Author + information
- Received June 8, 2018
- Revision received October 9, 2018
- Accepted October 12, 2018
- Published online January 16, 2019.
- Nikolas Lessmann, MSca,∗ (, )
- Pim A. de Jong, MD, PhDb,c,
- Csilla Celeng, MD, PhDb,
- Richard A.P. Takx, MD, MSc, PhDb,d,
- Max A. Viergever, PhDa,c,
- Bram van Ginneken, PhDe and
- Ivana Išgum, PhDa
- aImage Sciences Institute, University Medical Center Utrecht, Utrecht, the Netherlands
- bDepartment of Radiology, University Medical Center Utrecht, Utrecht, the Netherlands
- cUtrecht University, Utrecht, the Netherlands
- dDepartment of Radiology, St. Antonius Hospital, Nieuwegein, the Netherlands
- eDiagnostic Image Analysis Group, Radboud University Medical Center, Nijmegen, the Netherlands
- ↵∗Address for correspondence:
Dr. Nikolas Lessmann, University Medical Center Utrecht, Postbus 85500, Room Q.02.4.45, 3508 GA Utrecht, the Netherlands.
Objectives The aim of this study was to investigate sex differences in the prevalence, extent, and association of coronary artery calcium (CAC) and thoracic aorta calcium (TAC) scores with cardiovascular mortality in a population eligible for lung screening.
Background CAC and TAC scores derived from chest computed tomography (CT) might be useful biomarkers for individualized cardiovascular disease prevention and could be especially relevant in high-risk populations such as heavy smokers. Therefore, it is important to know the prevalence of arterial calcifications in male and female heavy smokers, and if there are differences in the predictive value calcifications carry.
Methods We performed a nested case–control study with 5,718 participants of the CT arm of the NLST (National Lung Screening Trial). Prevalence and extent of CAC and TAC were resampled to the full cohort to provide unbiased estimates of the typical calcium burden of male and female heavy smokers. Weighted Cox proportional hazards regression was used to assess differences in the association of CAC and TAC scores with all-cause and cardiovascular mortality.
Results CAC was substantially more common and more severe in men (prevalence: 81% vs. 60%; median volume: 104 mm³ vs.12 mm³). Women had CAC comparable to that of men who were 10 years younger. TAC was equally common in men and women, with a tendency to be more pronounced in women (prevalence: 92% vs. 93%; median volume: 388 mm³ vs. 404 mm³). Both types of calcification were associated with increased cardiovascular and all-cause mortality. TAC scores improved the prediction of coronary heart disease mortality over CAC in men, but not in women. In both sexes, TAC, but not CAC, was associated with cardiovascular mortality other than coronary heart disease.
Conclusions CAC develops later in women, whereas TAC develops equally in both sexes. CAC is strongly associated with coronary heart disease, whereas TAC is especially associated with extracardiac vascular mortality in either sex.
- cardiovascular disease
- coronary artery calcium
- lung cancer screening
- sex differences
- thoracic aorta calcium
Although cardiovascular disease is one of the major causes of death for both men and women, clear differences exist between the sexes in incidence rates of cardiovascular events (1,2). A strong predictor of cardiovascular events and mortality is the amount of arterial calcification that can be detected and quantified in cardiac or chest computed tomography (CT) scans (3). CT-based lung cancer screening of heavy smokers provides an opportunity for assessment of cardiovascular risk in addition to lung screening (4–7). However, in the lung cancer screening eligible population, the prevalence and extent of vascular calcification are likely higher than in the general population due to smoking history and advanced age. Reference values for coronary artery calcium (CAC) and thoracic aorta calcium (TAC) scores obtained from ungated lung screening CT and the associated risk for cardiovascular events and mortality are not well-established. Furthermore, although coronary atherosclerosis and coronary heart disease (CHD) are known to affect men more often than women (2), differences in aortic atherosclerosis are less well-established. The sex difference has been reported to be smaller for TAC than for CAC (8), which would be important information for correct interpretation of lung cancer screening–derived calcium scores and sex-specific management of these prevalent findings. Moreover, a limited number of studies suggest that TAC may be more relevant than CAC for prediction of noncardiac events and all-cause mortality (9,10). The purpose of this study was to investigate differences in A, B, and C of CAC and TAC between men and women, where A = prevalence, B = extent and C = association with cardiovascular and all-cause mortality between men and women in a lung cancer screening trial.
We performed a retrospective nested case-control study with data collected from the NLST (National Lung Screening Trial) (11). The design of the NLST, including eligibility criteria, imaging protocols, and participant characteristics was described in detail elsewhere (12). Briefly, the NLST compared screening for lung cancer with chest radiography against screening with low-dose chest CT. Eligibility criteria were a smoking history of at least 30 pack-years, active smoking within the previous 15 years, and an age of 55 to 74 years at enrollment. A total of 26,722 participants were randomized to the CT arm of the trial. These participants underwent 3 rounds of low-dose unenhanced, ungated chest CT imaging at 120 kVp in 1-year intervals in 1 of 33 participating screening centers in the United States. Enrollment began in August 2002, and participants were followed until the end of 2009, with a median follow-up time of 6.5 years.
In this case–control study, we included all participants in the CT arm of the NLST who died and a control group of approximately twice as many randomly chosen participants from the CT arm who were either censored or still alive at the end of the follow-up period. In total, 5,718 participants were included, of whom 3,553 (62%) were men, and of whom 1,724 (30%) died. Details of the selection are shown in Figure 1. In our statistical analysis, cases and control subjects were resampled to the full cohort size to remove the bias of the case–control design and make the results interpretable for the lung cancer screening eligible population (13).
Endpoints for the survival analysis were defined based on the cause of death as reported on the death certificates. These were coded using the International Classification of Diseases-10th edition, and a single underlying cause of death was determined. We performed separate analyses for multiple endpoints. These were all-cause mortality, death from cardiovascular disease (CVD), which was defined as death from any disease of the circulatory system (codes I00 to I99), death from CHD, which was any death from ischemic heart diseases (codes I20 to I25), cardiac arrest (code I46) or heart failure (code I50), and extracardiac vascular deaths, which were deaths from CVD, excluding CHD.
Calcifications of the coronary arteries and the thoracic aorta were scored in the baseline scans using automatic, in-house developed software. This software was previously validated with NLST scans (14). Briefly, the software uses a deep-learning approach based on 2 consecutive convolutional neural networks to detect calcifications and to label them according to the affected vascular bed. In this study, we calculated calcium scores for the total amount of calcification detected in either the coronary arteries (CAC score) or the thoracic aorta as far as visible, including the aortic arch (TAC score). Only voxels with an intensity of at least 130 Hounsfield units were taken into consideration, which is the standard threshold for calcium scoring (15). Because calcium scoring is normally performed in images with a slice thickness of 3 mm, all scans were resampled before the analysis to a slice thickness of 3 mm with a 1.5-mm increment. The detected calcifications were quantified by calculating the total calcium volume in millimeters cubed and a modified Agatston score that accounted for overlapping slices (16).
Differences in the prevalence and extent of calcification of the coronary arteries and the aorta were tested in 5-year age groups (55 to 59, 60 to 64, 65 to 70, 70 to 74 years). Because of the non-normal distribution of calcium scores, scores stratified by age group and sex were summarized with medians and interquartile ranges. We calculated weighted medians and prevalences to account for the bias introduced by the case-control design of our study. Because our study population included all deaths, but only a subset of the remaining participants of the full NLST low-dose CT cohort, control subjects were weighted with the inverse of their sampling probability. Prevalence of CAC and TAC was defined as an Agatston score ≥1 and was compared using chi-square tests with Rao-Scott second-order corrections to account for the sample weights (17). The extent of calcification in terms of calcium volume was compared using Mann-Whitney U-tests (18).
We used weighted Cox proportional-hazards regression (19) to examine the association between CAC and TAC scores and all-cause or cause-specific mortality. In addition to CVD death, we examined the association with cardiac (i.e., CHD) and extracardiac cardiovascular causes of death to investigate the hypothesis that CAC more closely associated with cardiac and TAC with extracardiac causes of death. In the models for time to a specific cause of death, participants who died of another cause were censored at the time of death. In all cases, separate models were fit for male and female participants.
All models were adjusted for cardiovascular risk factors based on questionnaires filled in by the participants at study entry. The variables included in the models were age, smoking status, pack-years, body mass index, race, education, marital status, and self-reported history of heart disease or stroke, diabetes, and hypertension. CAC and TAC scores were examined as stratified variables with Agatston score cutoff points for CAC of 0, 1 to 100, 101 to 1,000, >1,000, and for TAC of 0 to 1,000, 1,001 to 4,000, and >4,000. The cutoff points for CAC were higher than commonly used cutoff points because of the large number of elevated CAC scores in the study population (5,6). Both CAC and TAC were included in all models because we were interested in their complementary, rather than their individual, predictive value.
We tested for differences in baseline characteristics between male and female cases and control subjects using Kruskal-Wallis tests for continuous variables and chi-square tests or Fisher exact test, if frequencies were low, for categorical variables. Participants with missing survival status or unknown cause of death and participants with missing calcium scores because of missing or unreadable images were excluded from the analysis (Figure 1). Missing demographic and disease history data were imputed using regression techniques (20). Values of p < 0.05 were considered statistically significant. All statistical analyses were performed using R 3.4 (R Foundation, Vienna, Austria).
Baseline characteristics of the study population are shown for cases and control subjects, and according to sex in Table 1. In the study population, men were more likely married, better educated, and had, on average, accumulated more pack-years compared with women. Men were also more likely to have a diagnosis of heart disease or diabetes at study entry, whereas previous diagnoses of hypertension or stroke were equally likely in men and women.
Deceased study participants were more likely to be men than women (1,206 deaths vs. 518 deaths; 13.3 deaths/1,000 person-years vs. 8.1 deaths/1,000 person-years at risk in the NLST CT arm cohort; p < 0.001). Death from CVD, and specifically CHD, was also more frequent in men (3.6 CVD deaths vs. 1.9 CVD deaths and 2.4 CHD deaths per 1,000 person-years vs. 1.1 CHD deaths per 1,000 person-years; p < 0.001 in both cases). However, there was no significant difference in death rates between men and women for extracardiac vascular causes of death (1.1 deaths/1,000 person-years vs. 0.9 deaths/1,000 person-years; p = 0.153).
Prevalence and extent of arterial calcification
Arterial calcification was present in a substantial number of participants: 72% had calcified coronary arteries, and 93% had thoracic aorta calcification (after resampling to the NLST CT arm cohort). Both prevalence and extent of both CAC and TAC increased with age. The prevalence of CAC ranged from 62% in participants age 55 to 59 years to 92% in participants age 70 to 74 years. The prevalence of TAC was higher across all age groups, with presence of TAC in 88% of the participants at age 55 to 59 years and in virtually all participants at age 70 to 74 years (99%).
CAC was significantly more often prevalent in men than in women of similar age. The extent of CAC was also greater in men throughout all age groups. Women had, on average, as much CAC as men who were approximately 10 years younger (Figures 2A and 3A⇓⇓). Conversely, TAC was equally likely present in men and women of similar age (Figure 2B). The extent of TAC was comparable for men and women of similar age, except for the oldest participants aged 70 to 74 years, among whom women had more TAC than men (Figure 3B).
Zero CAC scores
TAC was present in almost the entire study population; however, 19% of the men and 40% of the women (after resampling to the NLST CT arm cohort) had no visible CAC despite their advanced age and extensive smoking history. However, subjects with a zero CAC score had lower prevalence of TAC compared with subjects with a non-zero CAC score (77% vs. 96% in men and 88% vs. 97% in women). In both sexes, a zero CAC score was associated with younger age and fewer pack-years, but not with smoking cessation. Marital status played a role only in women, who more often had a zero CAC score when they were married. Overall, participants with a zero CAC score were healthier than those who developed CAC (Table 2). Accordingly, death rates estimated for the NLST CT arm cohort were consistently lower for participants with CAC = 0 compared with those with CAC >0 at baseline (all p < 0.001). There were approximately 7.2 deaths/1,000 person-years (95% confidence interval [CI]: 5.9 to 8.6) versus 15.2 deaths/1,000 person-years (95% CI: 14.1 to 16.2) at risk among men and 4.5 deaths/1,000 person-years (95% CI: 3.6 to 5.3) versus 10.5 deaths/1,000 person-years (95% CI: 9.4 to 11.6) at risk among women. Specifically, death from CVD occurred at rates of 1.3 deaths/1,000 person-years (95% CI: 0.7 to 1.9) versus 4.5 deaths/1,000 person-years (95% CI: 4.0 to 5.1) and 0.8 deaths/1,000 person-years (95% CI: 0.5 to 1.2) versus 2.9 deaths/1,000 person-years (95% CI: 2.3 to 3.4) in men and women, respectively.
Inclusion of CAC and TAC significantly improved the c-statistic for prediction of all outcomes over models without these variables in both sexes (Tables 3 and 4⇓⇓). However, the improvement was not significantly greater in one of the sexes for any outcome (all p > 0.05). Increased levels of both CAC and TAC were associated with increased risk of all-cause and CVD mortality in men and women. CAC was more strongly associated with CVD mortality in women, whereas greater amounts of TAC were more strongly associated with CVD mortality in men than in women. However, these differences did not reach statistical significance (all p > 0.05). The association with CHD mortality was statistically significant for CAC in both sexes and for TAC only in men (p = 0.007), but not in women (p = 0.286). Conversely, increased levels of TAC were associated with increased risk of extracardiac CVD mortality in both sexes, but increased levels of CAC were not.
We investigated sex differences in arterial calcification visible on chest CT and in the associated risks independent of cardiovascular risk factors in a population of heavy smokers eligible for lung cancer screening. Men had calcified coronary arteries more often, which were also more severely calcified. Calcification of the thoracic aorta was highly prevalent in both sexes, with a tendency to be more pronounced in women. Survival analysis revealed that high CAC scores corresponded to higher hazard ratios in women for all outcomes. TAC scores demonstrated incremental predictive value over CAC and other risk factors, especially for prediction of mortality from cardiovascular causes other than CHD and for all-cause mortality.
Comparison with previous studies
It is known that men have a higher CAC burden than women (21–24). The present study confirmed that this also holds for lung cancer screening participants, in which a high calcification burden can be expected in both sexes because eligibility for screening requires the presence of 2 major risk factors for atherosclerosis: advanced age and extensive smoking history. In other publications, lower age–stratified median CAC scores were measured, e.g., in the MESA (Multi-Ethnic Study of Atherosclerosis) trial (23), as well as in a study of a large population of self-referred asymptomatic subjects (21). The incidence of CAC and the increase of CAC scores with age was reported to lag behind in women by 10 to 20 years (21,23,25,26). In this population of heavy smokers, we observed a similar age difference of approximately 10 years between men and women with similar CAC scores.
There is currently no consensus whether a similar sex difference exists for calcification of the aorta. Results of previous studies are inconclusive because some studies reported higher prevalence in women (8,27), some reported higher prevalence in men (28,29), and some reported no significant difference between the sexes (25,30). We observed no significant difference in most age groups for both prevalence and extent of TAC. However, the oldest women were more likely to have larger amounts of TAC than men of similar age (70 to 74 years), even though women were less excessive smokers. Remarkably, TAC was present in almost all individuals of this age.
A large number of studies demonstrated that CAC is a strong independent predictor of cardiovascular events and mortality in both sexes (15). Studies that examined sex differences in the ability of CAC to stratify cardiovascular risk are scarce, but they commonly reported equal power in both sexes (31,32). However, higher event rates were reported for women with high CAC scores compared with men with similar scores (33,34). Our results were in line with these findings because CAC was associated with cardiovascular mortality in both sexes independent of other risk factors and because hazard ratios were consistently higher in women.
In contrast to many other studies (15), a zero CAC score was not associated with better survival than a low CAC score. Motion artifacts might inhibit the distinction of true zero CAC scores from low CAC scores in ungated chest CT. Furthermore, although CVD mortality was significantly lower in subjects who had no CAC at baseline, it was still clearly present, regardless of sex. This suggests that there were both CAC- and non-CAC–related causes of cardiovascular mortality in heavy smokers, which limited the value of a zero CAC score as a predictor of the absence of CVD in this population (35). However, it is remarkable that a sizeable proportion of the lung cancer screening population, who were at least 55 years old and had ≥30 pack-years of cigarette smoking history, had not developed CAC.
The value of TAC in the prevention of cardiovascular events is currently not well defined. Most studies that investigated the predictive value of TAC for cardiovascular events and mortality derived TAC scores from cardiac CT scans. Unlike chest CT, these visualize only part of the thoracic aorta so that lesions near the carotid arteries, which are potentially relevant for cerebrovascular events, might have been missed. A study in chest CT scans from the Dutch-Belgian lung cancer screening trial NELSON (Nederlands-Leuvens Longkanker Screenings Onderzoek) showed a stronger association with extracardiac events for TAC than for CAC (9). Likewise, TAC, but not CAC, was associated with extracardiac CVD mortality in both sexes in our study. For prediction of CHD, the MESA trial showed incremental value of TAC over CAC, but only in women (36). Conversely, our study showed an association of TAC with CHD mortality in addition to CAC only in men. However, NLST participants were, on average, substantially older and heavier smokers than MESA participants.
Study Strengths and limitations
The present study was based on a subcohort of the NLST, which provided standardized imaging data for a population at increased cardiovascular risk that consisted of a large number of male and female subjects. A particular strength of this study was the substantial number of deaths that occurred in this population, which facilitated robust survival analysis even after stratification by sex. Another strength was the use of chest CT scans for evaluation of calcifications of the thoracic aorta, which ensured that calcifications of the aortic arch were included. This was often not the case in studies based on cardiac CT.
This study had several limitations. First, the findings of this study are limited to heavy smokers. In the general population, prevalence and severity of CAC and TAC are likely to be lower. Second, we derived calcium scores from ungated CT, which might have introduced an overestimation of the calcium burden compared with gated scans (37). Automatic calcium scoring might have also resulted in more false positive than false negative detections, which might have contributed to an overestimation, especially of the prevalence of arterial calcification (14). Third, only fatal events were recorded in the NLST, but sex differences in prediction of nonfatal cardiovascular events would also be highly relevant to better inform treatment decisions. Fourth, because the NLST primarily targeted lung cancer, lipid measurements were not performed, and intake of lipid-lowering medication or other preventive measures were not recorded. Fifth, the findings of this study are of a retrospective nature, and therefore, do not provide evidence for benefits from sex-specific preventive efforts in the lung cancer screening population.
Neither CAC nor TAC are currently part of guidelines for lung cancer screening with low-dose chest CT. However, reporting of CAC on ungated chest CT scans is increasingly recommended (38). The relevance of these findings is debated because virtually all heavy smokers eligible for lung cancer screening are at least in the intermediate-risk group for which preventive efforts are recommended, regardless of their calcium score (39). However, the findings of this study indicate that both CAC and TAC scores can support the identification of subjects who will experience an adverse cardiovascular event. Calcium scores can be useful biomarkers for individualized preventive efforts in men and women, especially if health care providers also take sex differences into account. In the high-risk population of heavy smokers, calcium scores can potentially help to identify individuals at highest risk for whom more aggressive and expensive preventive treatment beyond lipid management (e.g., anti-inflammatory drugs ) might be mandated. In addition, calcium scores may also play a role in lowering the burden of medication for low-risk heavy smokers who would otherwise be eligible for statins. Health care providers involved in lung cancer screening need to be aware that high CAC scores occur less frequently in women than in men, but also need to be aware that this difference does not exist for calcification of the aorta.
COMPETENCY IN MEDICAL KNOWLEDGE: The amount of coronary artery and thoracic aorta calcium visible on lung screening CT scans provides incremental prognostic value over demographic cardiovascular risk factors. CAC is more common in men, but is also present in a substantial number of female heavy smokers. Calcification of the thoracic aorta can be expected in nearly the entire lung cancer screening eligible population, with no significant differences between men and women, and provides a more powerful predictor of extracardiac CVD mortality than CAC.
TRANSLATIONAL OUTLOOK: The use of calcium scores for cardiovascular risk assessment, especially in addition to lung cancer screening, could greatly benefit from standardized reporting. Prospective studies are needed to develop evidence-based guidelines for management of prevalent cardiovascular findings in the high-risk population of heavy smokers.
The authors are grateful to the United States National Cancer Institute (NCI) for providing access to NCI’s data collected by the National Lung Screening Trial. The statements contained herein are solely ours and do not represent or imply concurrence or endorsement by NCI.
Drs. Išgum and Viergever were supported by an institutional research grant from PIE Medical Imaging and a research grant from the Netherlands Organization for Health Research and Development (ZonMw) in the framework of the research program IMDI (Innovative Medical Devices; 104003009), awarded with participation of Pie Medical Imaging. Drs. Išgum, van Ginneken, and Viergever were supported by a research grant from the Dutch Technology Foundation (STW) within the Deep Learning for Medical Image Analysis (DLMedIA) program, awarded with participation of PIE Medical Imaging, Philips Healthcare, Thirona, ScreenPoint Medical, and Delft Imaging Systems (P15-26). Drs. Išgum, de Jong, and Viergever were supported by a research grant from the Dutch Technology Foundation (STW) within the Population Imaging Genetics (ImaGene) program, awarded with participation of PIE Medical Imaging and 3mensio Medical Imaging (12726).
Dr. van Ginneken is co-founder and stockholder of Thirona; and has received royalties and research funding from Thirona, MeVis Medical Solutions, and Delft Imaging Systems. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Acronyms and Abbreviations
- coronary artery calcium
- coronary heart disease
- confidence interval
- computed tomography
- cardiovascular disease
- thoracic aorta calcium
- Received June 8, 2018.
- Revision received October 9, 2018.
- Accepted October 12, 2018.
- 2019 American College of Cardiology Foundation
- Mets O.M.,
- Vliegenthart R.,
- Gondrie M.J.,
- et al.
- Nasir K.,
- Roguin A.,
- Sarwar A.,
- Rumberger J.A.,
- Blumenthal R.S.
- Jacobs P.C.,
- Prokop M.,
- van der Graaf Y.,
- et al.
- Lessmann N.,
- van Ginneken B.,
- Zreik M.,
- et al.
- Hecht H.S.
- Therneau T.M.,
- Grambsch P.M.
- McClelland R.L.,
- Chung H.,
- Detrano R.,
- Post W.,
- Kronmal R.A.
- Jairam P.M.,
- de Jong P.A.,
- Mali W.P.,
- et al.
- Ohmoto-Sekine Y.,
- Yanagibori R.,
- Amakawa K.,
- et al.
- Kälsch H.,
- Lehmann N.,
- Möhlenkamp S.,
- et al.
- Bellasi A.,
- Lacey C.,
- Taylor A.J.,
- et al.
- Hecht H.S.,
- Cronin P.,
- Blaha M.J.,
- et al.
- Bernheim A.,
- Auffermann W.F.,
- Stillman A.E.