,952–3,436  0  0  2,622–3,100  0.2306  1   2  3,437–3,881  0.0049  0  3,101–3,495  0.2000  1   3  3,882–4,379  0.1763  1  3,496–4,061  0  0   4  >4,379  0.1287  1  >4,061  0.4038  2  DHEAS, µg/mL   0  ≤0.26  0.2547  1  ≤0.41    0   1  0.27–0.42  0.1520  1  0.42–0.61    0   2  0.43–0.61  0  0  0.62–0.87    0   3  0.62–0.90  0.0255  0  0.88–1.24    0   4  >0.90  0.2674  1  >1.24    0  Note: DHEAS = dehydroepiandrosterone sulfate; IGF-1 = Insulin-like growth factor-1; IGFBP3 = IGF-binding protein-3; IL-6 = interleukin-6; NT-proBNP = amino-terminal pro-B-type natriuretic peptide. aPoints determined by dividing the β coefficient by three times the β coefficient for age = 0.0912. IL-6 = interleukin-6. View Large Table 2. Biomarker Index Scoring   Women  Men  Biomarker Quintile  Biomarker Range  β Coefficient  Pointsa  Biomarker Range  β Coefficient  Pointsa  NT-proBNP, pg/mL   0  ≤76  0  0  ≤62  0  0   1  78–127  0.2173  1  63–121  0.1673  1   2  128–199  0.2324  1  122–211  0.3381  1   3  200–355  0.5158  2  212–485  0.6764  3   4  >355  0.8539  3  >485  1.141  5  IL-6, pg/mL   0  ≤1.8  0  0  ≤2.0  0  0   1  1.9–2.5  0.1333  0  2.1–2.7  0.1577  1   2  2.6–3.3  0.3954  1  2.8–3.6  0.1863  1   3  3.4–4.6  0.4857  2  3.7–5.1  0.4775  2   4  >4.6  0.7344  3  >5.1  0.5773  2  IGF-1, µg/L   0  ≤66  0.1144  0  ≤75    0   1  67–80  0.1014  0  76–94    0   2  81–96  0.1284  0  95–111    0   3  97–120  0  0  112–1,332    0   4  >120  0.3517  1  >133    0  IGFBP3, µg/L   0  ≤2,951  0.3066  1  ≤2,621  0.2073  1   1  2,952–3,436  0  0  2,622–3,100  0.2306  1   2  3,437–3,881  0.0049  0  3,101–3,495  0.2000  1   3  3,882–4,379  0.1763  1  3,496–4,061  0  0   4  >4,379  0.1287  1  >4,061  0.4038  2  DHEAS, µg/mL   0  ≤0.26  0.2547  1  ≤0.41    0   1  0.27–0.42  0.1520  1  0.42–0.61    0   2  0.43–0.61  0  0  0.62–0.87    0   3  0.62–0.90  0.0255  0  0.88–1.24    0   4  >0.90  0.2674  1  >1.24    0    Women  Men  Biomarker Quintile  Biomarker Range  β Coefficient  Pointsa  Biomarker Range  β Coefficient  Pointsa  NT-proBNP, pg/mL   0  ≤76  0  0  ≤62  0  0   1  78–127  0.2173  1  63–121  0.1673  1   2  128–199  0.2324  1  122–211  0.3381  1   3  200–355  0.5158  2  212–485  0.6764  3   4  >355  0.8539  3  >485  1.141  5  IL-6, pg/mL   0  ≤1.8  0  0  ≤2.0  0  0   1  1.9–2.5  0.1333  0  2.1–2.7  0.1577  1   2  2.6–3.3  0.3954  1  2.8–3.6  0.1863  1   3  3.4–4.6  0.4857  2  3.7–5.1  0.4775  2   4  >4.6  0.7344  3  >5.1  0.5773  2  IGF-1, µg/L   0  ≤66  0.1144  0  ≤75    0   1  67–80  0.1014  0  76–94    0   2  81–96  0.1284  0  95–111    0   3  97–120  0  0  112–1,332    0   4  >120  0.3517  1  >133    0  IGFBP3, µg/L   0  ≤2,951  0.3066  1  ≤2,621  0.2073  1   1  2,952–3,436  0  0  2,622–3,100  0.2306  1   2  3,437–3,881  0.0049  0  3,101–3,495  0.2000  1   3  3,882–4,379  0.1763  1  3,496–4,061  0  0   4  >4,379  0.1287  1  >4,061  0.4038  2  DHEAS, µg/mL   0  ≤0.26  0.2547  1  ≤0.41    0   1  0.27–0.42  0.1520  1  0.42–0.61    0   2  0.43–0.61  0  0  0.62–0.87    0   3  0.62–0.90  0.0255  0  0.88–1.24    0   4  >0.90  0.2674  1  >1.24    0  Note: DHEAS = dehydroepiandrosterone sulfate; IGF-1 = Insulin-like growth factor-1; IGFBP3 = IGF-binding protein-3; IL-6 = interleukin-6; NT-proBNP = amino-terminal pro-B-type natriuretic peptide. aPoints determined by dividing the β coefficient by three times the β coefficient for age = 0.0912. IL-6 = interleukin-6. View Large The associations of the Biomarker and Physiologic Indices with mortality were calculated using Cox proportional hazards models with each index modeled linearly among participants with both indices (N = 2,515). A likelihood ratio test was used to confirm that there was no improvement over the linear model by a model based on indicator variables for categories of the Biomarker Index. The proportional hazard assumption was tested with Schoenfeld residuals, and when violated, baseline hazards were stratified by sex and follow-up was truncated at 10 years to conform to the assumption. Models were built sequentially adjusting for age, sex, race, behaviors (current smoking, diet, any alcohol consumption, and ln(kcal) physical activity), body size (body mass index and waist circumference), prevalent clinical CVD or cancer, and, finally, incident CVD, cancer, or hospitalization for pneumonia during follow-up. All analyses were performed with STATA 14.2 (College Station, TX) or R version 2.13.0 (http://www.r-project.org). Results The mean (SD) age of the participants with all biomarkers (N = 3,321) was 74.5 (5.1) years, and 14.1% were black (Table 1). Compared with men, women were less likely to drink alcohol or have a history of coronary heart disease or stroke. Women had higher levels of adiponectin and IGFBP-3; lower levels of DHEAS, IL-6, and IGF-1; and similar NT-proBNP (Table 1). Among the 3,321 participants, there were 2,536 deaths over 19.5 years, and among the 2,525 participants with both indices, there were 1,003 deaths over 10 years. Cubic spline plots of the associations of the biomarkers with mortality in the test set showed departures from linearity, although data were sparse in the extremes of the distributions, as evidenced by the wide confidence intervals. Scoring of each biomarker is shown in Table 2. When the scores for each biomarker were summed into the total Biomarker Index, there was a clear gradation of mortality across the Biomarker Index score, with lower scores associated with lower mortality and higher scores associated with higher mortality (Table 3). Table 3. Incidence Rates for Mortality by Biomarker Index Score Biomarker Index  N  Person-Years  Number of Deaths  Incidence Ratea (95% CI)  0  77  720  15  20.8 (12.6, 34.5)  1  296  2,754  53  19.2 (14.7, 25.2)  2  610  5,574  147  26.4 (22.4, 31.0)  3  718  6,144  260  42.3 (37.5, 47.8)  4  576  4,754  243  51.1 (45.1, 58.0)  5  417  3,222  208  64.6 (56.3, 73.9)  6  287  2,053  178  86.7 (74.9, 100.4)  7  194  1,232  136  110.4 (83.3, 130.6)  8  126  627  106  169.1 (140.8, 204.5)  9  20  87  19  219.3 (139.9, 343.8)  Biomarker Index  N  Person-Years  Number of Deaths  Incidence Ratea (95% CI)  0  77  720  15  20.8 (12.6, 34.5)  1  296  2,754  53  19.2 (14.7, 25.2)  2  610  5,574  147  26.4 (22.4, 31.0)  3  718  6,144  260  42.3 (37.5, 47.8)  4  576  4,754  243  51.1 (45.1, 58.0)  5  417  3,222  208  64.6 (56.3, 73.9)  6  287  2,053  178  86.7 (74.9, 100.4)  7  194  1,232  136  110.4 (83.3, 130.6)  8  126  627  106  169.1 (140.8, 204.5)  9  20  87  19  219.3 (139.9, 343.8)  Note: CI = confidence interval. Wide confidence intervals for values of 0 and 9 with few deaths. aPer thousand person-years based on 10 years of follow-up. View Large Table 3. Incidence Rates for Mortality by Biomarker Index Score Biomarker Index  N  Person-Years  Number of Deaths  Incidence Ratea (95% CI)  0  77  720  15  20.8 (12.6, 34.5)  1  296  2,754  53  19.2 (14.7, 25.2)  2  610  5,574  147  26.4 (22.4, 31.0)  3  718  6,144  260  42.3 (37.5, 47.8)  4  576  4,754  243  51.1 (45.1, 58.0)  5  417  3,222  208  64.6 (56.3, 73.9)  6  287  2,053  178  86.7 (74.9, 100.4)  7  194  1,232  136  110.4 (83.3, 130.6)  8  126  627  106  169.1 (140.8, 204.5)  9  20  87  19  219.3 (139.9, 343.8)  Biomarker Index  N  Person-Years  Number of Deaths  Incidence Ratea (95% CI)  0  77  720  15  20.8 (12.6, 34.5)  1  296  2,754  53  19.2 (14.7, 25.2)  2  610  5,574  147  26.4 (22.4, 31.0)  3  718  6,144  260  42.3 (37.5, 47.8)  4  576  4,754  243  51.1 (45.1, 58.0)  5  417  3,222  208  64.6 (56.3, 73.9)  6  287  2,053  178  86.7 (74.9, 100.4)  7  194  1,232  136  110.4 (83.3, 130.6)  8  126  627  106  169.1 (140.8, 204.5)  9  20  87  19  219.3 (139.9, 343.8)  Note: CI = confidence interval. Wide confidence intervals for values of 0 and 9 with few deaths. aPer thousand person-years based on 10 years of follow-up. View Large The Biomarker Index was similarly distributed in the test and validation samples with a slight right skew (Figure 1, chi-square p value = .75). Estimates of the age, sex, and race adjusted hazard ratios (HRs) and 95% confidence intervals of the Biomarker Index with mortality were identical in the test (1.29 [1.24, 1.33]) and validation (1.29 [1.23, 1.35]) subsets, and measures of concordance were similar (Harrell’s C-statistic = 0.699 and 0.708 in the test and validation sets, respectively). Figure 1. View largeDownload slide Distribution of Biomarker Index score in test and validation samples of the Cardiovascular Health Study. Figure 1. View largeDownload slide Distribution of Biomarker Index score in test and validation samples of the Cardiovascular Health Study. Compared with participants with data on both indices, participants with only data for the Biomarker Index were on average of similar age and gender, but were more likely African American, had a larger waist circumference, had lower physical activity, and were more likely to have coronary heart disease or congestive heart failure, as well as slightly higher NT-proBNP and IL-6 (Supplementary Table 1). Among the 2,515 participants with data on both indices, combining the test and validation subsets, the mean (SD) of the Biomarker Index and Physiologic Index was 3.6 (1.9) and 4.4 (2.1), respectively. The indices were significantly correlated (Spearman correlation coefficient = 0.35, p < .0001). In separate unadjusted Cox models with the baseline hazard stratified by sex, the estimated HRs for the two indices were nearly identical at 1.34 for the Biomarker Index and 1.36 for the Physiologic Index (Table 4) with respective measures of concordance of 0.66 and 0.67. For the unadjusted Biomarker Index (Table 4, Model 1), the C-statistic was 0.656 and area under the curve was 0.685. For the unadjusted Physiologic Index, the C-statistic was 0.669 and area under the curve was 0.714 (test of comparison p value of .02). The HRs were attenuated after adjustment for age and for each other, but remained similar and statistically significant in all adjusted models. Table 4. Hazard Ratios for Death by Age and the Two Indices, With Various Levels of Adjustment Adjustment Variables  Age (per Year)  Physiologic Index (per 1 Index Point)  Biomarker Index (per 1 Index Point)  Sex  1.11 (1.10, 1.13)  1.36 (1.31, 1.40)  1.34 (1.30, 1.38)  Sex, race, and agea  1.08 (1.07, 1.09)  1.28 (1.23, 1.32)    1.09 (1.08, 1.10)    1.27 (1.23, 1.32)  + Both indices  1.07 (1.06, 1.08)  1.22 (1.18, 1.26)  1.21 (1.17, 1.25)  + Behaviors  1.07 (1.06, 1.09)  1.20 (1.15, 1.24)  1.20 (1.16, 1.24)  + Body size  1.06 (1.05, 1.08)  1.21 (1.17, 1.26)  1.19 (1.15, 1.23)  + Prevalent clinical CVD or cancer  1.06 (1.05, 1.08)  1.21 (1.16, 1.25)  1.18 (1.13, 1.22)  + Incident CVD, cancer, or hospitalization for pneumonia  1.05 (1.03, 1.07)  1.20 (1.15, 1.25)  1.19 (1.14, 1.24)  Adjustment Variables  Age (per Year)  Physiologic Index (per 1 Index Point)  Biomarker Index (per 1 Index Point)  Sex  1.11 (1.10, 1.13)  1.36 (1.31, 1.40)  1.34 (1.30, 1.38)  Sex, race, and agea  1.08 (1.07, 1.09)  1.28 (1.23, 1.32)    1.09 (1.08, 1.10)    1.27 (1.23, 1.32)  + Both indices  1.07 (1.06, 1.08)  1.22 (1.18, 1.26)  1.21 (1.17, 1.25)  + Behaviors  1.07 (1.06, 1.09)  1.20 (1.15, 1.24)  1.20 (1.16, 1.24)  + Body size  1.06 (1.05, 1.08)  1.21 (1.17, 1.26)  1.19 (1.15, 1.23)  + Prevalent clinical CVD or cancer  1.06 (1.05, 1.08)  1.21 (1.16, 1.25)  1.18 (1.13, 1.22)  + Incident CVD, cancer, or hospitalization for pneumonia  1.05 (1.03, 1.07)  1.20 (1.15, 1.25)  1.19 (1.14, 1.24)  Note: CVD = cardiovascular disease. The + symbol in the table indicates that each model with the + includes variables included in the prior model. aFirst line is adjusted for the Physiologic Index and second line for the Biomarker Index. View Large Table 4. Hazard Ratios for Death by Age and the Two Indices, With Various Levels of Adjustment Adjustment Variables  Age (per Year)  Physiologic Index (per 1 Index Point)  Biomarker Index (per 1 Index Point)  Sex  1.11 (1.10, 1.13)  1.36 (1.31, 1.40)  1.34 (1.30, 1.38)  Sex, race, and agea  1.08 (1.07, 1.09)  1.28 (1.23, 1.32)    1.09 (1.08, 1.10)    1.27 (1.23, 1.32)  + Both indices  1.07 (1.06, 1.08)  1.22 (1.18, 1.26)  1.21 (1.17, 1.25)  + Behaviors  1.07 (1.06, 1.09)  1.20 (1.15, 1.24)  1.20 (1.16, 1.24)  + Body size  1.06 (1.05, 1.08)  1.21 (1.17, 1.26)  1.19 (1.15, 1.23)  + Prevalent clinical CVD or cancer  1.06 (1.05, 1.08)  1.21 (1.16, 1.25)  1.18 (1.13, 1.22)  + Incident CVD, cancer, or hospitalization for pneumonia  1.05 (1.03, 1.07)  1.20 (1.15, 1.25)  1.19 (1.14, 1.24)  Adjustment Variables  Age (per Year)  Physiologic Index (per 1 Index Point)  Biomarker Index (per 1 Index Point)  Sex  1.11 (1.10, 1.13)  1.36 (1.31, 1.40)  1.34 (1.30, 1.38)  Sex, race, and agea  1.08 (1.07, 1.09)  1.28 (1.23, 1.32)    1.09 (1.08, 1.10)    1.27 (1.23, 1.32)  + Both indices  1.07 (1.06, 1.08)  1.22 (1.18, 1.26)  1.21 (1.17, 1.25)  + Behaviors  1.07 (1.06, 1.09)  1.20 (1.15, 1.24)  1.20 (1.16, 1.24)  + Body size  1.06 (1.05, 1.08)  1.21 (1.17, 1.26)  1.19 (1.15, 1.23)  + Prevalent clinical CVD or cancer  1.06 (1.05, 1.08)  1.21 (1.16, 1.25)  1.18 (1.13, 1.22)  + Incident CVD, cancer, or hospitalization for pneumonia  1.05 (1.03, 1.07)  1.20 (1.15, 1.25)  1.19 (1.14, 1.24)  Note: CVD = cardiovascular disease. The + symbol in the table indicates that each model with the + includes variables included in the prior model. aFirst line is adjusted for the Physiologic Index and second line for the Biomarker Index. View Large In the model with only age, the HR per year of age was 1.11 (1.10, 1.13; Table 4). Adjusting for the Biomarker Index reduced the HR for age to 1.09 (1.08, 1.10; 20% attenuation in age beta) and adjusting for the Physiologic Index reduced the HR for age to 1.08 (1.07, 1.09; 29% attenuation in age beta). Adjusting for both indices, the HR for age was 1.07 (1.06, 1.08; 39% attenuation in age beta); the HR for the Biomarker Index was 1.21 (1.17, 1.25); and the HR for the Physiologic Index was 1.22 (1.18, 1.26). Associations remained similar with additional adjustment for health behaviors, body size, and clinical CVD, cancer, and pneumonia hospitalization (Figure 2). Figure 2. View largeDownload slide Kaplan–Meier curve of cumulative mortality over the follow-up period by score of Biomarker Index. Figure 2. View largeDownload slide Kaplan–Meier curve of cumulative mortality over the follow-up period by score of Biomarker Index. Discussion In this sample of community-dwelling older adults, we show that the relative risk of mortality associated with an index of five new circulating biomarkers is similar to that of an index using physiologic markers that require more resource intensive ascertainment but that both sets of variables remain significant predictors of mortality even in the presence of each other. The Biomarker and Physiologic Indices attenuate age substantially and independently of each other, suggesting that each captures a part of biological aging that is distinct. Previously, we showed that the Physiologic Index and variations thereof are associated with death, disability, age-related chronic disease, frailty and vigor, telomere length, and many health behaviors in the Cardiovascular Health Study and Health, Aging, and Body Composition Study (1,2,4,13). In the Long Life Family Study, a family-based study of exceptional longevity, a variation of the Physiologic Index termed the Healthy Aging Index was also moderately and significantly heritable (14). Supporting prior analyses, the present data suggest that the two indices represent distinct aspects of the aging process. Thus, they may be feasible outcomes for benchmarking interventions being tested to alter the trajectory of aging. What aspects of aging biology may be reflected by the Biomarker Index? IL-6 is a cytokine produced by immune cells, vascular endothelium, adipose tissue, and muscle. It has proinflammatory and anti-inflammatory effects, rises with age in the absence of disease, and is robustly associated with myriad health outcomes (15,16). NT-proBNP is released primarily by ventricular myocytes in response to elevated filling pressure, acting hormonally on the kidneys to relieve cardiac stress through sodium and water excretion. NT-proBNP elevations occur in a number of cardiovascular aging and disease phenotypes, including coronary artery disease and myocardial infarction, left ventricular hypertrophy, atrial fibrillation, and valvular diseases, reflecting increased myocardial stretch. Higher NT-proBNP levels also occur with decreased renal clearance of NT-proBNP and are seen with advancing age (17,18). Various studies have shown NT-proBNP to be independently associated with mortality (19,20). Thus, NT-proBNP may be a marker of undiagnosed or subclinical cardiovascular damage that occurs with aging and disease, as well as co-existing renal dysfunction. DHEAS can be thought of as a pool for sex steroid hormones. More than 90% of estrogens in postmenopausal women and 30% of androgens in men are derived from DHEAS (21). DHEAS levels peak at birth and in the third decade of life and fall by 80%–90% by age 80, which partly explains the age-associated decline in estrogen and testosterone (22–24). The cause of this age-associated decline is unknown, although we reported that in older adults in the CHS All Stars study, CVD, male gender, and black race were associated with greater DHEAS decline over 9 years (25). Epidemiological data on an association between DHEAS and mortality are equivocal (26–29). The IGF pathway was the first pathway shown to influence aging and life span in animals, conceivably through growth and development and metabolic regulation (30). Studies in model systems are robust and suggest a causal role for IGF-1 and longevity in nonhuman systems, and cross-sectional epidemiologic studies of community-dwelling older adults confirmed that they have lower IGF-1 levels compared with younger persons (31–36). Nonetheless, although the Framingham Heart Study showed that greater baseline IGF-1 was associated with lower mortality (37), no association was initially detected in five other large cohort studies (10,38–41). IGF-1 was not strongly scored in the Biomarker Index, suggesting that at these ages it may be associated with mortality only in the extremes or in the context of other biomarker levels. We acknowledge several limitations in this analysis. First, the Biomarker Index was specifically scored by weighting its components using regression coefficients from Cox models of death. This methodology was not used to score the Physiologic Index whose components were monotonically associated with mortality and were scored with tertiles summed across the components. Consequently, the Physiologic Index may have been “disadvantaged” compared with the Biomarker Index in predicting death. However, we derived the Biomarker Index scoring in a randomly selected test sample and used it to predict death in a validation sample, where it performed similarly to the test sample, suggesting it was not overcalibrated. Second, the choice of biomarkers to include in the Biomarker Index was made from convenience in that the biomarkers were available in the CHS. However, we had good scientific rationale for including each candidate biomarker in the index. Before inclusion, each biomarker’s statistical appropriateness was additionally confirmed (ie, association with mortality independent of age and sex). In the future, high-throughput technology could allow affordable prospective measurement of hundreds or thousands of biomarkers simultaneously, which may increase predictive power and suggest new mechanistic associations. The challenge of incorporating large numbers of biomarkers into clinical practice may eventually be overcome by designing targeted “aging” panels using validated biomarkers, similar to targeted genetic prediction panels already used in clinical oncology. Third, we measured each biomarker only once and levels may have changed throughout time. Previous analyses suggest it is unlikely that participants improved significantly on any component of the Physiologic Index, but there is evidence that biomarkers can have varied trajectories that themselves may predict death better than a single measurement (42,43). Fourth, residual confounding may be present if confounders were not included in our models, such as dietary habits, although adjustment for major health behaviors and clinical disease minimally altered results. Fifth, although some different associations have been detected previously in CHS for some of these biomarkers (eg, association between adiponectin and mortality varying by CVD status (44)), the aim here was to find a biomarker generally predictive of mortality irrespective of disease status. Differences in study population and analytic design may have influenced results compared with prior reports. In conclusion, an index of five circulating biomarkers can predict death in older community-dwelling adults as accurately as an index of physiologic markers, and both indices substantially and independently attenuate the effect of age on mortality. Although the Biomarker Index needs further validation, these data suggest that Biomarker Index and the Physiologic Index measure important aspects of the underlying aging process, and they may serve as end points for studies seeking to develop interventions to improve aging. Supplementary Material Supplementary data is available at The Journals of Gerontology, Series A: Biological Sciences and Medical Sciences online. Funding This research was supported by contracts HHSN268201200036C, HHSN268200800007C, N01 HC55222, N01HC85079 through N01HC85083, N01HC85086, and grant HL080295 from the National Heart, Lung, and Blood Institute (NHLBI), with additional contribution from the National Institute of Neurological Disorders and Stroke (NINDS). Additional support was provided by AG023629, AG-15928, AG-20098, and AG-027058 from the National Institute on Aging (NIA). IGF-1 measurement was supported by 1R01AG031890 to R.C.K. A full list of principal CHS investigators and institutions can be found at CHS-NHLBI.org. Conflict of Interest None reported. Acknowledgments We thank Liana del Gobbo and Dariush Mozaffarian for their prior work on dietary indices in CHS that allowed calculation of the Alternative Healthy Eating Index. References 1. Sanders JL, Boudreau RM, Penninx BW, et al.  ; Health ABC Study. Association of a modified physiologic index with mortality and incident disability: the Health, Aging, and Body Composition Study. J Gerontol A Biol Sci Med Sci . 2012; 67: 1439– 1446. doi: 10.1093/gerona/gls123 Google Scholar CrossRef Search ADS PubMed  2. Newman AB, Boudreau RM, Naydeck BL, Fried LF, Harris TB. A physiologic index of comorbidity: relationship to mortality and disability. J Gerontol A Biol Sci Med Sci . 2008; 63: 603– 609. Google Scholar CrossRef Search ADS PubMed  3. Sanders JL, Boudreau RM, Fried LP, Walston JD, Harris TB, Newman AB. Measurement of organ structure and function enhances understanding of the physiological basis of frailty: the Cardiovascular Health Study. J Am Geriatr Soc . 2011; 59: 1581– 1588. doi: 10.1111/j.1532-5415.2011.03557.x Google Scholar CrossRef Search ADS PubMed  4. Sanders JL, Fitzpatrick AL, Boudreau RM, et al.   Leukocyte telomere length is associated with noninvasively measured age-related disease: the Cardiovascular Health Study. J Gerontol A Biol Sci Med Sci . 2012; 67: 409– 416. doi: 10.1093/gerona/glr173 Google Scholar CrossRef Search ADS PubMed  5. Sanders JL, Boudreau R, Newman A. Understanding the aging process using epidemiologic approaches. In: Newman A, Cauley J, eds. Epidemiology of Aging . Dordrecht, The Netherlands: Springer; 2012: 187– 214. Google Scholar CrossRef Search ADS   6. Sprott RL. Biomarkers of aging and disease: introduction and definitions. Exp Gerontol . 2010; 45: 2– 4. doi: 10.1016/j.exger.2009.07.008 Google Scholar CrossRef Search ADS PubMed  7. Newman AB, Arnold AM, Sachs MC, et al.   Long-term function in an older cohort – the Cardiovascular Health Study All Stars study. J Am Geriatr Soc . 2009; 57: 432– 440. doi: 10.1111/j.1532-5415.2008.02152.x Google Scholar CrossRef Search ADS PubMed  8. Fried LP, Borhani NO, Enright P, et al.   The Cardiovascular Health Study: design and rationale. Ann Epidemiol . 1991; 1: 263– 276. Google Scholar CrossRef Search ADS PubMed  9. Cushman M, Cornell ES, Howard PR, Bovill EG, Tracy RP. Laboratory methods and quality assurance in the Cardiovascular Health Study. Clin Chem . 1995; 41: 264– 270. Google Scholar PubMed  10. Kaplan RC, McGinn AP, Pollak MN, et al.   Total insulinlike growth factor 1 and insulin-like growth factor binding protein levels, functional status, and mortality in older adults. J Am Geriatr Soc . 2008; 56: 652– 660. doi: 10.1111/j.1532-5415.2007.01637.x Google Scholar CrossRef Search ADS PubMed  11. Del Gobbo LC, Kalantarian S, Imamura F, et al.   Contribution of major lifestyle risk factors for incident heart failure in older adults: the Cardiovascular Health Study. JACC Heart Fail . 2015; 3: 520– 528. doi: 10.1016/j.jchf.2015.02.009 Google Scholar CrossRef Search ADS PubMed  12. Psaty BM, Kuller LH, Bild D, et al.   Methods of assessing prevalent cardiovascular disease in the Cardiovascular Health Study. Ann Epidemiol . 1995; 5: 270– 277. Google Scholar CrossRef Search ADS PubMed  13. Sanders JL, Boudreau RM, Fried LP, Walston JD, Harris TB, Newman AB. Measurement of organ structure and function enhances understanding of the physiological basis of frailty: the Cardiovascular Health Study. J Am Geriatr Soc . 2011; 59: 1581– 1588. doi: 10.1111/j.1532-5415.2011.03557.x Google Scholar CrossRef Search ADS PubMed  14. Sanders JL, Minster RL, Barmada MM, et al.   Heritability of and mortality prediction with a longevity phenotype: the healthy aging index. J Gerontol A Biol Sci Med Sci . 2014; 69: 479– 485. doi: 10.1093/gerona/glt117 Google Scholar CrossRef Search ADS PubMed  15. Singh T, Newman AB. Inflammatory markers in population studies of aging. Ageing Res Rev . 2011; 10: 319– 329. doi: 10.1016/j.arr.2010.11.002 Google Scholar CrossRef Search ADS PubMed  16. Maggio M, Guralnik JM, Longo DL, Ferrucci L. Interleukin-6 in aging and chronic disease: a magnificent pathway. J Gerontol A Biol Sci Med Sci . 2006; 61: 575– 584. Google Scholar CrossRef Search ADS PubMed  17. Redfield MM, Rodeheffer RJ, Jacobsen SJ, Mahoney DW, Bailey KR, Burnett JCJr. Plasma brain natriuretic peptide concentration: impact of age and gender. J Am Coll Cardiol . 2002; 40: 976– 982. Google Scholar CrossRef Search ADS PubMed  18. Charloux A, Brandenberger G, Piquard F, Geny B. Dysregulation of pulsatility in aging IV. Pulsatile signaling and cardiovascular aging: functions and regulation of natriuretic peptide signaling. Ageing Res Rev . 2008; 7: 151– 163. doi: 10.1016/j.arr.2007.12.001 Google Scholar CrossRef Search ADS PubMed  19. Witham MD, Gillespie ND, Hutcheon SD, Struthers AD, McMurdo ME. B-type natriuretic peptide is associated with mortality in older functionally impaired patients. J Am Geriatr Soc . 2005; 53: 1991– 1995. doi: 10.1111/j.1532-5415.2005.53555.x Google Scholar CrossRef Search ADS PubMed  20. Beleigoli AM, Boersma E, Diniz MdFH, Vidigal PG, Lima-Costa MF, Ribeiro AL. C-reactive protein and B-type natriuretic peptide yield either a non-significant or a modest incremental value to traditional risk factors in predicting long-term overall mortality in older adults. PLoS One . 2013; 8: e75809. doi: 10.1371/journal.pone.0075809 Google Scholar CrossRef Search ADS PubMed  21. Labrie F, Bélanger A, Simard J, Van L-T, Labrie C. DHEA and peripheral androgen and estrogen formation: intracinology. Ann NY Acad Sci . 1995; 774: 16– 28. Google Scholar CrossRef Search ADS PubMed  22. Longcope C. Metabolism of dehydroepiandrosterone. Ann NY Acad Sci . 1995; 774: 143– 148. Google Scholar CrossRef Search ADS PubMed  23. Orentreich N, Brind JL, Rizer RL, Vogelman JH. Age changes and sex differences in serum dehydroepiandrosterone sulfate concentrations throughout adulthood. J Clin Endocrinol Metab . 1984; 59: 551– 555. doi: 10.1210/jcem-59-3-551 Google Scholar CrossRef Search ADS PubMed  24. Vermeulen A. Dehydroepiandrosterone sulfate and aging. Ann NY Acad Sci . 1995; 774: 121– 127. Google Scholar CrossRef Search ADS PubMed  25. Sanders JL, Boudreau RM, Cappola AR, et al.   Cardiovascular disease is associated with greater incident dehydroepiandrosterone sulfate decline in the oldest old: the Cardiovascular Health Study All Stars study. J Am Geriatr Soc . 2010; 58: 421– 426. doi: 10.1111/j.1532-5415.2010.02724.x Google Scholar CrossRef Search ADS PubMed  26. Tchernof A, Labrie F. Dehydroepiandrosterone, obesity and cardiovascular disease risk: a review of human studies. Eur J Endocrinol . 2004; 151: 1– 14. Google Scholar CrossRef Search ADS PubMed  27. Maggio M, Lauretani F, Ceda GP, et al.   Relationship between low levels of anabolic hormones and 6-year mortality in older men: the Aging in the Chianti Area (InCHIANTI) study. Arch Intern Med . 2007; 167: 2249– 2254. doi: 10.1001/archinte.167.20.2249 Google Scholar CrossRef Search ADS PubMed  28. Glei DA, Goldman N. Dehydroepiandrosterone sulfate (DHEAS) and risk for mortality among older Taiwanese. Ann Epidemiol . 2006; 16: 510– 515. doi: 10.1016/j.annepidem.2005.09.010 Google Scholar CrossRef Search ADS PubMed  29. Enomoto M, Adachi H, Fukami A, et al.   Serum dehydroepiandrosterone sulfate levels predict longevity in men: 27-year follow-up study in a community-based cohort (Tanushimaru study). J Am Geriatr Soc . 2008; 56: 994– 998. doi: 10.1111/j.1532-5415.2008.01692.x Google Scholar CrossRef Search ADS PubMed  30. Kenyon C. The plasticity of aging: insights from long-lived mutants. Cell . 2005; 120: 449– 460. doi: 10.1016/j.cell.2005.02.002 Google Scholar CrossRef Search ADS PubMed  31. Landin-Wilhelmsen K, Wilhelmsen L, Lappas G, et al.   Serum insulin-like growth factor I in a random population sample of men and women: relation to age, sex, smoking habits, coffee consumption and physical activity, blood pressure and concentrations of plasma lipids, fibrinogen, parathyroid hormone and osteocalcin. Clin Endocrinol (Oxf) . 1994; 41: 351– 357. Google Scholar CrossRef Search ADS PubMed  32. Papadakis MA, Grady D, Tierney MJ, Black D, Wells L, Grunfeld C. Insulin-like growth factor 1 and functional status in healthy older men. J Am Geriatr Soc . 1995; 43: 1350– 1355. Google Scholar CrossRef Search ADS PubMed  33. Goodman-Gruen D, Barrett-Connor E. Epidemiology of insulin-like growth factor-I in elderly men and women: the Rancho Bernardo Study. Am J Epidemiol . 1997; 145: 970– 976. Google Scholar CrossRef Search ADS PubMed  34. Harris TB, Kiel D, Roubenoff R, et al.   Association of insulin-like growth factor-I with body composition, weight history, and past health behaviors in the very old: the Framingham Heart Study. J Am Geriatr Soc . 1997; 45: 133– 139. Google Scholar CrossRef Search ADS PubMed  35. O’Connor KG, Tobin JD, Harman SM, et al.   Serum levels of insulin-like growth factor-I are related to age and not to body composition in healthy women and men. J Gerontol A Biol Sci Med Sci . 1998; 53: M176– M182. Google Scholar CrossRef Search ADS PubMed  36. Boonen S, Lysens R, Verbeke G, et al.   Relationship between age-associated endocrine deficiencies and muscle function in elderly women: a cross-sectional study. Age Ageing . 1998; 27: 449– 454. Google Scholar CrossRef Search ADS PubMed  37. Roubenoff R, Parise H, Payette HA, et al.   Cytokines, insulin-like growth factor 1, sarcopenia, and mortality in very old community-dwelling men and women: the Framingham Heart Study. Am J Med . 2003; 115: 429– 435. Google Scholar CrossRef Search ADS PubMed  38. Cappola AR, Xue QL, Ferrucci L, Guralnik JM, Volpato S, Fried LP. Insulin-like growth factor I and interleukin-6 contribute synergistically to disability and mortality in older women. J Clin Endocrinol Metab . 2003; 88: 2019– 2025. doi: 10.1210/jc.2002-021694 Google Scholar CrossRef Search ADS PubMed  39. Saydah S, Graubard B, Ballard-Barbash R, Berrigan D. Insulin-like growth factors and subsequent risk of mortality in the United States. Am J Epidemiol . 2007; 166: 518– 526. doi: 10.1093/aje/kwm124 Google Scholar CrossRef Search ADS PubMed  40. Laughlin GA, Barrett-Connor E, Criqui MH, Kritz-Silverstein D. The prospective association of serum insulin-like growth factor I (IGF-I) and IGF-binding protein-1 levels with all cause and cardiovascular disease mortality in older adults: the Rancho Bernardo Study. J Clin Endocrinol Metab . 2004; 89: 114– 120. doi: 10.1210/jc.2003-030967 Google Scholar CrossRef Search ADS PubMed  41. Harrela M, Qiao Q, Koistinen R, et al.   High serum insulin-like growth factor binding protein-1 is associated with increased cardiovascular mortality in elderly men. Horm Metab Res . 2002; 34: 144– 149. doi: 10.1055/s-2002-23198 Google Scholar CrossRef Search ADS PubMed  42. Sanders JL, Guo W, O’Meara ES, et al.   Trajectories of IGF-I predict mortality in older adults: the Cardiovascular Health Study. J Gerontol A Biol Sci Med Sci . 2017. doi:10.1093/gerona/glx143. 43. Cappola AR, O’Meara ES, Guo W, Bartz TM, Fried LP, Newman AB. Trajectories of dehydroepiandrosterone sulfate predict mortality in older adults: the Cardiovascular Health Study. J Gerontol A Biol Sci Med Sci . 2009; 64: 1268– 1274. doi: 10.1093/gerona/glp129 Google Scholar CrossRef Search ADS PubMed  44. Kizer JR, Benkeser D, Arnold AM, et al.   Associations of total and high-molecular-weight adiponectin with all-cause and cardiovascular mortality in older persons: the Cardiovascular Health Study. Circulation . 2012; 126: 2951– 2961. doi: 10.1161/CIRCULATIONAHA.112.135202 Google Scholar CrossRef Search ADS PubMed  © The Author(s) 2018. Published by Oxford University Press on behalf of The Gerontological Society of America. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices)

Journal

The Journals of Gerontology Series A: Biomedical Sciences and Medical SciencesOxford University Press

Published: Apr 12, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off