Coronary artery calcium (CAC) at Offspring eighth examination cycle.

A subsample of FHS participants underwent a chest CT at the 8^th examination cycle using a multi-slice multidetector CT scanner (LightSpeed Ultra; General Electric Milwaukee, WI) for the assessment of coronary artery calcium (CAC). Participants were scanned twice by trained technicians following an established protocol, and then examined by experienced readers who identified calcification along the course of the coronary arteries, as reported previously [16]. CAC scores were calculated from the two scans based on a modified Agatston score [17].

Carotid Intima Media Thickness (IMT) at Offspring eighth examination cycle.

FHS study participants underwent carotid ultrasonography of both right and left carotid arteries at the 8^th examination cycle using a Toshiba SSH-140A imaging machine as previously described [18]. Briefly, two measurements of carotid arteries were performed using longitudinal views of both the common carotid artery (CCA) and internal carotid artery (ICA) by a trained and certified sonographer, in a standardized protocol, and then reviewed by a radiologist, both of them blinded to participants’ clinical information. To assess the intima-media thickness (IMT), the sonographer manually traced intima-media interface lines to quantify the degree of thickening of the carotid artery walls. The mean IMT of the CCA was measured over a segment of 1 cm long, located approximately at 5 mm proximal to the common carotid bulb, so CCA-IMT was defined as the mean of the maximal IMT measurement for the right and left CCA. The maximum IMT of the ICA was defined as the maximum wall thickness in either the right or left ICA extending from the bulb to 10 mm above the carotid sinus, so ICA-IMT was defined as the mean of the maximal IMT measurements for the carotid artery bulb and the ICA on the right and left sides [19]. For this investigation, we defined overall IMT of the CCA and ICA as the mean of the mean IMT values of right and left sides after standardization.

Association of biomarkers with components of subclinical atherosclerosis.

We natural logarithmically-transformed values of CAC to normalize their skewed distributions; ln(CAC+1) was used as the dependent variable in analyses. We standardized the average of the values of the CCA and ICA IMT, and obtained an average IMT for both carotid arteries.

We used multivariable linear regression models to evaluate the association of the biomarkers of aging (independent variables, separate model for each) with CAC and IMT (dependent variables, separate models for each). We initially adjusted all models for age and sex, and then further adjusted for BMI, SBP, hypertension medication use, diabetes, smoking status, and total cholesterol/HDL ratio. We also evaluated potential effect modification of the relation between biomarkers of aging and components of subclinical atherosclerosis by sex by including corresponding interaction terms in the models. A P-value <0.05 for the interaction was considered statistically significant.

Association of biomarkers with all-cause mortality and CVD.

Follow-up time started after the 7^th examination cycle, which served as the baseline for this analysis. After confirming that the assumption of proportionality of hazards was met, we used multivariable Cox regression to relate biomarkers of aging (independent variables, separate model for each) to time to death (dependent variable) adjusting for age, sex, BMI, SBP, hypertension medication use, diabetes, current smoking status, total cholesterol/HDL ratio, and eGFR. To evaluate the conjoint association of the biomarkers of aging with time to death, we included IGF-1, ADMA, and F2-Isoprostanes together in a stepwise regression model using an entry and retention criterion of 0.1 for statistical significance level and forced age, sex, smoking status, BMI, SBP, hypertension medication, diabetes, total cholesterol/HDL ratio, and eGFR into the model. LTL was excluded from the stepwise selection model to retain a larger sample size. We allowed for different baseline hazards by stratifying models according to prevalent CVD status in all Cox regression models.

Creation of a biomarker score.

We created a biomarker score as follows: first, we used a stepwise selection process in the Cox regression model including three biomarkers (IGF-1, ADMA, and F2-Isoprostanes; LTL was not included to retain a larger sample size). Then, we multiplied the beta estimate for each biomarker obtained by the stepwise model with the individual’s biomarker value/concentration (e.g. β_ADMA*ADMA). This created three new variables (products) for each participant which we summed to get the biomarker score, i.e. score = β_ADMA*ADMA+β_IGF−1*IGF1+β_F2*F2. The biomarker score was categorized as tertiles in ascending order (tertile 1 with the lowest values vs. tertile 3 with the highest values of the biomarker score). We created a Kaplan-Meier survival curve to graphically present the survival time according to tertiles of the biomarker score. In secondary analysis we created a biomarker score including all four biomarkers (IGF-1, ADMA, F2-Isoprostanes, and LTL) using a subsample including data from participants with available data on all 4 biomarkers.

Discussion.

Principal findings. We observed that higher ADMA levels were associated with higher CAC scores, whereas shorter LTL and lower IGF-1 values were associated with higher IMT values cross-sectionally, adjusting for age, sex, and other cardiovascular risk factors. Shorter LTL and lower IGF-1 concentrations, and higher levels of ADMA and F2-Isoprostanes were associated with higher risk of all-cause mortality prospectively when modeled individually. In addition, IGF-1, ADMA and F2-Isoprostanes levels were jointly associated with all-cause mortality when modeled together. In secondary analyses, ADMA and F2-Isoprostanes (joint model) were positively and LTL was inversely associated with risk of incident CVD.

LTL and CVD outcomes.

Epidemiological studies in humans have reported an association between shorter LTL and higher risk of CVD [21] and all-cause mortality [22], but effect sizes have been small and directionality of associations has been inconsistent [23]. It is not clear whether short LTL is a biomarker, a risk factor, or a consequence of aging, and age-related diseases [24]. It is also possible that LTL shortening is along the causal pathway between oxidative stress and CVD [23].

LTL and subclinical CVD.

Our findings regarding the association of LTL with IMT are in accordance with other studies. Shorter LTL has been associated with presence of carotid plaque [25], incident CCA-IMT [26], and presence of higher CCA-IMT and ICA-IMT [27]. In the latter study, using data from FHS Offspring participants, the authors observed an inverse association between LTL and ICA-IMT in the overall sample, whereas an association of LTL and CCA-IMT was statistically significant only among obese men [27]. Our investigation differed from this prior study in two respects. One, our participants were at least a decade older than participants from the other FHS investigation. Second, we defined our outcome, carotid IMT, as the average of the means of CCA-IMT and ICA-IMT after standardization. The inverse association of LTL and mean carotid -IMT in our sample remained statistically significant after adjustment for age, sex, BMI and other CVD risk factors. Our finding that LTL is associated with IMT is consistent with most published reports [26,28]. However, consistent with a report by Fernandez-Alvira et al. from the Progression of Early Subclinical Atherosclerosis (PESA) study [29], we did not observe an association between LTL and CAC score. However, in a cross-sectional study of asymptomatic middle-aged adults, shorter LTL was significantly associated with higher CAC score [30]. The latter study differs from ours in that the sample was smaller in size and participants were younger and healthier, but the effect size of the association between LTL and CAC was small [30]. We included a much larger sample of participants, but we did not observe an association between LTL and CAC after adjustment for other variables such as age, smoking, and obesity. Longitudinal studies are warranted to define better whether there is a significant association between shortening LTL and progression of CAC.

LTL and all-cause mortality.

Previous studies have reported an inverse association between LTL and all-cause mortality consistent with our findings [31–35]. A recent meta-analysis [22] of 25 investigations based on prospective data collection reported an association between shorter LTL and higher risk of all-cause mortality, with a moderate effect size of the association that was similar in the two youngest groups (<75 years and 75–80 years) but weaker in the oldest group (over 80 years). Further subgroup analyses demonstrated that LTL measurement technique, sex, age, ethnicity, and the number of covariates included contributed to the between-study heterogeneity [22]. Similar to previous studies, in our investigation the effect size for the association of LTL with risk of all-cause mortality was modest. It is not yet clear whether shortening LTL triggers a process that will lead to earlier mortality or whether there are other biological processes that cause shortening of LTL [34]. To add to this complex scenario, other factors could interfere in this association like telomerase activity and amount of oxidative stress, which we did not account for in our statistical models.

IGF-1 and subclinical CVD.

In concordance with our findings, two other studies [45,46] reported an inverse association of IGF-1 concentrations and carotid IMT. Other epidemiological studies with small-to-moderate sample sizes did not find association between lower IGF-1 concentrations and IMT [47], or observed higher levels of IGF-1 related to higher CCA IMT [48]. However, a study investigating this association, reported a U-shaped association in a small sample of older people without known CVD [49]. In the present investigation, we did not observe any association between IGF-1 and CAC. Studies on the association of low concentrations of IGF-1 with subclinical cardiovascular disease have yielded contradictory results. It is unclear why lower IGF-1 is associated with higher carotid IMT but no with higher coronary calcium in the present investigation. It has been proposed that subclinical disease in the coronary and carotid arteries may reflect different atherosclerotic processes/measures [49].

IGF-1 and all-cause mortality.

The mechanisms by which a low IGF-1 serum level may be associated with reduced mortality are not clear. It has been proposed that IGF-1 may protect against atherosclerosis (as opposed to the opposite pro-atherosclerotic effects of oxidative stress and inflammation) and increase the production of NO and NOS in endothelial cells. IGF-1 also is involved in the vascular aging process by preventing oxidative smooth muscle cell apoptosis and reducing proinflammatory cytokine production in atherosclerotic plaques [50]. The relation between circulating IGF-1 concentrations and mortality is complex even in epidemiological studies. Low circulating IGF-1 concentrations have been associated with increased risk of cardiovascular mortality in some studies [51,52] but not in others [43]. Circulating concentrations of IGF-1 have also been associated with all-cause mortality, in a U shape manner [41]. In a national sample of more than 20,000 participants that was followed between 4 and 12 years, IGF binding protein (BP)-3 concentrations but not IGF-1 levels was associated with cardiovascular or all-cause mortality [43]. This sample differed from ours in that it is a nationally representative sample and included people 20 years or older. The effect of IGF-1 on aging (and related outcomes) may have been attenuated by including younger people in the latter study. Contrary to the aforementioned national sample, we observed an inverse association between IGF-1 and all-cause mortality in our sample. Our sample was followed for a longer period than the national sample, resulting in a larger number of events and greater statistical power to elucidate associations.

ADMA and subclinical CVD.

In the current investigation, participants with higher levels of ADMA had higher CAC scores than those with lower levels in accordance with at least two previous studies. [57,58] although these latter studies differed from our investigation in that their sample size was smaller, and included mostly black participants [57] or Japanese patients with CKD [58]. Higher levels of ADMA have been associated with the presence of subclinical atherosclerosis as defined by CAC and carotid IMT; however, we did not observe a significant association between blood ADMA levels and IMT. Contrary to our investigation, two other community-based studies with moderate sample sizes (575 and 922 participants, respectively) observed associations between circulating ADMA concentrations and carotid IMT in a non-white sample [59,60]. In a recent publication using data from FHS Offspring participants, higher levels of blood ADMA were associated with presence of greater ICA/bulb-IMT but not with CCA-IMT [61]. We assessed carotid-IMT as an average of both ICA-IMT and CCA-IMT, so it is possible that the effect of ICA-IMT may have been attenuated by averaging ICA and CCA.

ADMA and all-cause mortality.

The association between ADMA and all-cause mortality has been investigated in population-based studies [62] and in clinical studies with CAD [63] or terminal renal disease patients [64], but findings are still inconsistent. In a previous analysis using data from the Framingham Heart Study [62], we observed a significant association of ADMA with all-cause mortality only among non-diabetic participants. In the report by Leong et al., a community-based study of an all-women cohort [64], the risk for all-cause mortality was slightly higher with higher ADMA levels, but did not reach statistical significance after 24 years of follow-up. In the current Framingham sample that includes middle-aged to older men and women with a large follow-up period and more incident fatal events than previous studies, we observed a positive association of ADMA and all-cause mortality.

F2-Isoprostanes and subclinical CVD.

Unlike our study, at least two studies [69,70] observed positive associations of F2-Isoprostane levels with coronary calcification. In one, a community-based study of young black and white subjects, F2-Isoprostanes was assessed in blood, rather than urine [69]. In the other study [70], the sample consisted of a small group of Japanese patients with type 2 diabetes. In a case-control study with a small sample size (n = 30 patients), those with greater carotid or iliofemoral IMT (>0.5mm) had increased urinary levels of F2-Isoprostanes compared to those with lower IMT [71].

Previous evidence has linked F2-Isoprostanes, a marker of oxidative stress, to CVD especially in individuals with clinical CVD. Ours sample was comprised of individuals without overt clinical CVD and with a low prevalence of subclinical CVD; this may explain the lack of statistically significant associations between F2-isoprostane and IMT or CAC in the present investigation.

F2-Isoprostanes and all-cause mortality.

In the present study, we observed a significant association between urinary F2-Isoprostanes and all-cause mortality. Our finding is in accordance with a large German cohort study of adults (n~8,000) with a follow up of 14 years, in which F2-Isoprostane urinary levels were associated with CVD mortality [68].