Introduction

The calcium score introduced by Agatston et al. in 1990 [1], remains to date the most frequently used tool to quantify coronary artery calcium (CAC) seen on a computed tomography (CT) scan of the heart. Many relevant publications have demonstrated that the risk of cardiovascular events correlates well with the presence and severity of CAC [2–4]. However, ethnicity, age and sex influence the prevalence and extent of CAC, as noted in the Multi-Ethnic Study of Atherosclerosis (MESA) [5–7]. The main implication of this observation is that a numerically similar CAC score may carry a different prognostic weight in patients of different ethnicity and gender. Therefore a uniform interpretation of the significance of pre-specified CAC thresholds without considering the demographics of the individual, is likely fraught with a risk estimation error. A second important concept that emerged from the recent literature surrounding CAC is that atherosclerosis imaging may help predict the biological age of an individual that may be very different from his chronological age [8]. In fact, since CAC accumulation is strictly linked with ageing as well as atherosclerosis accumulation, a middle age man with extensive CAC resembles a more senior subject as far as his “coronary age”. Shaw et al. [8] addressed these concepts in a recent publication where the chronological age of an individual was increased or reduced based on the extent of CAC detected on a cardiac CT scan. The authors supported their suggestion with mortality data collected from a cohort of 10,377 patients referred by primary care physicians for CAC screening and followed for an average of 5 years. However, over 95% of the subjects were Caucasians and demonstrated a very uniform background (mostly from middle Tennessee). In contrast, in this analysis we utilized data from the MESA study to generate age-specific curves of CAC score appropriate for sex and ethnicity. We propose a method to calculate an individual’s “coronary age” based on the CAC score measured in his/her coronary artery tree compared to the median CAC measured in a similar population. As the subjects included in the MESA study were coronary artery disease naïve and include 4 main Ethnic groups, these data may be valuable to the physician trying to estimate risk in patients from various ethnic groups.

Methods and results

Patients and imaging methods

MESA is multicenter study designed to assess the prevalence and extent of, as well as the prognosis related to subclinical atherosclerosis in a multi racial cohort consisting of four ethnicities: White, Black, Hispanic and Chinese [5]. The race of the patients enrolled was obtained by means of questionnaires and the participants could select only one answer. This may potentially have introduced some classification error. In total, 6,814 people with no clinically known cardiovascular disease at baseline were enrolled. Specifically, the investigators enrolled 2419 Whites (48% men), 1898 Blacks (44.5% men), 803 Chinese (48.6% men) and 1494 Hispanics (48% men) [6]. Patients were recruited at 6 sites in the United States: Columbia University, New York, NY, Johns Hopkins University, Baltimore, MD, Northwestern University, Chicago, IL, University of California Los Angeles, Los Angeles, Ca, University of Minnesota, Twin Cities, MN and Wake Forest University, Winston-Salem, NC. The coronary artery calcium score was calculated according to the Agatston method [1]. CT images were obtained either with a C-150 electron beam computed tomography scanner (GE Imatron, San Francisco, Ca) at the Chicago, New York and Los Angeles locations or a 4-slice multidetector computer tomography scanner (Lightspeed GE, Milwaukee, WI and Volume Zoom, Siemens, Erlangen, Germany) at the other sites. All scans were acquired with prospective triggering and were performed with a calibration phantom positioned under the patient lying on the radiological cradle. The images were read centrally at the Harbor-UCLA laboratory (Torrence, Ca). The reported intrascan variability was similar for all types of CT scanners used and was ∼18% [9]. The volume and mass calcium scores were not collected and therefore could not be used in these analyses.

A detailed description of the aims of the study can be obtained from other published reports [5–7] and online at http://www.mesa-nhlbi.org (last accessed April 20, 2007).

Statistical analyses

The curves presented in this study were calculated utilizing the online MESA calcium calculator available at http://www.mesa-nhlbi.org/calcium/input.aspx. The calculator provides a calcium score percentile (age, sex and ethnicity specific) for any measured CAC score in patients of Chinese, Hispanic, Black or White ethnicity. Using the calculator we obtained the 50th percentile for each year of age in the range of 45–84, for each ethnic group and both sexes, for a total of 8 groups. These 8 datasets were entered into an excel spreadsheet (Microsoft Corporation 2003, Redmond, WA, USA) and graphs were created for each ethnicity and sex plotting age on the y-axis (dependent variable), and CAC score on the x-axis (independent variable). A polynomial curve fit was performed to obtain the equations for the calcium score curves. A univariate non-linear regression analysis was performed for each pair of CAC score and age corresponding to the 8 subgroups using SPSS 15.0 for windows (SPSS Inc, Chicago, IL).

Our assumption was that an individual demonstrating a CAC score greater than his/her chronological age should be considered as old as the individual in the next age group whose median CAC score corresponds to that of the individual under consideration.

The analyses based on MESA data showed that the calculated 50th CAC score percentile was highest for white patients of both sexes with white males having the highest CAC scores overall. Furthermore, CAC was found at a younger age in the white ethnic group. No significant differences were noted for the 50th percentile of CAC score between Blacks and Hispanics in the respective sex categories. However, the CAC scores were significantly (p < 0.01) higher in males than females and CAC appeared at a younger age in males for these groups. Chinese subjects had the lowest scores in both sexes and Chinese females had the lowest CAC scores overall.

The polynomial curves provided excellent fit for all datasets resulting in R2 value of 0.999. The fitted curves are shown in Figs. 1 and 2. The resulting polynomial equations are shown in Table 1. Utilizing these equations a hypothetical 45-year-old white man with a CAC score of 79 resembles a 65-year-old individual of the same race and sex as far as his coronary age is concerned. For a 45-year-old black man with a CAC score of 79 the calculated coronary age is 75 due to a substantial difference in CAC accumulation between races.

Fig. 1
figure 1

Median calcium score curves in men: Graphs plotting the 50th percentile calcium scores in men for four ethnicities (White, Black, Chinese and Hispanic). The coronary calcium scores are plotted on the x-axis with age on the y-axis. The graphs also show the trend lines (solid lines) and the predictive mathematical equations. These predictive trendiness have excellent fit with the observed data (R2=0.995–0.999)

Full size image
Fig. 2
figure 2

Median calcium score curves in women: Graphs plotting the 50th percentile calcium scores in women for the four ethnicities (White, Black, Chinese and Hispanic). The coronary calcium scores are plotted on the x-axis with age on the y-axis. The graphs also show the trend lines (solid lines) and the predictive mathematical equations. These predictive trendiness have excellent fit with the observed data (R2=0.998-0.999)

Full size image
Table 1 Predictive equations for vascular age based on calcium score and for the 50th percentile of calcium score based on a subject’s age
Full size table

Table 1 also shows the polynomial equations to calculate the median calcium score from the chronological age of an individual. For instance, for a 62 year-old man the calculated median calcium score is 44.

Discussion

In this manuscript we elaborated on the concept of biological age introduced by Shaw et al. [8]. We propose to assess an individual’s coronary age based on his CAC score compared with the median CAC score of individuals of the same sex, age and race. The novelty resides in the utilization of data collected from a large, ethnically diverse and asymptomatic cohort, unlike the monotonic group originally studied by Shaw et al. [8]. The suggestion that coronary age may carry prognostic significance came from the observation that CAC scores percentiles predicted the occurrence of a hard outcome better than absolute scores and traditional risk factors in a cohort of patients referred for CAC screening by primary care physicians [3, 10]. CAC percentiles were age and sex specific scores derived from a homogeneous population in middle Tennessee including over 90% white individuals [3]. A high percentile suggests an accelerated atherosclerosis process, since the individual demonstrates more CAC than expected for his age. Hence this subject is relatively older than his chronological age. In a subsequent study, Shaw et al utilized data on all cause mortality derived from 10,377 patients referred for CAC screening by primary care physicians [7]. In that study, utilizing absolute CAC scores and linear regression models the authors calculated a calcium age for each patient. They then estimated the number of years of life lost with Cox regression models [7]. The conclusion was that a score < 10 may detract as many as 10 years from a 70 year old man (i.e., the man appears to be 10 years younger), and a score > 400 may add as many as 30 years (30 years older) to a 40 year old man. The advantage of that analysis over our study was the availability of actual mortality data. Nonetheless, the data were only applicable to white patients with risk factors for atherosclerosis and could not be extended to asymptomatic, low-risk subjects belonging to other ethnic groups. In the current study, therefore, we offer a method to extend such prognostically important concepts to subjects from the general population with a low referral bias. Furthermore, the calcium age represents a simple integration of subclinical atherosclerosis assessment, ethnicity, age and gender that can be easily grasped by patients as physicians endure to explain the significance of the results of an imaging test. If these concepts were validated in prospective studies, we envision inserting a coronary age in each patient’s report sent to the referring physician or to the patient himself, as this would facilitate the interpretation of the results.

In summary, the predictive equations presented in this manuscript can be used to estimate the coronary age of men and women of diverse ethnicities from the general population based on their Agatston score. Unfortunately, there are currently no multi-ethnic databases that contain volume and mass scores that could be utilized to derive similar equations. Coronary age may provide information regarding the true biological age of the coronary arteries of the individual under study and can indicate the severity of damage that has occurred during the process of aging. This method should allow patients to better understand the meaning of their test results and may ensure a better compliance with lifestyle modifications and preventive therapeutic regimens. The results of the long-term follow-up from the MESA study are awaited soon and will help validate the utility of coronary age as a means to represent the risk inherent with the burden of atherosclerosis.