Telomere length was measured from DNA extracted from leukocytes in peripheral blood (Berglund et al., 2016). A quantitative polymerase chain reaction-based technique was carried out to compare the telomere sequence copy number in each participant’s sample (T) to a single-copy reference gene from β-hemoglobin (S). The resulting relative length was represented as T/S ratio. Relative telomere length was further adjusted for batch effect, and 10 outliers (beyond Mean ±4*SD) were omitted in the present analyses.

Genome-wide methylation levels were measured from leukocytes using Illumina’s Infinium HumanMethylation 450K BeadChip according to the manufacturer’s protocol and quantified by beta-values (Wang et al., 2018b). DNAmAges of Horvath and Hannum versions incorporate methylation levels of 353 and 71 age-related CpGs trained from multiple sample types and blood sample accordingly through a penalized regression model (Hannum et al., 2013; Horvath, 2013). In contrast to DNAmAges trained via regressing on age, the third version of methylation age estimator, PhenoAge, were trained on a composite clinical measure of phenotypic age, and eventually included 513 CpG sites (Levine et al., 2018). Further, GrimAge adopted a two-step development method, in which methylation data were used to predict a set of biomarkers (plasma proteins and smoking pack-year), and the methylation-predicted biomarkers were then used to predict mortality risk. As a result, a total number of 1030 CpG sites were taken into account (Lu et al., 2019). All DNAmAges were combined using penalized regression models and generated from the online DNA Methylation Age Calculator (Horvath, 2019).

Physiological age considered a set of physiological biomarkers assessed from the immediate blood test, blood test in lab, urine strip test, and physical examination data that were available in all waves of IPT. First, we included one measurement for each individual to form a sub-sample in which one measurement was randomly selected when repeated measurements for a single individual were available. Pearson correlations were examined using measurements of age and candidate biomarkers in the sub-sample. As a result, nine and five eligible age-associated biomarkers (Pearson correlation >0.1) were included in the development of physiological age for men and women separately. Supplementary file 1B illustrates the biomarker-age correlations in detail. We then performed principal component analysis to created principal components (PCs) and applied a method proposed by Klemera and Doubal (2006) to combine CA and PCs into a single physiological age value in men and women separately using sub-sample. Second, we calculated PCs and physiological age for all available repeated measurements in men and women separately using loadings of biomarkers and weights of CA and PCs which were estimated from the sub-sample analysis.

Four cognitive domains were assessed through a battery of in-person cognition testing, including crystallized, fluid, memory, and perceptual speed abilities (Reynolds et al., 2005). Scores on all cognitive measures were recorded to percentage correct of the total possible points for each respective test. A general cognitive ability score was derived through the principal component analysis (PCA) of the tests. Component scoring coefficients from the first component extracted at IPT1, excluding demented individuals, were used to construct a cognitive function measure at the first and subsequent IPTs using test scores standardized to the mean and SD of each test at IPT1. T-score scaling (M = 50, SD = 10) was then applied to the components.

Four types of specific functional measurements were taken into consideration in the development of FAI (Finkel et al., 2019). Vision and hearing were self-reported on a scale of 1 to 5 and combined to create a measure of self-reported sensory ability. Muscle strength, walking speed time, and lung function were tested and recorded by trained nurses. The four indicators were standardized separately on the basis of the values from IPT two and then summed to create a composite score.

FI was introduced to conceptualize the vulnerability of a given person to a range of age-related adverse outcomes. FI in SATSA was constructed from 42 self-reported health deficits, such as symptoms, diseases, disability, mood, and activities in daily living. FI was calculated as the ratio of the number of deficits presented in a given person to the total number of deficits considered in the study (n = 42 in SATSA). Details of FI items are described in Supplementary file 1C and elsewhere (Jiang et al., 2017).

We constructed BA residuals by regressing out the CA-related part from respective BA. As BAs were assessed in a longitudinal manner and BA levels within twin pairs were assumed to be related due to shared familial factors, we adopted mixed models with fixed effects for sex and CA, the latter as a natural spline term with three degrees of freedom, to allow for non-linear relationships, and random intercepts at the twin-pair and subject level:

with β and µ denoting fixed and random effects, i, j, k being indicators for twin pair, individual, and measurement, respectively, and ns() representing a natural spline term with parameters as specified by the degrees of freedom. The resulting predicted residuals BAResidualijk have thereby been adjusted for CA as well as systematic constant differences between twin pairs and individuals.

Note: The content above has been extracted from a research article, so it may not display correctly. Click HERE to view the original source.

Please log in to submit your questions online.
Your question will be posted on the Bio-101 website. We will send your questions to the authors of this protocol and Bio-protocol community members who are experienced with this method. you will be informed using the email address associated with your Bio-protocol account.

We use cookies on this site to enhance your user experience. By using our website, you are agreeing to allow the storage of cookies on your computer.