Vascular calcifications as prognostic markers of CVD events in chest CT

Transcription

1 Vascular calcifications as prognostic markers of CVD events in chest CT Peter Jacobs

2 Vascular calcifications as prognostic markers of CVD events in chest CT Thesis, Utrecht University with a summary in Dutch ISBN Author P.C.A. Jacobs Cover design Rosa Kuiper Lay-out Roy Sanders Print Gildeprint Drukkerijen B.V., Enschede, The Netherlands

3 Vascular calcifications as prognostic markers of CVD events in chest CT Vaatwandverkalking in CT thorax als voorspeller van hart- en vaatziekte (met een samenvatting in het Nederlands) Proefschrift ter verkrijging van de graad van doctor aan de Universiteit Utrecht op gezag van de rector magnificus, prof.dr. J.C. Stoof, ingevolge het besluit van het college voor promoties in het openbaar te verdedigen op dinsdag 2 februari 2010 des middags te 4.15 uur door Peter Cornelis Arjan Jacobs geboren op 4 juli 1978 te Rotterdam

4 Promotoren: Prof.dr. Y. van der Graaf Prof.dr. W. P. Th. M. Mali Financial support by the Radiologen Stichting Utrecht for the publication of this thesis was gratefully acknowledged. Financial support was also provided by Apotheek Groenendaal Heemstede (L. Maussen), AstraZeneca BV, Schering-Plough BV and Bristol-Myers Squibb BV.

5

6 Contents Chapter 1 General Introduction 9 Part 1 Subclinical coronary and aortic calcifications in routine care, diagnostic chest CT as predictor of CVD events Chapter 2 Prevalence of incidental findings in CT screening of the chest: a systematic review. JCAT. 2008;32: Chapter 3 Semi-quantitative assessment of cardiovascular disease markers in multislice CT of the chest: inter- and intra-observer agreement. JCAT. Accepted. 39 Chapter 4 Use of unrequested information from routine care, diagnostic chest CT: the case of coronary and aortic calcifications. Submitted 55 Part 2 Part 2: Calcium scoring as part of non-gated, low-dose CT screening for lung cancer in asymptomatic heavy smokers Chapter 5 Coronary artery calcium scoring in low-dose ungated computed tomography screening for lung cancer: interscan agreement. AJR. Accepted. 71 Chapter 6 Coronary artery calcium can predict all-cause mortality and cardiovascular events on low-dose computed tomography screening for lung cancer. Submitted. 87

7 Chapter 7 Comparing coronary artery calcium and thoracic aortic calcium for prediction of all-cause mortality and cardiovascular events on lowdose non-gated computed tomography. Atherosclerosis. Accepted. 105 Chapter 8 Coronary artery calcium screening as part of low-dose computed tomography screening for lung cancer improves cardiovascular risk prediction in men. Submitted. 123 Chapter 9 Smoking cessation and risk of cardiovascular disease among participants of a lung cancer screening trial. Submitted. 141 Chapter 10 General Discussion 155 Chapter 11 Summary 165 Nederlandse samenvatting 173 Dankwoord 181 Curriculum Vitae 187

8

9 CHAPTER 1 General Introduction

10 Chapter 1 Cardiovascular disease (CVD) is very common in the general population affecting more than 50% of adults over 60 years of age in the western world, and despite advances in (preventive) treatment, the number of CVD deaths remains high. Over the past decade 30-35% of all annual deaths in the Netherlands were caused by CVD 1. Moreover, up to 50% of first manifestations of CVD are sudden cardiac death or acute myocardial infarction 2. This emphasizes the importance of early detection of asymptomatic individuals at high risk to develop CVD events, especially since most important risk factors are essentially modifiable. Aggressive controlling of blood pressure and lipids through administering preventive drugs in combination with lifestyle coaching (quitting smoking, exercise, modification of dietary habits) has been shown effective to reduce the risk of future CVD events 3. Among several other strategies, this has prompted computed tomography (CT) screening of asymptomatic subjects for the presence of coronary artery calcium (CAC). These calcified atherosclerotic plaques are in fact an intermediate between traditional cardiovascular risk factors (e.g., hypertension, hypercholesterolaemia, smoking) and incident CVD events. Large follow-up studies have established CAC as a strong and independent risk marker for future CVD events and all-cause death 4-6. Consensus statements have confirmed that the CAC score is helpful in stratifying asymptomatic subjects at intermediate risk of future coronary heart disease (CHD) (10%-20% 10-year risk) into low (<10% 10-year risk; no preventive measures indicated) and high risk (>20% 10-year risk; preventive measures indicated) 7. Thoracic aorta calcium (TAC) is a comparatively new risk marker that can be scored as part of CT screening for CAC. Although TAC is partly correlated with CAC, previous reports have demonstrated important differences in the amount of TAC and CAC within an individual and across specific groups of patients 8,9, suggesting a (partly) different aetiology of atherosclerosis in the coronary and noncoronary arterial beds. If this is true, detection of TAC could be complementary to CAC for predicting noncoronary CVD events (e.g., stroke, aortic aneurysms, peripheral arterial occlusive disease). Electron-beam CT (EBCT) and ECG-gated multidetector-row CT (MDCT) are the gold standard examinations to screen for the presence and extent of CAC. A novel approach is to use low-dose, non-gated CT to identify subjects at intermediate or high risk of future CVD events. This thesis explores two different strategies to use CAC and TAC scored on nongated CT as risk predictors for future CVD events. First, we have focused on the possibility of scoring subclinical CAC and TAC as prognostic markers on routine care, diagnostic chest CT in a clinical care population. Second, we have investigated the feasibility and implications of scoring CAC and TAC as part of a low-dose CT lung cancer screening study. 10

11 General Introduction Part 1: Subclinical coronary and aortic calcifications in routine care, diagnostic chest CT as predictor of CVD events Subclinical or unrequested imaging findings are - with the advances of CT technology over the last decade - an increasingly frequent element of a diagnostic CT examination. Review articles have shown a prevalence of as high as 40% of subclinical, extracolonic findings in CT colonography screening studies 10,11. Comparable reviews for chest CT examinations have not yet been performed. Chapter 2 is a systematic review of the prevalence of subclinical findings from chest CT screening studies. Nevertheless, the clinical importance of these findings remains unclear. Follow-up studies are required to gain a better understanding of the natural course of these findings and to investigate their prognostic relevance for predicting future disease. Also, demonstrating a lack of prognostic value is clinically relevant since this will help the radiologist to determine what subclinical findings should or should not be reported. To this end, the multicenter PROgnostic Value of unrequested Information in Diagnostic Imaging (PROVIDI) study was designed. It consists of clinical care patients in whom a routine care, diagnostic chest CT was performed between January 2002 and December For Part 1 of this thesis we focused on subclinical cardiovascular imaging characteristics. In the setting of screening studies, subclinical CAC has already been established as a strong risk factor for CVD events 2,12,13. Other subclinical cardiovascular imaging characteristics have been suggested as possible predictors of future CVD events: thoracic aorta calcification, aortic and mitral valve calcifications, cardiothoracic ratio, elongation of the thoracic aorta One important issue to address when starting to score these emerging imaging characteristics in a variety of non-gated chest CT protocols is the reproducibility of the measurements. Selection and validation of potentially valuable prognostic markers is in part dependent on the ability to score them in a reproducible way. Chapter 3 presents detailed definitions of how to score cardiovascular imaging characteristics semi-quantitatively. Based on these definitions we evaluated the inter and intra-observer reproducibility of a predefined set of possible prognostic imaging markers for CVD events. In Chapter 4 we have used follow-up data from the PROVIDI cohort to study the prognostic value of CAC and TAC - two of the most prevalent examples of subclinical findings in routine care, diagnostic chest CT for the development of cardiovascular disease. Part 2: Calcium scoring as part of non-gated, low-dose CT screening for lung cancer in asymptomatic heavy smokers People with a history of heavy smoking are not only at risk of developing lung cancer, but also at risk for cardiovascular disease which results from a combination of tobacco-induced 11

12 Chapter 1 systemic inflammatory responses, systemic oxidative stress, and enhanced levels of circulating procoagulant factors. In fact, the estimated risk for smokers of a dying from a CVD event is almost twice that of dying from lung cancer 17. The usefulness and cost-effectiveness of CT screening for lung cancer is currently being investigated in a number of randomized controlled trials (RCT) To combine this approach with the simultaneous screening for subclinical manifestations of CVD could be a very effective way to detect these two major smoking-induced diseases at an early stage. To reduce radiation exposure in a screening program requiring a number of serial CT examinations, current practice is to use non-gated, low-dose CT for lung cancer screening. Compared with gold standard techniques as EBCT or ECG-gated MDCT, calcium scoring in non-gated, low-dose CT suffers more from motion artefacts caused by the beating heart. This will result in decreased inter and intra-observer reproducibility and interscan agreement of an individual s calcium score 23,24. Nevertheless, to use non-gated, low-dose CT as a screening tool, the principal requirement should be a high interscan agreement of stratifying people in appropriate calcium score-based CVD risk categories 25. In the evaluation of the effectiveness of CAC screening especially in a high-risk population of heavy smokers - not only the prognostic value of CAC as single risk factor, but also the added value of CAC scores above and beyond the baseline CVD risk of the population should be taken into account. If, because of their longstanding smoking behaviour and its effect on other important CVD risk factors, all participants eligible for lung cancer screening would fall in the highest CVD risk category the benefits of additional screening for CAC would obviously be negligible. An analysis of reclassification of baseline CVD risk after adding CAC scores can further elucidate whether screening is effective in a population of heavy smokers. If so, the question rises whether more intensified smoking cessation counselling would benefit participants with an average smoking history of >40 pack-years or whether these people are perhaps exposed to so much tobacco that their risk of CVD events remains essentially unchanged. Chapters 5 to 9 are based upon data derived from the screen group of the Dutch-Belgian (NELSON) lung cancer CT screening trial, a population-based multicenter RCT among heavy (former) smokers who were followed-up for all-cause death and CVD events from 2004 to Chapter 5 describes the interscan agreement of CAC scores in non-gated, low-dose CT as part of lung cancer screening. Chapter 6 is the first attempt to investigate the independent prognostic value of CAC for predicting all-cause mortality and cardiovascular events when using non-gated, low-dose CT in a population of heavy smokers. In Chapter 7 the cross-sectional relationship between CAC and TAC in heavy smokers is 12

13 General Introduction presented. Also, the associations of CAC and TAC with different types of cardiovascular events (coronary artery disease versus non-cardiac CVD events) were investigated to establish whether CAC and TAC differentially predict types of CVD events. In Chapter 8 we determined the baseline distribution of CVD risk according to traditional CVD risk factors in a population of heavy smokers and investigated in how many cases adding the CAC score to these traditional risk factors would improve the correct classification of participants into their appropriate risk category. In Chapter 9, finally, we examined the effect of smoking cessation on the risk of subsequent CVD events in these longstanding smokers and, furthermore, compared the CAC and TAC scores of former smokers according to the time since they had given up their smoking habit. 13

14 Chapter 1 References 1. Last accessed Arad Y, Spadaro LA, Goodman K, Newstein D, Guerci AD. Prediction of coronary events with electron beam computed tomography. J Am Coll Cardiol. 2000;36: De Backer G, Ambrosioni E, Borch-Johnsen K et al. European guidelines on cardiovascular disease prevention in clinical practice. Third Joint Task Force of European and Other Societies on Cardiovascular Disease Prevention in Clinical Practice. Eur Heart J. 2003;24: Budoff MJ, Shaw LJ, Liu ST et al. Long-term prognosis associated with coronary calcification: observations from a registry of 25,253 patients. J Am Coll Cardiol. 2007;49: Greenland P, LaBree L, Azen SP, Doherty TM, Detrano RC. Coronary artery calcium score combined with Framingham score for risk prediction in asymptomatic individuals. JAMA. 2004;291: Shaw LJ, Raggi P, Schisterman E, Berman DS, Callister TQ. Prognostic value of cardiac risk factors and coronary artery calcium screening for all-cause mortality. Radiology. 2003;228: US Preventive Services Task Force. Screening for coronary heart disease: recommendation statement. Ann Intern Med. 2004;140: Takasu J, Budoff MJ, O Brien KD et al. Relationship between coronary artery and descending thoracic aortic calcification as detected by computed tomography: The Multi-Ethnic Study of Atherosclerosis. Atherosclerosis Wong ND, Sciammarella M, Arad Y et al. Relation of thoracic aortic and aortic valve calcium to coronary artery calcium and risk assessment. Am J Cardiol. 2003;92: Hara AK. Extracolonic findings at CT colonography. Semin Ultrasound CT MR. 2005;26: Xiong T, Richardson M, Woodroffe R, Halligan S, Morton D, Lilford RJ. Incidental lesions found on CT colonography: their nature and frequency. Br J Radiol. 2005;78: Kondos GT, Hoff JA, Sevrukov A et al. Electron-beam tomography coronary artery calcium and cardiac events: a 37-month follow-up of 5635 initially asymptomatic low- to intermediate-risk adults. Circulation. 2003;107: LaMonte MJ, FitzGerald SJ, Church TS et al. Coronary artery calcium score and coronary heart disease events in a large cohort of asymptomatic men and women. Am J Epidemiol. 2005;162: Koos R, Kuhl HP, Muhlenbruch G, Wildberger JE, Gunther RW, Mahnken AH. Prevalence and clinical importance of aortic valve calcification detected incidentally on CT scans: comparison with echocardiography. Radiology. 2006;241: Mahnken AH, Muhlenbruch G, Das M et al. MDCT detection of mitral valve calcification: prevalence and clinical relevance compared with echocardiography. AJR Am J Roentgenol. 2007;188: Witteman JC, Kok FJ, van Saase JL, Valkenburg HA. Aortic calcification as a predictor of cardiovascular mortality. Lancet. 1986;2: Ezzati M, Lopez AD. Estimates of global mortality attributable to smoking in Lancet. 2003;362: Blanchon T, Brechot JM, Grenier PA et al. Baseline results of the Depiscan study: a French randomized pilot trial of lung cancer screening comparing low dose CT scan (LDCT) and chest X-ray (CXR). Lung Cancer. 2007;58:

15 General Introduction 19. Henschke CI, McCauley DI, Yankelevitz DF et al. Early Lung Cancer Action Project: overall design and findings from baseline screening. Lancet. 1999;354: Lopes PA, Picozzi G, Mascalchi M et al. Design, recruitment and baseline results of the ITALUNG trial for lung cancer screening with low-dose CT. Lung Cancer. 2009;64: MacRedmond R, Logan PM, Lee M, Kenny D, Foley C, Costello RW. Screening for lung cancer using low dose CT scanning. Thorax. 2004;59: van Iersel CA, de Koning HJ, Draisma G et al. Risk-based selection from the general population in a screening trial: selection criteria, recruitment and power for the Dutch-Belgian randomised lung cancer multi-slice CT screening trial (NELSON). Int J Cancer. 2007;120: Horiguchi J, Yamamoto H, Hirai N et al. Variability of repeated coronary artery calcium measurements on low-dose ECG-gated 16-MDCT. AJR Am J Roentgenol. 2006;187:W1-W Shemesh J, Evron R, Koren-Morag N et al. Coronary artery calcium measurement with multi-detector row CT and low radiation dose: comparison between 55 and 165 mas. Radiology. 2005;236: Kim SM, Chung MJ, Lee KS, Choe YH, Yi CA, Choe BK. Coronary calcium screening using low-dose lung cancer screening: effectiveness of MDCT with retrospective reconstruction. AJR Am J Roentgenol. 2008;190:

16

17 Part 1 Subclinical coronary and aortic calcifications in routine care, diagnostic chest CT as predictor of CVD events Chapter 2 Prevalence of incidental findings in CT screening of the chest: a systematic review. P.C.A. Jacobs, W.P.Th.M. Mali, D.E. Grobbee, Y. van der Graaf. JCAT.2008;32: Chapter 3 Semi-quantitative assessment of cardiovascular disease markers in multislice CT of the chest: inter- and intra-observer agreement. P.C.A. Jacobs, M. Prokop, A.L. Oen, Y. van der Graaf, D.E. Grobbee, W.P.Th.M. Mali. Accepted for JCAT. Chapter 4 Use of unrequested information from routine care, diagnostic chest CT: the case of coronary and aortic calcifications. P.C.A. Jacobs, M.J.A. Gondrie, W.P.Th.M. Mali, A.L. Oen, M. Prokop, D.E. Grobbee, Y. van der Graaf, on behalf of the PROVIDI Study Group. Submitted.

18

19 CHAPTER 2 Prevalence of incidental findings in CT screening of the chest: a systematic review

20 Chapter 2 Abstract Objective To perform a systematic review on the prevalence of incidental findings in CT screening studies of the chest. Methods We selected CT screening studies of the chest (screening for coronary artery disease (coronary calcium and CT coronary angiography) and lung cancer screening). Screening protocols, descriptions of baseline characteristics, range of incidental findings, and recommendations for follow-up were abstracted. Results Eleven chest CT screening studies were identified. The proportion of people with at least one imaging abnormality requiring follow-up varied widely between studies (3-41.5%). This was largely due to considerable variation in follow-up recommendations for incidental findings across studies. Analyzed by subgroup, 7.7% (95% CI, %) of 6421 participants in coronary artery disease (CAD) screening had further investigations compared to 14.2% (95% CI, %) of 4531 participants in lung cancer screening. Conclusions In this review, 7.7% and 14.2% of patients undergoing either CAD or lung cancer screening with computed tomography were found to have clinically significant incidental findings requiring additional investigations. 20

21 Prevalence of incidental findings in CT screening of the chest Introduction Major advances over the last decade in the performance of computed tomography (CT) have revolutionized the ability of radiologists to detect the most subtle, subclinical manifestations of disease. This has prompted a number of large scale CT screening studies of the chest for the early detection of diseases such as lung cancer and coronary heart disease Large scale acceptance and implementation of these types of screening programmes has so far been hampered primarily by a lack of demonstrated benefit on disease survival. Early detection of disease can theoretically result in a substantial reduction of mortality, but it may also induce so-called lead-time bias or overdiagnosis bias Another important and less frequently addressed element in the evaluation of these studies is the likelihood of so-called incidental findings. Increased performance of CT has led to a steep increase in their number. An incidental finding is usually defined as an imaging abnormality not related to the indication for obtaining the CT scan. In the case of CT screening, every abnormal finding that cannot be related pathophysiologically to the target disease should be regarded as incidental. These we can roughly divide into four main categories: (A) findings with immediate therapeutic consequences (e.g. newly diagnosed cancer); (B) findings with unquestionable clinical and/or prognostic relevance, requiring follow-up and/or therapeutic intervention (e.g. substantial calcification of the cardiac valves); (C) findings with possible clinical and/or prognostic relevance requiring follow-up (e.g. indeterminate pulmonary nodules); and (D) findings without a proven clinical and/or prognostic significance not requiring follow-up (e.g. fatty degeneration of back musculature). In the literature on incidental findings, most reports make a distinction between clinically significant and clinically non-significant findings. Clinically significant is defined as an imaging abnormality requiring further (diagnostic) work-up. In CT screening of healthy individuals with a low a priori chance of serious disease, the vast majority of these clinically significant incidental findings will belong to category C. Follow-up of these lesions will sometimes turn out to have been a legitimate decision that detects disease at a curable stage or discloses a subclinical manifestation of a future disease suitable for preventive therapeutic measures. Due to the expected high rate of false-positive findings, however, follow-up will have been unnecessary in a more substantial number of cases. This will possibly cause the patient more anxiety (maybe even iatrogenic harm) and certainly will add to the overall costs of the screening procedure 17. Therefore, a comprehensive assessment of CT as a screening tool should include a thorough appreciation of the prevalence and nature of these incidental findings. Previously, abdominal incidental findings detected at CT colonography have been reviewed 18,19, but to date this has not been performed for CT screening of the chest. The purpose of this study was to perform a systematic review of the literature regarding the prevalence of incidental findings detected in CT screening of the chest. 21

22 Chapter 2 Materials & Methods Identification and selection of the literature A search of relevant publications dealing with the prevalence and significance of incidental findings detected with CT screening of the chest was performed using the PUBMED and EMBASE databases (January 2007) and the reference lists of included studies were searched (described in full in Appendices 1 and 2). For completeness, the search strategy was not restricted by language, but finally only reports in English were included. The searches were restricted to cover a period from 1990 to the present and the search was last updated on 29 May This sensitive search identified a large number of titles. Each title and abstract was reviewed by two authors to assess the relevance for the present review. Two categories of CT screening studies were identified for data extraction: (1) CAD screening and (2) lung cancer screening. The eligibility criteria were: participants were screened with standard single-slice helical CT, multidetector row CT (MDCT) and electron-beam CT (EBCT) as a lot of screening studies for coronary heart disease during the 90s (that is prior to the development of third and fourth generation MDCT scanners) were performed with EBCT; in studies using CT as a follow-up instrument, separate data on the prevalence of incidental findings in the baseline CT had to be provided; studies had to include data on the number of significant findings and had to specify how they defined a clinically significant finding. Any disagreements between the two reviewers were resolved in a consensus meeting. Relevant papers were retrieved and the full text appraised independently by two authors. Data extraction Data extraction included details on type of screening and the screening protocol, participant characteristics and range of incidental lesions reported. Our principal focus of interest was on the number of lesions deemed clinically significant by the authors. Almost unanimously the definition of clinically significant was based on whether or not the lesion was considered to require any further medical evaluation (most often additional imaging). Additional measures abstracted from these studies were the proportion of participants with one or more incidental findings, the exact location and nature of a finding and the proportion of people with newly diagnosed cancers (excluding the target disease lung cancer in lung cancer screening). Early detection might be life-saving. We therefore included them as a separate category. Results Summarized characteristics of Identified Studies Our primary literature search identified 586 references. On the basis of titles and abstracts, 145 articles were read in full. Of these, 10 studies of chest CT screening met our eligibility criteria 22

23 Prevalence of incidental findings in CT screening of the chest for inclusion in this review. Searching the reference list of these articles, one more article could be identified. Table 1 summarises the study characteristics of these 11 studies. Seven are screening studies for the presence of coronary artery disease and four are lung cancer screening studies 5, In addition we found two articles presenting data on specific incidental findings in the Early Lung Cancer Action Project (ELCAP) screening study population 30,31. Although not meeting the eligibility criteria for this review, their most important conclusions will also be discussed. There was no overlap of participants across these studies. Four of the eleven studies were conducted in the USA. The remaining seven took place in Finland, Germany, Ireland, the Netherlands, Switzerland, Israel and Japan. The study populations were generally large: six out of eleven studies exceeded 1000 patients and, in total, people were scanned. Mean age was consistent across most studies averaging 56.7 years (range: 20-87). Smoking status was not or incompletely available in two studies 20,23. Across the other studies, the mean proportion of current or former smokers was 57.6% (range %). CAD Screening Four studies are screening studies for the presence of coronary calcium. These studies used electron-beam CT (EBCT) systems with comparable scanning protocols (single breath hold, 3mm slice thickness, ECG-gated and reconstruction at maximum field-of-view (FOV)) 20,22,23,25. The remaining three studies are CT coronary angiography studies 21,24,26. The studies by Haller 21 and Onuma 24 are performed in patients suspected of CAD, whereas the study by Gil 26 was conducted in asymptomatic subjects. They used a 16-slice or a 64-slice scanner; datasets were reconstructed with a large FOV to include the entire chest with a slice thickness of 1mm, respectively 5mm. Both enhanced and non-enhanced scans were performed. Other CT parameters included a tube voltage of kvp, tube currents ma and a table feed of mm/rotation. In all but one study, reviewing of the scans was performed in at least two different tissue windows. In all studies images were reviewed by experienced radiologists (Table 1). In five studies, a clinically significant finding was defined as a finding requiring further medical follow-up. However, in Haller 21 and Elgin 20 imaging abnormalities were subdivided into minor or major categories. In this case, all nonminor findings were regarded as requiring follow-up. The studies by Hunold 23 and Gil 26 were the only studies to include a number of non-coronary cardiac abnormalities, whereas the other studies restricted themselves to extra-cardiac abnormalities. For matters of comparability we decided not to include these additional cardiac findings in this review. Only two studies reported - albeit limited information on clinical follow-up 22,25. 23

24 Chapter 2 Table 1 Study characteristics Reference Type of scanner Protocol specifications Range of findings reported and definitions CAC screening Schragin et al (25) Electron-beam CT Retrospective ECG-triggering; ST*: 3mm; scanning range: aortic root cardiac apex; reviewing: soft tissue + lung windows printed onto film Noncardiac findings; significant: requiring followup Horton et al (22) Electron-beam CT Retrospective ECG-triggering; ST: 3mm; scanning range: pulmonary trunk cardiac apex; FOV#: 350mm; reviewing: three windows Noncardiac findings; significant: requiring followup Elgin et al (20) Electron-beam CT Retrospective ECG-triggering; ST: 3mm; scanning range: not specified; reviewing: bone + lung windows Noncoronary findings; nonminor: requiring noninvasive imaging test Hunold et al (23) Electron-beam CT Retrospective ECG-triggering; ST: 3mm; scanning range: pulmonary trunk cardiac apex; contrast studies after native scan; reviewing: soft tissue window only Noncoronary findings; numbers of findings with diagnostic consequences separately provided Haller et al (22) Multislice CT (16-slice) Retrospective ECG-triggering; ST: 1mm; collimation: 16x0.75mm; 400 mas at 120 kv; scanning range: not specified; contrast enhanced; reviewing: three windows Noncardiac findings; minor: no immediate work-up Onuma et al (24) Multislice CT (16 and 64-slice) Retrospective ECG-triggering; ST: 5mm; collimation: 16x0.75/32x0.6mm; mas at 120 kv; scanning range: pulmonary trunk cardiac apex; contrast studies after native scans; reviewing: soft tissue + lung windows Noncardiac findings; significant: requiring followup Population characteristics and inclusion criteria Self-referred and referred; consecutive patients not part of a research protocol Referred patients, symptomatic for coronary artery disease(cad); consecutive referred and self-referred asymptomatic patients Consecutive, asymptomatic, year old active duty army personnel; part of periodic physical examination Consecutive referred patients, symptomatic for or suspected of CAD Consecutive referred patients, symptomatic for CAD Consecutive referred patients suspected for CAD 24

25 Prevalence of incidental findings in CT screening of the chest Table 1 (continued) Study characteristics Reference Type of scanner Protocol specifications Range of findings reported and definitions Gil et al (26) Multislice CT (16-slice) Retrospective ECG-gating; ST: 1mm; collimation 16x0.75mm; 400 mas at 140kV; scanning range: apex lungs diaphragmatic surface of the heart; contrast studies after native scan; reviewing: 4 windows + reformations Lung cancer screening Noncardiac findings; significant: requiring followup Swensen et al (27) Multislice CT (Low-dose CT) ST: 5mm; pitch: 1.5; 40 mas at 120 kv; scanning range: sternal notch crista iliaca; reviewing: three windows No limitations; significant: requiring follow-up MacRedmond et al (5) Single-slice CT (Low-dose CT) ST: 10mm; pitch: 2.0; <50 mas; tube voltage not specified; scanning range: not specified; reviewing: not specified No limitations; significant: requiring follow-up Van de Wiel et al (28) Multislice CT (16-slice) (Low-dose CT) ST: 1mm; pitch: 1.5; 20 mas at kv; scanning range: lung apex lung bases; reviewing: soft tissue + lung windows No limitations; possibly clinically relevant: requiring follow-up Vierriko et al (29) Single and Multislice CT (16-slice) (High-resolution CT) ST: mm; pitch: not specified; mas at kv; scanning range: lung apex costophrenic angle; reviewing: not specified No limitations; number of findings with diagnostic consequences separately provided *ST: slice thickness #FOV: field-of-view Population characteristics and inclusion criteria Self-referred, asymptomatic for CAD consecutive patients not part of a research protocol Self-referred; age >50years; >20 pack years; still smoking at age 40; no history of cancer and medically fit Self-referred; age >50 years; >10 pack years; still smoking at age 45; no history of cancer and medically fit Recruited; age 50-75; >15 cigarettes/day for more than 25 years; quit smoking less than 10 years ago; no history of cancer unless curatively treated >5 years ago; medically fit Medically fit asbestos exposed workers recruited from screening programme( ) and asbestos related disease patients recruited from clinics of occupational medicine 25

26 Chapter 2 Five studies reported the total number of patients in which at least one incidental finding was scored ranging from 8 to 58.1% with a mean prevalence of 34% (95% CI, %). The proportion of people with at least one imaging abnormality requiring further medical follow-up varied widely between studies (3-41%). The mean prevalence of these clinically significant findings was 7.7% (95% CI, 7.0%-8.3%). The wide range suggests that - although most studies employed the same definition of a clinically significant finding the actual decision to recommend a follow-up investigation differed substantially from one clinic to another. To further illustrate these differences in recommendations, Table 3 presents data on the imaging abnormalities that were qualified as clinically significant in at least two of six studies subdivided according to their location and nature. Especially in the case of pulmonary nodules, pleural disease, lymphadenopathy and liver abnormalities there is considerable variation in the number of detected findings requiring further medical followup between these studies. People diagnosed with a pulmonary nodule are the category most often referred for further testing. Almost half of all imaging abnormalities requiring follow-up consists of suspicious (non-calcified and 3-10mm) pulmonary nodules (185/413). Suspicious lung nodules were observed in 3.4% (185/5421) of participants (95% CI, %) (Table 2). Also, 11 cases of newly diagnosed cancer were detected as a result of 5163 screening CT examinations (0.21%, 95% CI, %). These constituted 7 cases of lung cancer. Lung Cancer Screening Three studies had comparable low-dose CT (LDCT) scanning protocols specifically suitable for lung cancer screening 5,27,28, using both single-slice and multidetector CT systems with slice thickness of 1-10mm. Effective radiation dose was estimated to be between 0.4 and 1.6mSv. Other CT parameters included tube voltages of kVp, tube currents <50mA and a pitch of In the study by Vierikko 29 high-resolution CT (HRCT) images were obtained with a slice thickness of mm, tube voltage of kVp and a tube current of mA. The study by Swensen 27 was the only extending the scanning range to include a large part of the abdomen to the level of the iliac spines (Table 2). All but one study evaluated CT images in at least the lung and soft-tissue window and reviewing was performed by experienced radiologists in all studies. Three studies clearly stated that clinically significant findings were findings requiring further medical follow-up. Although comparable in design and baseline population characteristics, these four studies had substantially different outcomes on a number of important points (Table 2). The proportion of clinically significant incidental findings in these high-risk populations was 14.2% (95% CI, %). The highest prevalence of clinically significant findings was reported by MacRedmond 5 (26.9%, 95% CI, %). Prevalence rates in the other studies were substantially lower ( %). The mean proportion of patients with at least one incidental finding was 65.2% 26

27 Prevalence of incidental findings in CT screening of the chest Table 2 Frequency and Range of Incidental Findings encountered on Chest CT Reference No. of Partcipans Median age (range) Patients with incidental lesions (%) Significant findings(%) Newly diagnosed incidental cancer (%) Lung nodules (%) Current and former smoking CAC screening Schragin et al (25) (21-86) 278 (20.5%) 57 (4.2%) 1 (0.07%) 46 (3.4%) 12.8% current smokers; 27.9% former smokers Horton et al (22) (23-87) Not specified 103 (7.8%) 1 (0.08%) 65 (4.9%) 6.7% current smokers; 18.3% former smokers Elgin et al (20) (39-46) 79 (8%) 54 (5.4%) Not specified Not specified 13% current smokers Hunold et al (23) (20-86) 953 (53%) 50 (2.8%) 3 (0.17%) 8 (0.44%) Smoking status not available Haller et al (22) (not specified) 41 (24.7%) 8 (4.8%) 2 (1.2%) Not specified 42% cigarette smokers Onuma et al (24) (not specified) 292 (58.1%) 114 (22.7%) 4 (0.8%) 33 (6.6%) 13% current smokers; 39% former smokers Gil et al (26) (40-86) Not specified 107 (41.5%) Not specified 52 (20.2%) 20.1% current smokers, 28.3% former smokers Lung cancer screening CAC (%) Swensen et al (27) (50-85) Not specified 210 (14%) 14 (0.92%) Not specified 100% current or former smokers MacRedmond et al (5) (50-74) 276 (61.5%) 121 (26.9%) * - 64 (14.3%) 100% current or former smokers Van de Wiel et al (28) (not specified) 1409 (73%) 169 (7%) Not specified 1306 (67.7%) 100% current or former smokers Vierriko et al (29) (45-87) 277 (43.8%) 46 (7.3%) (21.8%) ** 19.9% current smokers; 58% former smokers * This study was the only study to count (severe) emphysema as clinically significant finding. For matters of comparison, we subtracted emphysema from the total number of clinically significant findings reported in their study manuscript. ** This study only included the most evident coronary artery calcifications in their report. 27

28 Chapter 2 Table 3 Significant findings reported in at least two CAC screening studies Imaging finding Schragin(25) (N=1356) Horton(22) (N=1326) Hunold(23) (N=1812) Haller(21) (N=166) Onuma(24) (N=503) Gil (26) (N=258) Pulmonary nodule 46 (3.4%) 53 (4.0%) 8 (0.44%) 4 (2.4%) 25 (5.0%) 49 (19.0%) Pulmonary nodule (>1cm) 1 (0.07%) 12 (0.90%) * * 8 (1.6%) 3 (1.2%) Infiltrates 3 (0.22%) 24 (1.8%) - 3 (1.8%) 10 (2.0%) 8 (3.1%) Interstitial disease 3 (0.22%) - 1 (0.06%) 2 (1.2%) - 3 (1.2%) Lymphadenopathy 2 (0.15%) - 6 (0.33%) 5 (3.0%) 1 (0.20%) 4 (1.5%) Liver disease 1 (0.07%) 8 (0.6%) 14 (0.77%) - NS 12 (4.6%) Sclerotic/osteolytic bone - 2 (0.15%) (0.8%) Breast abnormalities - 2 (0.15%) 2 (0.11%) - 3 (0.60%) - Esophageal thickening - 1 (0.08%) 3 (0.17%) (0.4%) Aortic ectasia/aneurysm 1(0.07%) - 8 (0.44%) 2 (1.2%) 6 (1.2%) 4 (1.5%) Aortic dissection (0.06%) - 1 (0.20%) - Pleural disease (0.11%) 2 (1.2%) 16 (3.2%) - Thymic lesion (0.06%) (1.1%) Diaphragmatic hernia (0.11%) 2 (1.2%) - 6 (2.3%) Thyroid disease (0.40%) 8 (3.1%) Adrenal mass (3.1%) 28

29 Prevalence of incidental findings in CT screening of the chest (95% CI, 63.5%-66.9%). Reporting of coronary artery calcifications ranged from all detectable calcifications 28 to only the most evident calcifications 29. Accordingly, their prevalence ranged from 14.3% to as high as 67.7% (Table 2). In the ELCAP population 31, calcifications were detected in 2706 of 4250 participants (64%). Calcifications were invariably classified as not clinically significant findings. Table 4 presents a detailed overview of those clinically significant findings detected in at least two studies. Only the study by MacRedmond 5 reports a substantial number of significant findings related to airway disease. In contrast, the study by van der Wiel 28 explicitly defines all cases of pulmonary fibrosis and bronchiectasis a priori as not clinically relevant. This again highlights the variation in follow-up recommendations made based upon the initial CT scan. Apart from the study by MacRedmond 5, the majority of findings were located in the upper abdomen most often consisting of indeterminate masses in the kidneys or liver. Only the study by Swensen 27 detected a considerable number of aortic aneurysms (51/1520). A total of 43 newly detected cancers in 2602 prevalence scans were reported. Of these, 28 were prevalent lung cancer cases (+ 1 mesothelioma) and as such constituted the target disease of screening. All fourteen other cancer cases came from just one study 27 and included 2 gastric tumors, 1 pheochromocytoma, 1 pancreatic adenocarcinoma and 4 renal cell carcinomas. In comparison, in a study conducted in 9263 ELCAP participants individuals (0.77%) presented with a mediastinal mass. Only 3 cases (0.03%) of newly diagnosed cancer were detected (one thymic carcinoma and 2 esophageal cancers). 29

30 Chapter 2 Table 4 Significant findings reported in at least two lung cancer screening studies Imaging finding Swensen(27) (N=1520) MacRedmond(5) (N=449) Van de Wiel(28) (N=1929) Vierikko(29) (N=633) Airway disease Tracheal diseae 7 (0.46%) (0.30%) Consolidation/Infiltrate 2 (0.13%) 11 (2.5%) # 4 (0.60%) Bronchiectasis 11 (0.72%) 44 (9.8%) # - Interstitial disease - 6 (1.3%) # - Interstitial disease 61 (4.0%) 41 (9.1%)* 50 (2.6%) 16 (2.5%)* Renal mass 37 (2.4%) Not specified 48 (2.5%) Not specified Renal calculus 24 (1.6%) Not specified - Not specified Hydronephrosis - Not specified 2 (0.10%) Not specified Liver disease - 41 (9.1%)* 76 (3.9%) 16 (2.5%)* Breast abnormalities 20 (1.3%) - 1 (0.05%) - Aortic aneurysm 51 (3.4%) 1 (0.2%) - - Pericardial disease 9 (0.59%) - 1 (0.05%) - Pleural disease 4 (0.26%) 4 (0.9%) # 5 (0.80%) Lymphadenopathy 5 (0.80%) (1.0%) Adrenal mass 36 (2.4%) - 1 (0.05%) 9 (1.4%) Gastric tumor 2 (0.13%) 1 (0.2%) - - Thyroid disease - 9 (2.0%) 9 (0.47%) - * In these two studies no distinction was made between renal and liver disease. Figures reported here are therefore totals of both types of findings. # Although figures relating to these types of findings are reported in this study, the authors explicitly state that they considered these findings a priori not clinically relevant. 30

31 Prevalence of incidental findings in CT screening of the chest Discussion We reviewed CT screening studies of the chest that reported on the prevalence of incidental findings. Results of CAD screening and lung cancer screening suggest that the prevalence of clinically significant incidental findings is high. In the seven CAD screening studies, clinically significant findings were described in 7.7% of patients. The proportion of patients with clinically significant findings in lung cancer screening studies was about twice as high: 14.2%. From these data, it seems that there is a linear relationship between the scanning range and the proportion of clinically significant incidental findings detected. Cardiac CT has a limited scanning range and field of view (on average depicting only one to two-thirds of the lung fields) as opposed to lung cancer screening (at least the entire thorax and a variable part of the upper abdomen). Alternatively, the higher percentages in the lung cancer screening studies might be explained by the history of heavy smoking in all of its participants. In the CAD screening studies on average only 1 in three participants were active or former smokers. Only in the study by Schragin et al 25 a structured longer term follow-up study using the National Death Index registration was additionally performed, trying to ascertain the clinical significance of previously scored imaging abnormalities on all-cause mortality; one of seven deaths in this cohort (n=1356) could be, retrospectively, attributed to an incidental finding at baseline screening. An important shortcoming in this study is the very low number of personyears of follow-up. Secondly, since their study population consisted of referred or self-referred people undergoing cardiac CT, cardiovascular deaths in this cohort could not be attributed to incidental findings. Thirdly, the only endpoint in this design is all-cause mortality, while especially in a short period of follow-up it seems to be much more rewarding to look at incidental lesion-related morbidity (hospitalization, interventional procedures). Since data on incidental findings usually are a by-product of the original research questions in these screening studies, it is not unlikely that our literature search (based on keywords in titles and abstracts) missed a number of studies providing data on this issue as part of their discussion. By focusing our search strategy on identifying original articles from CT screening studies on CAC and lung cancer and reading through these papers to see whether incidental findings were reported, we have tried to limit this as much as possible. Due to the heterogeneity of the different types of study design in each type of screening study (e.g. type of scanners, scanning protocols, baseline population characteristics), we feel that pooling all the data does not contribute to a thorough appreciation of the issue of incidental findings. Therefore, we have limited ourselves to presenting proportion estimates per subgroup. Furthermore, clinical significance of an imaging finding was determined by whether or not a recommendation for follow-up was issued. Although in some cases (e.g. pulmonary nodules) firm guidelines exist on when a lesion should or should not be followedup 32, it is likely that in many other cases this decision is based on a more individual judgement 31

32 Chapter 2 by one of the many radiologists that have reviewed these images. These subjective factors affect the generalization of the results. Our results are consistent with results from the study by Messersmith et al 33 that showed one or more incidental findings in 45% of helical CT scans of the abdomen in 321 consecutive emergency department patients; about half of these lesions were graded as moderate or severe, roughly corresponding to what is usually designated as clinically significant. A study by Xiong et al 19 reviewed the prevalence of incidental findings in CT colonography studies; they found incidental lesions in 40% of 3280 patients and a 13.8% prevalence of lesions prompting further investigation. Compared with CT screening of the chest, these studies on CT colonography on average report a slightly higher number of incidental findings and a definitely higher prevalence of incidental cancer cases (2.7% of extracolonic cancers reported in the study by Xiong 19 ). The study by Swensen et al 27 is the only study in the present review that has described a comparable prevalence (0.92% (14/1520)) of incidentalomas. These two differences can partly be explained by the fact that most patients in these CT colonography studies were in their sixties - approximately a decade older than in our study - and that they generally describe symptomatic patients, where the disease prompting the CT investigation is not unlikely to be a priori of extracolonic origin. Furthermore, we conclude that studies reporting incidental findings although generally using comparable definitions to classify their findings according to importance show a great variation in the type and number of findings that the authors deemed clinically significant. Even where firm guidelines exist as to whether a lesion should be followed-up - as in the case of pulmonary nodules the variation in number of nodules recommended for additional investigations was substantial. Therefore, this review shows that in the event of an incidental finding detected in CT screening there is lack of uniformity in recommendations regarding the follow-up strategy of that finding. Also, the number of cases of serious disease (e.g. incidental cancers) in the studies presently reviewed seems to be low relative to the total number of findings deemed clinically significant. This is not altogether unexpected given the low pre-test probability of serious disease in these (relatively) healthy subjects. This low prevalence of serious conditions is one of the arguments critics have raised against CT as a screening tool and a balanced assessment of CT screening should include a cost-effectiveness analysis including the issue of incidental findings. Even a proportion of 8% incidental findings - comparable to the lower proportions described in this review would cause 1 in 12 participants of a screening study to undergo additional work-up. Imaging characteristics related to a normal pattern of ageing and encountered among a majority of the elderly population (coronary calcification, decreased bone density, decreased lung density, pleural thickening) have been ignored or classified as not clinically significant by these studies. These findings might prove to be of prognostic significance in predicting 32

33 Prevalence of incidental findings in CT screening of the chest serious and potentially preventable morbidity and mortality 34. A thorough appreciation of incidental findings in CT screening cannot only be attained by calculating the potential costs brought about by additional work-up, but also has to take into account the potentially valuable information contained in these images. In conclusion, incidental findings are common feature in CT screening of the chest. However, they are inconsistently reported and followed-up due to a lack of consensus and the exact definition of a clinically significant incidental finding. This review shows the need for more uniform decision making algorithms in dealing with incidental findings. This can be achieved by further research on the longer term follow-up of the full range of incidental findings. 33

34 Chapter 2 References 1. Henschke CI, McCauley DI, Yankelevitz DF et al. Early Lung Cancer Action Project: overall design and findings from baseline screening. Lancet. 1999;354: Gohagan JK, Marcus PM, Fagerstrom RM et al. Final results of the Lung Screening Study, a randomized feasibility study of spiral CT versus chest X-ray screening for lung cancer. Lung Cancer. 2005;47: Sone S, Li F, Yang ZG et al. Results of three-year mass screening programme for lung cancer using mobile low-dose spiral computed tomography scanner. Br J Cancer. 2001;84: Diederich S, Wormanns D, Semik M et al. Screening for early lung cancer with low-dose spiral CT: prevalence in 817 asymptomatic smokers. Radiology. 2002;222: MacRedmond R, Logan PM, Lee M, Kenny D, Foley C, Costello RW. Screening for lung cancer using low dose CT scanning. Thorax. 2004;59: Swensen SJ, Jett JR, Hartman TE et al. CT screening for lung cancer: five-year prospective experience. Radiology. 2005;235: Haberl R, Becker A, Leber A et al. Correlation of coronary calcification and angiographically documented stenoses in patients with suspected coronary artery disease: results of 1,764 patients. J Am Coll Cardiol. 2001;37: Nallamothu BK, Saint S, Bielak LF et al. Electron-beam computed tomography in the diagnosis of coronary artery disease: a meta-analysis. Arch Intern Med. 2001;161: Budoff MJ, Diamond GA, Raggi P et al. Continuous probabilistic prediction of angiographically significant coronary artery disease using electron beam tomography. Circulation. 2002;105: Kondos GT, Hoff JA, Sevrukov A et al. Electron-beam tomography coronary artery calcium and cardiac events: a 37-month follow-up of 5635 initially asymptomatic low- to intermediate-risk adults. Circulation. 2003;107: O Malley PG, Taylor AJ, Gibbons RV et al. Rationale and design of the Prospective Army Coronary Calcium (PACC) Study: utility of electron beam computed tomography as a screening test for coronary artery disease and as an intervention for risk factor modification among young, asymptomatic, active-duty United States Army Personnel. Am Heart J. 1999;137: Nasir K, Michos ED, Blumenthal RS, Raggi P. Detection of high-risk young adults and women by coronary calcium and National Cholesterol Education Program Panel III guidelines. J Am Coll Cardiol. 2005;46: Marcus PM, Bergstralh EJ, Zweig MH, Harris A, Offord KP, Fontana RS. Extended lung cancer incidence follow-up in the Mayo Lung Project and overdiagnosis. J Natl Cancer Inst. 2006;98: O Rourke RA, Brundage BH, Froelicher VF et al. American College of Cardiology/American Heart Association Expert Consensus document on electron-beam computed tomography for the diagnosis and prognosis of coronary artery disease. Circulation. 2000;102: Fenton JJ, Deyo RA. Patient self-referral for radiologic screening tests: clinical and ethical concerns. J Am Board Fam Pract. 2003;16: Black WC. Overdiagnosis: An underrecognized cause of confusion and harm in cancer screening. J Natl Cancer Inst. 2000;92: Stone JH. Incidentalomas--clinical correlation and translational science required. N Engl J Med. 2006;354:

35 Prevalence of incidental findings in CT screening of the chest 18. Hara AK. Extracolonic findings at CT colonography. Semin Ultrasound CT MR. 2005;26: Xiong T, Richardson M, Woodroffe R, Halligan S, Morton D, Lilford RJ. Incidental lesions found on CT colonography: their nature and frequency. Br J Radiol. 2005;78: Elgin EE, O Malley PG, Feuerstein I, Taylor AJ. Frequency and severity of incidentalomas encountered during electron beam computed tomography for coronary calcium in middle-aged army personnel. Am J Cardiol. 2002;90: Haller S, Kaiser C, Buser P, Bongartz G, Bremerich J. Coronary artery imaging with contrast-enhanced MDCT: extracardiac findings. AJR Am J Roentgenol. 2006;187: Horton KM, Post WS, Blumenthal RS, Fishman EK. Prevalence of significant noncardiac findings on electron-beam computed tomography coronary artery calcium screening examinations. Circulation. 2002;106: Hunold P, Schmermund A, Seibel RM, Gronemeyer DH, Erbel R. Prevalence and clinical significance of accidental findings in electron-beam tomographic scans for coronary artery calcification. Eur Heart J. 2001;22: Onuma Y, Tanabe K, Nakazawa G et al. Noncardiac findings in cardiac imaging with multidetector computed tomography. J Am Coll Cardiol. 2006;48: Schragin JG, Weissfeld JL, Edmundowicz D, Strollo DC, Fuhrman CR. Non-cardiac findings on coronary electron beam computed tomography scanning. J Thorac Imaging. 2004;19: Gil BN, Ran K, Tamar G, Shmuell F, Eli A. Prevalence of significant noncardiac findings on coronary multidetector computed tomography angiography in asymptomatic patients. J Comput Assist Tomogr. 2007;31: Swensen SJ, Jett JR, Sloan JA et al. Screening for lung cancer with low-dose spiral computed tomography. Am J Respir Crit Care Med. 2002;165: van de Wiel JC, Wang Y, Xu DM et al. Neglectable benefit of searching for incidental findings in the Dutch--Belgian lung cancer screening trial (NELSON) using low-dose multidetector CT. Eur Radiol Vierikko T, Jarvenpaa R, Autti T et al. Chest CT screening of asbestos-exposed workers: lung lesions and incidental findings. Eur Respir J. 2007;29: Henschke CI, Lee IJ, Wu N et al. CT screening for lung cancer: prevalence and incidence of mediastinal masses. Radiology. 2006;239: Shemesh J, Henschke CI, Farooqi A et al. Frequency of coronary artery calcification on low-dose computed tomography screening for lung cancer. Clin Imaging. 2006;30: MacMahon H, Austin JH, Gamsu G et al. Guidelines for management of small pulmonary nodules detected on CT scans: a statement from the Fleischner Society. Radiology 2005 Nov. 237; Messersmith WA, Brown DF, Barry MJ. The prevalence and implications of incidental findings on ED abdominal CT scans. Am J Emerg Med. 2001;19: O Malley PG, Taylor AJ, Jackson JL, Doherty TM, Detrano RC. Prognostic value of coronary electron-beam computed tomography for coronary heart disease events in asymptomatic populations. Am J Cardiol. 2000;85:

36 Chapter 2 Appendix I Literature search results Identified on scanning titles and abstracts n = 586 Excluded Reading full text papers n = 441 * n = 145 Excluded Full papers for appraisal n = 133 ** n = 12 Flowchart of identification and inclusion of studies. * Reasons for exclusion: articles not in the English language; aricles not reporting relevant data for the present review based on reading through Titles and Abstracts only. ** Reasons for exclusion: articles not meeting one or more of the eligibility criteria set for this review based on reading the full text papers. 36

37 Prevalence of incidental findings in CT screening of the chest Appendix II Search Strategy The following search terms were used to identify relevant articles: Pubmed/Embase ( ) Lung cancer screening exp Lung neoplasms/ all subheadings exp Carcinoma, Non-small-cell lung/ exp Carcinoma, Small-cell lung/ lung cancer$ lung neoplasm$ small cell carcinoma 1 OR 2 OR 3 OR 4 OR 5 OR 6 exp Mass screening/ all subheadings exp Tomography, X-ray computed/ all subheadings CT screening OR CT scan$ 9 OR 10 7 AND 8 AND 11 Last search 26 January 2007 (449 results) Coronary artery calcium screening exp Coronary disease/ all subheadings Coronary artery calcium coronary calcification$ CAC CC 1 OR 2 OR 3 OR 4 OR 5 ancillary OR accidental OR noncardiac OR incidenta$ OR noncoronary OR extracardiac 6 AND 7 exp Tomography, X-ray computed/ all subheadings CT screening OR CT scan$ 9 OR 10 8 AND 11 Last search 26 January 2007 (106 results) 37

38

39 CHAPTER 3 Semi-quantitative assessment of cardiovascular disease markers in multislice CT of the chest: inter- and intra-observer agreement

40 Chapter 3 Abstract Objective To investigate the inter- and intra-observer agreement for the semi-quantitative assessment of markers of subclinical cardiovascular disease (CVD) as identified by routine care, diagnostic computed tomography (CT) scans of the chest, in order to improve the quality of reporting of these incidental findings. Methods Two observers independently evaluated 109 consecutive chest CT scans in routine care, clinical patients from one tertiary referral center. All non-gated, contrast-enhanced scans were acquired on a 16-slice CT scanner. Images were scored for the presence of aortic wall abnormalities, coronary artery calcifications, calcifications of the heart valves, the thoracic aorta and the proximal supra-aortic arteries. Furthermore, the presence of left ventricular scarring and elongation of the aorta were recorded. All markers were scored on a semiquantitative scale. Inter- and intra-observer agreement are presented as weighted kappa and intraclass correlation coefficients (ICC). Results Inter- and intra-observer agreement for individual markers was good to excellent with weighted κ coefficients of for inter-observer agreement and weighted κ coefficients of for intra-observer agreement. Conclusion Semi-quantitative assessment of subclinical CVD markers in routine care, diagnostic chest CT scans is possible with good to excellent inter- and intra-observer agreement. Use of these definitions in clinical practice will enable a more standardized assessment and reporting of incidental findings in diagnostic chest CT. 40

41 Semi-quantitative assessment of cardiovascular disease markers in multislice CT of the chest Introduction Cardiovascular disease (CVD) remains the leading cause of death in developed countries 1. Since many risk factors for CVD are potentially modifiable by specific preventive strategies 2, timely detection and management of risk factors in asymptomatic individuals could result in a substantial reduction of their cardiovascular risk 3. Early detection through screening of asymptomatic individuals for an increased risk of a future CVD event has been investigated using a number of different screening tools 4-7, but for a variety of reasons these are currently not recommended by expert panels An alternative strategy to improve early detection of patients at risk could be to use the potentially valuable prognostic information contained in routine care, diagnostic computed tomography (CT) scans of patients not otherwise known to suffer from CVD. A number of reports have shown that these subclinical markers of disease sometimes referred to as incidental findings because they are unrelated to the clinical indication for obtaining the CT scan are very prevalent 11. Reporting a routine care, diagnostic chest CT examination, clinical radiologist will very often encounter atherosclerotic changes in the coronary arteries and the thoracic aorta 12,13. Diagnostic CT images can accurately visualize calcifications of the heart valves when compared with echocardiography And it is establishing its role in the assessment of myocardial infarctions 16. However, the clinical importance of these findings is uncertain. As a consequence, there is currently a complete lack of standardized reporting of these findings by clinical radiologists. To improve the existing situation and to be able to use this information for optimized targeting of preventive therapy in the future requires an accurate, reproducible and standardized assessment of these disease markers. Although in recent years a number of prognostic studies have been performed based on the semi-quantitative scoring of a single prognostic CVD marker in CT 12,13, no semi-quantitative scoring method of the full range of subclinical CVD markers has been standardized and evaluated for inter-observer agreement. In this study we present clear-cut definitions for the semi-quantitative scoring of subclinical atherosclerotic and degenerative changes in the heart and great vessels of routine care, clinical patients undergoing diagnostic multislice CT of the chest, and to determine its inter and intra-observer agreement. This scoring system will enable a more standardized reporting of these incidental, subclinal findings in diagnostic chest CT. 41

42 Chapter 3 Methods Patient population As part of a multicenter case-cohort study investigating the prognostic value of cardiovascular disease markers, contrast-enhanced chest CT scans of 152 patients (aged >40 years) recruited from one of the participating centers (a tertiary referral hospital) were retrospectively identified. All scans were performed between July and October Patients included received routine care and were not part of a research protocol at the time of their CT. Patients with a primary pulmonary malignancy or metastatic disease from any other type of cancer were excluded (n=43). The rationale for exclusion of these patients is that, given their primary diagnosis of cancer, it is highly unlikely that an evaluation of potentially prognostic CVD markers will improve their prognosis. The remaining 109 patients in this report ranged from 41 to 88 years of age (median: 61). This study was approved by our medical ethics review board. Imaging protocol and analysis All investigations were performed with a 16-slice scanner (Brilliance-16, Philips, Cleveland, Ohio). For contrast-enhanced multislice CT of the chest the following parameters were employed: 16x0.75mm collimation, seconds rotation time, 120 kv and mas. These data sets were reconstructed with a 5mm slice thickness using 4mm increment. Ninety ml non-ionic contrast solution was applied with a flow rate of 3mL/s followed by 25 ml of saline flush (3mL/s). One board certified radiologist with five years of experience (observer 1) and a research fellow with two years of experience in reading chest CT scans (observer 2) independently interpreted the CT images on an Easyvision workstation (Philips Medical Systems, Best, The Netherlands) blinded for any clinical information. Intra-observer variability was assessed in a sample consisting of the first 25 consecutive patients of this study scored twice by the same research fellow after a one month interval between readings. A score form was used consisting of eight cardiovascular disease markers. Both readers were trained to use this score form under the supervision of a board certified chest radiologist with more than 15 years of experience using a training set of 50 patients randomly selected from months prior to July Table 1 provides a concise summary of all definitions used in this score form. Aortic wall abnormalities (Table1) were subdivided into aortic plaques thickness, aortic wall irregularity, plaque ulceration and aortic wall calcification. The thickness of an aortic plaque was scored by measuring the maximal diameter of the aortic plaque including the wall and superimposed calcifications when present, and was graded as: grade 0, absent; grade 1, 42

43 Semi-quantitative assessment of cardiovascular disease markers in multislice CT of the chest Table 1 Definitions used for grading of cardiovascular imaging abnormalities as presented in the CVC score obtained from diagnostic CT of the chest. (More extensive definitions are presented in the Methods section) Grades Imaging finding (potential range) Aortic wall abnormalities plaque thickness (0-2) absent <5mm 5mm wall irregularity (0-2) absent <50% wall circumference 50% wall circumference plaque ulceration (0-1) absent present wall calcification (0-3) absent 3 foci 4-5 foci or 1 calcification extending over 3 slices >5 foci or > 1 calcification extending over 3 slices Coronary calcifications (0-3) (score per main branch: absent 1-2 foci LM,LAD,LCX,RCA) 1 >2 foci or 1 calcification extending over 2 slices extensively calcified arteries covering multiple segments AVL 2 calcifications (0-4) absent absent 1 leaflet, spotty 1 leaflet, linear 2 leaflets, linear MVL 3 calcifications (0-3) absent absent 1 leaflet, spotty 1 leaflet, linear MVA 4 calcifications (0-1) absent absent calcifications at the margin of the MVL plane Elongation of aorta (0-1) absent present Supra-aortic artery calcifications (0-2) absent calcifications in 1 supraaortic artery calcifications in >1 supraaortic arteries LV 5 scarring (0-1) absent focal thinning of ventricular wall or subendocardial hypoattenuating rim present 1 LM, left main; LAD, left anterior descending; LCX. Left circumflex; RCA, right coronary artery; 2 AVL, aortic valve leaflet; 3 MVL, mitral valve leaflet; 4 MVA, mitral valve annulus; 5 LV, left ventricle 43

44 Chapter 3 moderate (<5mm); grade 2, severe ( 5mm) 13. The irregularity of the aortic wall was graded as: grade 0, absent (no irregularity in the aortic wall contour); grade 1, moderate irregularity involving less than 50% of wall circumference; grade 2, more than 50% of wall circumference irregularly thickened (Figure 1). Ulceration of an aortic plaque was defined as a focal dissection within the aortic wall resembling a small crater with sharp angled overhanging edges 17. Ulcerations were classified as: 0, absent and 1, present. The number and size of aortic wall calcifications were assessed and graded as: grade 0, absent; grade 1, mild ( 3 focal calcifications); grade 2, moderate (4-5 focal calcifications or one calcification extending over 3 slices); grade 3, severe (>5 focal calcifications or >1 calcification extending over 3 slices). All items were scored for the ascending and descending aorta separately and subsequently added into a composite score for degenerative changes of the whole thoracic aorta (0-16). A B Figure 1 Axial 16-slice CT images of different grades of aortic wall irregularity in the descending part of the thoracic aorta (arrows demonstrate the example area): A grade 1 (irregularity covering <50% of wall circumference); B grade 2 (>50% of wall circumference); C plaque ulceration. C Coronary calcifications (Table 1) were specified according to their location in the left main artery (LM), left anterior descending artery (LAD), left circumflex artery (LCX) and right coronary artery (RCA) and scored separately (Figure 2). In case of doubt defining the 44

45 Semi-quantitative assessment of cardiovascular disease markers in multislice CT of the chest exact junction of LM, LAD and LCX calcifications were assigned to the LAD. Calcifications were assessed using the following scale: grade 0, absent; grade 1, mild (1-2 focal (limited to 2 slices) calcifications); grade 2, moderate (>2 focal calcifications or a single calcification extending over >2 slices); grade 3, severe (fully calcified coronary arteries extending over multiple segments). The four separate scores were summed into a single composite score for the coronary artery tree (0-12) 12. A B Figure 2 Axial 16-slice CT images of different grades of coronary artery calcification in the left anterior descending (LAD) branch (arrows demonstrate the example area): A grade 1 (2 focal calcifications); B grade 2 (1 calcification >8mm in length); C grade 3 (extensive calcification covering multiple segments). C Calcifications of the heart valves (Table 1) were anatomically subdivided into calcification of the aortic valve leaflets (AVL), the mitral valve leaflets (MVL) and the mitral annulus (MVA). MVL calcifications present as linear calcifications between the area of the left atrium and ventricle and are therefore positioned horizontally to the scanning plane. MVA calcifications (0, absent; 1, present) are visible as round or curled spots at the periphery of this horizontal leaflet plane. AVL and MVL calcifications were graded as: grade 0, absent; grade 1, mild (single spotty calcification); grade 2, moderate (single linear calcification); grade 3, 45

46 Chapter 3 A B C Figure 3 Axial 16-slice CT images of different grades of aortic valve calcification (arrows): A grade 1 (spotty calcification on single leaflet); B grade 2 (linear calcification on single leaflet); C grade 3 (two leaflets involved); D grade 4 (all leaflets involved). D severe (2 leaflets involved); grade 4 (only for AVL), all three leaflets affected 14,15 (Figure 3). Elongation of the aorta (Table 1) was defined as the anterior displacement and/or kinking of the descending thoracic aorta and aortic arch as measured by the midline of the thoracic aorta crossing the midline of the vertebral body (0, absent; 1, present) between the level of the fourth thoracic vertebra and the point were the descending aorta crosses the diaphragm 17,18. Atherosclerotic plaques of the supra-aortic arteries (Table 1) were defined as spotty or circumferential calcification at or just beyond the origin of the brachiocephalic trunk, the left common carotid or the left subclavian artery. These calcifications were graded as: grade 0, absent; grade 1, moderate (only 1 branch affected); grade 2, severe (2 or more branches affected). Left ventricular scarring (Table 1) was scored as: grade 0, absent; grade 1, when focal thinning of the ventricular wall was noticed or when a subendocardial hypoattenuating rim was present in the left ventricular wall indicating a previous myocardial infarction. 46

47 Semi-quantitative assessment of cardiovascular disease markers in multislice CT of the chest Statistical analysis Kappa (κ) statistics were calculated to assess the inter- and intra-observer variability in grading the eight cardiovascular markers. Since the number of categories for some of these ordinal parameters is high, thereby increasing the chance of misclassification, we used a weighted kappa to take into account partial agreement between two readings 19. This is done by assigning a weight between 0 (complete disagreement) and 1 (complete agreement) to each pair of grade scores. The more these scores differed, the lower the weight assigned to them. For interpretation of the κ statistic we refer to Landis et al 20 : a value of is considered excellent, good, moderate, fair and <0.20 poor agreement. Furthermore, for the parameters with a high number of categories (aortic wall defects (0-16) and coronary calcifications (0-12) we can reasonably assume that the categorical scale can be transformed into a continuous scale, thereby allowing the application of a two-way intraclass correlation coefficient (ICC) which analyzes the amount of clustering between two measurements ranging from 1 (complete agreement) to 0 (complete disagreement) 21. Results Our study population of 109 patients comprised 62 men and 47 women. Median age was 61 years (range: 41-88). In Table 2 we present the frequency distributions of all 8 cardiovascular disease parameters as scored by observer 1 in routine care, diagnostic chest CT investigations. Apart from the composite scores for aortic wall abnormalities and coronary artery calcifications, grade 0 scores ( = absence of a disease marker) are represented most often for all markers. Table 3 shows the measures of inter- and intra-observer variability. Inter-observer variability (n=109) expressed in weighted κ (w-κ) values ranged from 0.54 for left ventricular scarring to 0.89 for atherosclerotic changes to the supra-aortic arteries. Inter-observer variability measured by ICC values was 0.89 for the composite score of aortic wall abnormalities and 0.90 for coronary calcifications. Elongation of the aorta showed lowest intra-observer agreement as measured by weighted κ (w-κ =0.55). Atherosclerosis of the supra-aortic arteries (w-κ =0.96) and coronary artery calcifications (w-κ =0.91; ICC=0.98) showed the highest level of intra-observer agreement. ICC for aortic wall abnormalities is

48 Chapter 3 Table 2 Prevalence (score>0) and distribution of cardiovascular imaging abnormalities in 109 patients undergoing diagnostic CT of the chest as part of daily clinical practice scored by the most experienced of two readers (Prevalence: percentage of patients with one or more abnormalities rounded off to the nearest integer, Median: median number of abnormalities per patient, Range: range of abnormalities actually scored per parameter per patient) Observer 1 Imaging finding (potential range) Prevalence (%) Median (range) Aortic wall defects (0-16) 88 3 (0-11) Coronary calcifications (0-12) 63 2 (0-12) AVL 1 calcifications (0-4) 17 0 (0-4) MVL 2 calcifications (0-3) 13 0 (0-3) MVA 3 calcifications (0-1) 9 - Elongation of aorta (0-1) 36 - Atherosclerosis of supra-aortic arteries (0-2) 40 0 (0-2) LV 4 scarring (0-1) 7-1 AVL, aortic valve leaflet; 2 MVL, mitral valve leaflet; 3 MVA, mitral valve annulus; 4 LV, left ventricle. Table 3 Inter and intra-observer agreement of multislice CT imaging findings in 109 patients undergoing diagnostic CT of the chest as part of daily clinical practice Inter-observer agreement Intra-observer agreement w-kappa 1 ICC 2 (95% CI) w-kappa ICC (95% CI) Aortic wall defects ( ) ( ) Coronary calcifications ( ) ( ) AVL 3 calcifications MVL 4 calcifications MVA 5 calcifications Elongation of aorta Atherosclerosis supra-aortic artery LV 6 scarring w-kappa, weighted kappa; 2 ICC, intraclass correlation coefficient; 3 AVL, aortic valve leaflet; 4 MVL, mitral valve leaflet; 5 MVA, mitral valve annulus; 6 LV, left ventricle. 48

49 Semi-quantitative assessment of cardiovascular disease markers in multislice CT of the chest Discussion The semi-quantitative scoring of subclinical atherosclerotic and degenerative changes of the heart and thoracic aorta incidentally detected on diagnostics multislice chest CT performed as part of daily clinical practice shows an overall good to excellent inter- and intra-observer variability (Table 3). Published data on reproducibility of cardiovascular risk indicators scored semi-quantitatively on non-gated CT images are incomplete or non-existent. In the past numerous studies using ECG-gating techniques have assessed the reproducibility of (semi)automatic scoring techniques for coronary artery calcifications Also, a recent population-based, ECG-gated CT study, investigating the presence and progression of subclinical CVD (MESA), demonstrated sufficient reproducibility (inter-observer variability %) for extracoronary calcifications in the thoracic aorta, aortic valve and mitral annulus 25. Although semi-quantitative scoring of individual CVD markers has been employed for prognostic research in various publications, none of those studies have reported data on the inter-observer agreement of their measurements. Tanne et al 13 used a visual grading of calcification thickness in the thoracic aorta to investigate the associated risk of cerebrovascular events without reporting inter- or intra-observer agreement. Similarly, the only study reporting to have used visual grading in the assessment of coronary artery calcifications (absent, mild, moderate, severe) in non-gated CT scans did not report observer variability statistics 12. On other potential CVD markers, like aortic elongation or left ventricular scarring, no prognostic studies have altogether been reported. In this study we have tried to enhance the inter-observer agreement by extensive training of the readers in uniformly scoring of CVD markers and subsequent consensus meetings discussing the doubtful cases prior to scoring the original dataset. Nevertheless, only moderate agreement could be obtained for left ventricular scarring (LVS) and MVA/MVL calcifications. Detection of LVS using multislice CT is a new and evolving indication resulting in limited experience among readers and lack of hard definitions on what exactly constitutes an old myocardial infarction. Distinguishing MVL from MVA calcifications has proven to be more difficult when not using retrospective ECG-gating techniques, accounting for the lower inter-observer agreement for those two markers. Aortic wall abnormalities as a composite marker consisted of 4 items (plaque thickness, wall irregularity, ulceration and calcification) separately scored for the descending and ascending part of the thoracic aorta. Combining separate items into a single marker increases the weighted κ and ICC statistic substantially. However, when presented separately, three of these items individually showed good to very good inter-observer agreement. In the case of aortic wall ulceration only poor agreement could be achieved (w-κ 0.38). In part this might be explained by the low prevalence of this marker (2.8%) compared to other markers. 49

50 Chapter 3 We did not find substantial difference in inter-observer agreement of these markers between the ascending and descending thoracic aorta. In scoring calcifications of the coronary arteries a common difficulty in cardiac CT studies occurs in defining the cut-off point between the left main trunk and the left circumflex and left anterior descending branches. Not surprisingly, most CAC inter-observer disagreement in our study occurred in scoring calcified plaques in the left main artery. A limitation of our study is that evidence for the actual prognostic value of a number of these markers - using a semi-quantitative scoring method is limited or has yet to be established. Therefore, working with the best available definitions of potential markers, we had to make a selection which can ultimately prove to be too extensive or too limited. It follows that these semi-quantitative scores have neither been compared with existing (semi)automatic scores nor have they been validated against a secondary gold standard. The present study is simply an evaluation of the agreement of a set of related imaging abnormalities by two observers. Although assessing the inter-observer agreement of a scoring method before its validation can be considered a limitation to our report, we argue that, before any assessment of the potential prognostic value of these markers can take place, a standardized and reproducible scoring method based on solid definitions is required. Whether or not one or more of these individual markers will be used for future prognostic purposes, using definitions form this scoring method can help clinical radiologist towards a more standardized way of reporting these incidental findings. Another limitation of our study is that we only performed our readings retrospectively on thick (5mm) axial slabs. Performing real-time readings on a CT workstation and, consequently, being able to view images in multiple planes using coronal and sagittal reformations or manually adapting the section width to view thin (1mm) sections, would certainly have improved the detection and correct grading of some of the CVD markers. However, our objective was to present a feasible score form for everyday use in clinical practice. Even in its present form this scoring method requires approximately 5 minutes to carefully complete, depending on the amount of focal atherosclerotic and/or degenerative lesions. This clearly takes too much time to be feasible in daily clinical practice. However, exclusion of those markers that turn out to have low prognostic value in multivariable analysis will shorten this reviewing time. Therefore, ultimately the time required to use this score form should be commensurate with the advantage of being able to reproducibly quantify imaging predictors of cardiovascular disease in routine care, diagnostic chest CT. A standardized and reproducible scoring method is a precondition for the further investigation of the potential prognostic value of these CVD markers. Multivariable analysis will show which imaging characteristics will be strong predictors and which ones will be not. The ultimate aim will be to establish a simple prognostic rule for clinical radiologists in order to quantify a patient s future risk of CVD events. 50

51 Semi-quantitative assessment of cardiovascular disease markers in multislice CT of the chest In conclusion, we present a semi-quantitative scoring method, using clear-cut definitions, for assessment of incidental, subclinical markers of cardiovascular disease in routine care, diagnostic chest CT scans with good to excellent inter- and intra-observer agreement. Use of these definitions in clinical practice will enable a more standardized assessment and reporting of incidental findings in diagnostic chest CT. 51

52 Chapter 3 References 1. American Heart Association. Heart Disease and Stroke Statistics Update. clinicalupdates/. Last accessed Yusuf S, Hawken S, Ounpuu S et al. Effect of potentially modifiable risk factors associated with myocardial infarction in 52 countries (the INTERHEART study): case-control study. Lancet. 2004;364: Greenland P, Gaziano JM. Clinical practice. Selecting asymptomatic patients for coronary computed tomography or electrocardiographic exercise testing. N Engl J Med. 2003;349: Balady GJ, Larson MG, Vasan RS, Leip EP, O Donnell CJ, Levy D. Usefulness of exercise testing in the prediction of coronary disease risk among asymptomatic persons as a function of the Framingham risk score. Circulation. 2004;110: De Bacquer D, De Backer G, Kornitzer M, Myny K, Doyen Z, Blackburn H. Prognostic value of ischemic electrocardiographic findings for cardiovascular mortality in men and women. J Am Coll Cardiol. 1998;32: Blumenthal RS, Becker DM, Yanek LR et al. Detecting occult coronary disease in a high-risk asymptomatic population. Circulation. 2003;107: Nallamothu BK, Saint S, Bielak LF et al. Electron-beam computed tomography in the diagnosis of coronary artery disease: a meta-analysis. Arch Intern Med. 2001;161: Greenland P, Bonow RO, Brundage BH et al. ACCF/AHA 2007 clinical expert consensus document on coronary artery calcium scoring by computed tomography in global cardiovascular risk assessment and in evaluation of patients with chest pain: a report of the American College of Cardiology Foundation Clinical Expert Consensus Task Force (ACCF/AHA Writing Committee to Update the 2000 Expert Consensus Document on Electron Beam Computed Tomography) developed in collaboration with the Society of Atherosclerosis Imaging and Prevention and the Society of Cardiovascular Computed Tomography. J Am Coll Cardiol. 2007;49: Gibbons RJ, Balady GJ, Bricker JT et al. ACC/AHA 2002 guideline update for exercise testing: summary article. A report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines (Committee to Update the 1997 Exercise Testing Guidelines). J Am Coll Cardiol. 2002;40: Us Preventive Services Task Force. Screening for coronary heart disease: recommendation statement. Ann Intern Med. 2004;140: Jacobs PC, Mali WP, Grobbee DE, van der GY. Prevalence of incidental findings in computed tomographic screening of the chest: a systematic review. J Comput Assist Tomogr. 2008;32: Shemesh J, Henschke CI, Farooqi A et al. Frequency of coronary artery calcification on low-dose computed tomography screening for lung cancer. Clin Imaging. 2006;30: Tanne D, Tenenbaum A, Shemesh J et al. Calcification of the thoracic aorta by spiral computed tomography among hypertensive patients: associations and risk of ischemic cerebrovascular events. Int J Cardiol. 2007;120: Koos R, Kuhl HP, Muhlenbruch G, Wildberger JE, Gunther RW, Mahnken AH. Prevalence and clinical importance of aortic valve calcification detected incidentally on CT scans: comparison with echocardiography. Radiology. 2006;241:

53 Semi-quantitative assessment of cardiovascular disease markers in multislice CT of the chest 15. Mahnken AH, Muhlenbruch G, Das M et al. MDCT detection of mitral valve calcification: prevalence and clinical relevance compared with echocardiography. AJR Am J Roentgenol. 2007;188: Nikolaou K, Knez A, Sagmeister S et al. Assessment of myocardial infarctions using multidetector-row computed tomography. J Comput Assist Tomogr. 2004;28: Prokop M, Galanski M. Spiral and Multislice Computed Tomography of the Body. Stuttgart, New York: Thieme; Balm R, Eikelboom BC, van Leeuwen MS, Noordzij J. Spiral CT-angiography of the aorta. Eur J Vasc Surg. 1994;8: Altman DG. Practical statistics for medical research. London: Chapman & Hall; Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics. 1977;33: Armitage P, Berry G. Statiscal methods in medical research. 3rd edn ed. Oxford: Blackwell Science; McCollough CH, Ulzheimer S, Halliburton SS, Shanneik K, White RD, Kalender WA. Coronary artery calcium: a multiinstitutional, multimanufacturer international standard for quantification at cardiac CT. Radiology. 2007;243: Sabour S, Rutten A, van der Schouw YT et al. Inter-scan reproducibility of coronary calcium measurement using Multi Detector-Row Computed Tomography (MDCT). Eur J Epidemiol. 2007;22: Kopp AF, Ohnesorge B, Becker C et al. Reproducibility and accuracy of coronary calcium measurements with multidetector row versus electron-beam CT. Radiology. 2002;225: Budoff MJ, Takasu J, Katz R et al. Reproducibility of CT measurements of aortic valve calcification, mitral annulus calcification, and aortic wall calcification in the multi-ethnic study of atherosclerosis. Acad Radiol. 2006;13:

54

55 CHAPTER 4 Use of unrequested information from routine care, diagnostic chest CT: the case of coronary and aortic calcifications

56 Chapter 4 Abstract Backgound Increase in volume of CT investigations will likely result in a sharp rise of unrequested information. Clinical relevance of these subclinical findings is unknown. This is the first followup study to investigate the prognostic relevance of two examples of unrequested information - coronary (CAC) and aortic calcifications (TAC) as contained in routine diagnostic chest CT in a clinical care population. Methods Subjects (>40 years) came from a clinical care-based cohort of patients who had a chest CT in one of eight medical centers located throughout the Netherlands. Patients with a history of cardiovascular disease (CVD) were excluded. Three trained reviewers visually graded all coronary (0-12) and aortic calcifications (0-8) using a simple semi-quantitative scoring method. Results Through linkage with the national death registry and the national registry of hospital discharge diagnoses, 240 fatal and 275 non-fatal CVD events could be ascertained. Compared with subjects with no calcium, the adjusted risk for a CVD event was 3.7 times higher (95% CI, ) among patients with severe coronary calcification (CAC score 6) and 2.7 times higher (95% CI, ) among patients with severe aortic calcification (TAC score 5). Conclusions Subclinical vascular calcifications on CT are strong predictors of incident CVD events in a routine clinical care population. This is a first step to illustrate the use of prognostic information contained in routine care, diagnostic imaging to identify patients who could benefit from primary preventive efforts. 56

57 Use of unrequested information from routine care, diagnostic chest CT Introduction Over the last decades, advances in diagnostic imaging techniques have been impressive. At the same time, this success has resulted in an increase of unrequested information during the diagnostic work-up of patients. Unrequested information can be defined as imaging abnormalities of potential clinical relevance that are unexpectedly discovered and unrelated to the clinical indication. Although unrequested information occurs across different imaging modalities and anatomic regions, computed tomography (CT) of the chest provides numerous exemplary cases of unrequested information: vascular calcifications, pulmonary emphysema, pulmonary nodules, or skeletal osteoporosis. Due to lack of follow-up, the clinical importance of unexpected abnormalities is largely unknown. Performing follow-up studies is the only way to determine which imaging abnormalities have or do not have clinical relevance. Unrequested information could contain valuable prognostic information on a variety of prevalent diseases in an ageing western population providing a new means of identifying patients that could profit from primary prevention efforts. Non-invasive assessment of coronary artery calcium (CAC) by various CT techniques has been widely used as a prognostic marker of cardiovascular disease (CVD) in large prospective screening studies 1-5. Vascular calcifications have a high prevalence in the normal population 6 and will thus be detected frequently when reading diagnostic CT scans in a clinical care population. To date, no follow-up studies have been performed to investigate whether vascular calcifications can be used as prognostic marker in a population receiving routine clinical care. In this study we examined whether subclinical coronary and aortic calcifications detected unexpectedly by routine diagnostic chest CT are associated with future CVD events. Methods and Materials Population All subjects are participants of the PROVIDI study, a multicenter retrospective cohort study among persons 40 years of age and older who had a chest CT in one of 8 participating hospitals between January 2002 and December The original cohort consisted of patients. Based on CT reports, a research physician excluded all patients (n=9077) with a poor prognosis defined as (1) diagnosis of primary lung cancer (including mesothelioma), or (2) diagnosis of distant metastatic disease from other types of cancer. This exclusion criterion was applied since it is highly unlikely that in patients with a poor prognosis detection of subclinical disease markers will alter clinical decision making. Subjects with suspected cardiovascular disease or patients known with symptomatic cardiovascular disease were excluded (n=2303). The institutional review boards of all participating centers approved this study. 57

58 Chapter 4 Ater exclusion, a total of patients were eligible for this study. All 1653 patients from one randomly chosen study center were additionally excluded to be used as an external validation sample for future research. Chest CT scans were thus available for patients. Study design We used a case-cohort design in which a representative subcohort (n=1285) is sampled from the baseline cohort 7. In the data analysis, all members of the subcohort are weighted by the inverse sampling fraction to reach valid estimates for the full cohort. This design provides an efficient way to avoid scoring CT characteristics for all subjects of the baseline cohort, but in stead only in all incident cases and a representative baseline sample. Chest CT protocols To mimic day-to-day clinical practice and to increase the generalizability of our results, we chose to include CT scans from a wide spectrum of scanner types and imaging protocols. All scans were obtained with dual, 4-, 8-, 16-, 40-, or 64-slice scanners of different vendors. The field-of-view (FOV) of all eligible chest CT protocols had to include the heart and full length of the thoracic aorta. Slice thickness varied according to the CT indication and corresponding protocol. Contrast- and non-enhanced CT scans were used. One research physician abstracted and classified the CT indication on the basis of information from CT reports. Scoring of CT characteristics CT scans were scored by three readers: one board certified radiologist with ten years of experience and two research physicians with two and three years of experience in reading chest CT scans. All readers were blinded for patient characteristics and outcome status. CT images were stored in DICOM format and read using Dicomworks, version All CT scans were analyzed in the axial plane. Readings were performed using a score form including information on type of CT protocol (section thickness, tube load (mas) and voltage (kvp), and the use of contrast agent) and scan quality (defined as good, adequate, or poor). The extent of vascular calcifications was scored using simple visual grading (0-3) as previously described 8 and summed into a single score for the coronary artery tree (0-12) and the thoracic aorta (0-8). Table 1 provides exact definitions. The lower cut-off level for the descending aorta was defined as the mid-level of the 11 th thoracic vertebra. Use of the score form was trained under the supervision of an experienced board certified chest radiologist using a training set of 50 randomly selected patients. Reproducibility of visual grading was evaluated. Briefly, weighted kappa for inter -and intra-observer variability of aortic wall abnormalities (including calcifications) were 0.72 and 0.88, for calcifications of the supra-aortic arteries 0.89 and 0.96, and for calcifications in the coronary arteries 0.77 and 0.91, respectively. 58

59 Use of unrequested information from routine care, diagnostic chest CT Follow-up We recorded incident fatal and nonfatal CVD events for a mean of 17.8 months. End-point status was obtained through linkage of patients with the National Death Registry and the National Registry of Hospital Discharge Diagnoses from January 2002 to December Database linkage was performed with a validated probabilistic method In these databases, cause of death and the occasion of hospitalization are coded according to the International Classification of Diseases, 9 th and 10 th revision (ICD-9, ICD-10) 13. Correct designation of causes of death has been established in a comparison study with patient medical records 14. Fatal and nonfatal CVD events were defined by ICD-9 codes as coronary heart disease (CHD) (codes ), heart failure (code 428), peripheral arterial disease (PAOD) (codes 440, ), aortic aneurysm (code 441), cerebrovascular disease (codes ), or non-rheumatic valvular disease (code 424). In case of multiple valid end points in the same patient, cause of death prevailed over hospital admissions or else the first hospital discharge diagnosis was used. Secondary analyses were performed for a selection of cases from the primary end point: all fatal and nonfatal coronary events (codes ) and all fatal and nonfatal non-cardiac events (PAOD, codes 440, ; aortic aneurysm, code 441; and cerebrovascular events, codes ). Table 1 Definitions used for visual grading of calcified plaques on routine diagnostic chest CT Grades Finding (potential range) Aortic wall calcification (0-3) (score for ascending and descending aorta separately) Supra-aortic artery calcification (0-2) Coronary calcification (0-3) (score per main branch: LM,LAD,LCX, RCA) absent 3 foci 4-5 foci or 1 calcification extending over 3 slices absent calcifications in 1 supra-aortic artery calcifications in >1 supra-aortic arteries absent 1-2 foci >2 foci or 1 calcification extending over 2 slices >5 foci or 1 calcification extending over 3 slice calcified arteries covering multiple segments LM denotes left main; LAD, left anterior descending; LCX, left circumflex; RCA, right coronary artery. Data analysis We calculated annualized event rates for all three end points stratified according to coronary and aortic calcium scores. We used Cox proportional-hazards regression to estimate hazard ratios for any CVD event, coronary events, and any non-cardiac event according to calcium scores. Sum scores for visually graded coronary and aortic calcifications were analyzed as categorical and as continuous variables (per SD). 59

60 Chapter 4 After estimating the hazard ratios for all non-cardiac events according to the aortic sum score, we performed additional regression analyses for all cases of stroke, looking at calcifications in each anatomically distinct part of the thoracic aorta (ascending, descending, arch). Apart from crude associations, all models were adjusted for age, sex, indication of CT scan, image quality, and type of medical center in which the CT scan was performed (tertiary/ secondary). We also tested the interaction of aortic and coronary sum scores with section thickness (continuous) and the use of contrast agent (yes/no) by entering interaction terms in the model. We had missing values for CT characteristics in <3%. We used regression methods implemented in SPSS software (SPSS 14.0, Chicago, Illinois) to impute missing values. All other analyses were performed with R software, version 6.2. Results A total of 121 subjects (44 of 559 cases (7.9%), and 77 of 1285 subjects from the subcohort (6.0%)) were excluded from analyses because CT scans could not be retrieved from CT databases. Table 2 summarizes the baseline characteristics of the subcohort and the various case-groups. Subjects with coronary and non-cardiac events were older than people in the subcohort, but only subjects with coronary events were more often male. A total of 515 subjects had a CVD event during follow-up: fatal CVD events occurred in 240 patients, and 275 patients had a nonfatal event. Table 3 shows a gradual increase in annualized event rate for any CVD event with increasing CAC and TAC score categories. The prevalence of CAC (sum score >0) was 67% for people in the subcohort compared with 88% in subjects with CVD events, 90% for coronary events, and 85% for non-cardiac events. The corresponding figures for TAC > 0 were 61% (subcohort), 82%, 80%, and 86%, respectively. Table 4 shows that the risk of any CVD event increases with an increase in category of TAC and CAC scores after adjustment for age, sex, CT indication, scan quality, and type of medical center. A statistically significant increase in hazard ratio occurs already in case of mild calcification (TAC and CAC scores = 1-2). The risk of any CVD event was increased by a factor of 2.7 (95% CI, ) and 3.7(95% CI, ), respectively, among patients with severe TAC or CAC (TAC score 5; CAC score 6) as compared with those without any calcium. In order to compare the strength of the associations between the two calcium measures, Table 4 also shows the increase in risk of any CVD event associated with CAC and TAC as continuous measures (per 1 SD increase in the TAC and CAC sum scores). After adjustment, an increase of 1 SD in TAC and CAC sum scores resulted in a 46% and 41% increase in the risk for any CVD event. 60

61 Use of unrequested information from routine care, diagnostic chest CT Table 2 Baseline characteristics Variable Subcohort Any CVD Event Coronary Event Non-cardiac Event (N=1208) (N=515) (N=310) (N=128) Age (yr) 61.5± ± ± ±11.3 Male sex (%) CT indication (%) Pulmonary disease Haematological malignancy Mediastinal disease Ruled-out pulmonary malignancy Pulmonary embolism Other Image quality (%) Good Adequate Poor Tertiary medical center (%) Use of contrast agent (%) Section thickness (mm) 1-3 mm mm >6 mm Thoracic aortic calcium (sum score) None (0) Mild (1-2) Moderate (3-4) Severe ( 5) Coronary artery calcium (sum score) None (0) Mild (1-2) Moderate (3-5) Severe ( 6) Values are means±sd or proportions. Table 3 Annualized event rates for any CVD event (n=515) according to coronary artery calcium score and thoracic aorta calcium score categories TAC risk category Annualized event rates CAC risk category None (0) 1.0 (95) None (0) 0.7 (62) Mild (1-2) 1.9 (127) Mild (1-2) 2.2 (111) Moderate (3-4) 5.0 (170) Moderate (3-5) 3.1 (150) Severe ( 5) 5.6 (123) Severe ( 6) 5.9 (192) CAC denotes coronary artery calcium score; CVD, cardiovascular disease; TAC, thoracic aorta calcium score. Data are percentages. Data in parentheses are numbers of patients. Median follow-up per case-group: 17.8 months for total CVD events. 61

62 Chapter 4 Table 4 Risk of any CVD event, coronary events, and non-cardiac events associated with increasing coronary artery calcium score and thoracic aorta calcium score in clinical care patients undergoing routine diagnostic chest CT End points Hazard ratios (95% CI) Crude Full Model Crude Full Model Any CVD Event (n=515) TAC categories CAC categories None (0) None (0) Mild (1-2) 1.8 ( ) 1.4 ( ) Mild (1-2) 2.8 ( ) 2.2 ( ) Moderate (3-4) 4.4 ( ) 2.6 ( ) Moderate (3-5) 3.8 ( ) 2.5 ( ) Severe ( 5) 4.9 ( ) 2.7 ( ) Severe ( 6) 6.9 ( ) 3.7 ( ) Continuous 1.76 ( ) 1.46 ( ) Continuous 1.73 ( ) 1.41 ( ) Coronary Events (n=310) TAC categories CAC categories None (0) None (0) Mild (1-2) 1.5 ( ) 1.2 ( ) Mild (1-2) 3.0 ( ) 2.4 ( ) Moderate (3-4) 4.0 ( ) 2.3 ( ) Moderate (3-5) 4.3 ( ) 2.9 ( ) Severe ( 5) 4.6 ( ) 2.4 ( ) Severe ( 6) 9.1 ( ) 5.0 ( ) Continuous 1.76 ( ) 1.43 ( ) Continuous 1.87 ( ) 1.54 ( ) Non-cardiac Events (n=128) TAC categories CAC categories None (0) None (0) Mild (1-2) 2.6 ( ) 1.9 ( ) Mild (1-2) 2.5 ( ) 1.8 ( ) Moderate (3-4) 5.6 ( ) 3.5 ( ) Moderate (3-5) 2.7 ( ) 1.7 ( ) Severe ( 5) 5.5 ( ) 3.3 ( ) Severe ( 6) 4.5 ( ) 2.3 ( ) Continuous 1.71 ( ) 1.45 ( ) Continuous 1.48 ( ) 1.20 ( ) CAC denotes coronary artery calcium score; CVD, cardiovascular disease; CI, confidence interval; TAC, thoracic aorta calcium. Full Model: adjusted for age, sex, clinical indication for chest CT, image quality, and type of medical center. Hazard ratios for continuous TAC/CAC scores are calculated per 1 standard deviation increase in CAC (SD=3.022) or TAC (SD=1.952). Non-cardiac events were fatal and nonfatal cases of stroke, aortic aneurysm/dissection and peripheral arterial occlusive disease. Looking at coronary and non-cardiac events separately (Table 4, lower half), we could demonstrate a more differentiated pattern in the associations with TAC and CAC scores. Risk for coronary events increased 5-fold among subjects with severe CAC compared with those without any coronary calcium, whereas having severe TAC resulted in an increase in risk by a factor of 2.4. Conversely, risk of non-cardiac events (stroke, aortic aneurysm/dissection, PAOD) increased by a factor of 3.3 for subjects with severe TAC compared with patients without any aortic calcium, whereas people with severe CAC showed an increase of risk by a 62

63 Use of unrequested information from routine care, diagnostic chest CT factor of 2.3. No statistically significant interaction was found between section thickness and TAC (P=0.46) or CAC (P=0.89) and the use of intravenous contrast and TAC (P=0.40) or CAC (P=0.81) in relation to risk of CVD events. Table 5 shows the risk of stroke associated with increasing amounts of calcium in each of the three anatomically distinct parts of the thoracic aorta. Strongest associations were demonstrated for calcifications in the ascending aorta. Moderate calcification of the ascending aorta increased the risk of stroke by a factor 3.0 compared with no calcification, whereas moderate calcification of the descending aorta or supra-aortic branches increased the risk by a factor of 1.8 and 1.4, respectively. Table 5 Risk of stroke (n=58) associated with calcium in three anatomically distinct parts of the thoracic aorta Site Hazard ratios (95% CI) Crude Full Model Ascending aorta (0-3) None (0) Mild (1) 3.2 ( ) 2.2 ( ) Moderate (2) 4.7 ( ) 3.0 ( ) Severe (3) Inf Inf Descending aorta (0-3) None (0) Mild (1) 3.0 ( ) 1.8 ( ) Moderate (2) 3.9 ( ) 1.8 ( ) Severe (3) 5.9 ( ) 3.0 ( ) Supra-aortic branches (0-2) None (0) Moderate (1) 2.4 ( ) 1.4 ( ) Severe (2) 3.1 ( ) 1.7 ( ) CI denotes confidence interval. Full Model: adjusted for age, sex, clinical indication for chest CT, image quality, and type of medical center. Due to the limited number of stroke events, no cases occurred in the highest TAC category for the ascending aorta and consequently no hazard ratio could be calculated. Discussion We examined whether unrequested information from routine diagnostic chest CT in a clinical care population could be used as predictor of future disease. Severe coronary (CAC) or aortic (TAC) calcification based on visual grading of calcified plaques - increased the risk of any CVD event almost 4-fold and threefold, respectively, after a mean follow-up of close to one and a half year. CAC was found to be a stronger predictor of coronary events, whereas TAC was stronger associated with non-cardiac events. 63

64 Chapter 4 More than 60 million diagnostic CT scans are performed annually in the US and this number will likely increase as the population continues to age 15. These diagnostic images contain unrequested, subclinical findings that could contain important prognostic information. Obtaining this information is free, as it is contained in routine care and comes at no additional exposure to ionizing radiation. Currently, no data are available on the natural course and prognosis of these findings. Results from this study suggest that CT images obtained as part of routine clinical care do indeed convey important ancillary information. We investigated the prognostic value of coronary and aortic calcifications since these are highly prevalent 6 and long recognized as prognostic markers of CVD in asymptomatic screening populations 1-5. The prevalence of any CAC in our study (67%) is comparable with these previous studies (53%-69%) 2,16,17. Like these screening studies we show that CAC is a strong predictor of future CVD events. However, our results are the first to demonstrate this association in a routine clinical care population. Furthermore, we used a simple visual grading to score calcifications on a heterogeneous set of CT protocols. Existing calcium scoring typically involves non-contrast enhanced, ECG-gated CT protocols with semi-automatic calcium scoring software. The scoring method we used consists of simple definitions, is reproducible, can be performed with a minimum of additional reviewing time and can be applied to a whole range of different CT protocols (including non-gated CT). Prognostic information on CVD risk can be appended to the CT report. Together with all relevant clinical data available from a patient s medical record, the clinician can decide whether additional work-up is required. This will result in a more consistent and clinically useful way of reporting unrequested subclinical findings. CVD is one of the leading causes of death and hospitalizations in western society 18. Major advances in knowledge about CVD risk factors have been made in recent decades and international guidelines for the primary prevention of CVD have been actively developed However, for a variety of reasons substantial groups of people at intermediate or high risk of CVD events according to existing guidelines are currently not recognized and treated as such 22,23. This has prompted expert panels to state that the search for new strategies to detect patients who would benefit most from intensive primary prevention efforts is a clinically important objective 24,25. The present study is an attempt at defining such a new strategy. Limitations Unfortunately, no information from patient s medical records was available for this study. This might have influenced our results in two ways. First, exclusion of patients with symptomatic CVD could now only be performed based on information from CT reports, possibly resulting in a number of symptomatic CVD patients being missed. This would have caused an overestimation of our risk ratios. However, we excluded approximately 16% of patients from the baseline clinical care cohort based on information form CT reports (2303/14366) which 64

65 Use of unrequested information from routine care, diagnostic chest CT is even slightly more than the prevalence of symptomatic CVD (11.2%) in the general US population of whites 26. Second, lack of information from medical records prevented us from adjusting our risk estimates for traditional CVD risk factors (hypertension, hyperlipidaemia, diabetes, smoking). Based on previous reports we expect that only limited attenuation of our age and sex-adjusted risk estimates will occur when adjusting additionally for other CVD risk factors 3. Conclusion Subclinical coronary and aortic calcifications detected on routine care, diagnostic chest CT in a clinical care population can be used as a prognostic marker of future CVD events. This offers a novel approach in trying to employ potentially valuable prognostic information contained in routine diagnostic imaging and maybe used to identify patients who could benefit from primary preventive efforts. 65

66 Chapter 4 References 1. Arad Y, Goodman KJ, Roth M, Newstein D, Guerci AD. Coronary calcification, coronary disease risk factors, C-reactive protein, and atherosclerotic cardiovascular disease events: the St. Francis Heart Study. J Am Coll Cardiol. 2005;46: Kondos GT, Hoff JA, Sevrukov A et al. Electron-beam tomography coronary artery calcium and cardiac events: a 37-month follow-up of 5635 initially asymptomatic low- to intermediate-risk adults. Circulation. 2003;107: LaMonte MJ, FitzGerald SJ, Church TS et al. Coronary artery calcium score and coronary heart disease events in a large cohort of asymptomatic men and women. Am J Epidemiol. 2005;162: O Malley PG, Taylor AJ, Jackson JL, Doherty TM, Detrano RC. Prognostic value of coronary electron-beam computed tomography for coronary heart disease events in asymptomatic populations. Am J Cardiol. 2000;85: Shaw LJ, Raggi P, Schisterman E, Berman DS, Callister TQ. Prognostic value of cardiac risk factors and coronary artery calcium screening for all-cause mortality. Radiology. 2003;228: Odink AE, van der Lugt A, Hofman A et al. Association between calcification in the coronary arteries, aortic arch and carotid arteries: the Rotterdam study. Atherosclerosis. 2007;193: Prentice RL. A case-cohort design for epidemiologic cohort studies and disease prevention trials. Biometrika. 1986;73: Shemesh J, Henschke CI, Farooqi A et al. Frequency of coronary artery calcification on low-dose computed tomography screening for lung cancer. Clin Imaging. 2006;30: De Bruin A, Kardaun JW, Gast A, Bruin E, van Sijl M, Verweij G. Record linkage of hospital discharge register with population register: experiences at Statistics Netherlands. Stat J UN Econ Comm Eur. 2004;21: Herings RM, Bakker A, Stricker BH, Nap G. Pharmaco-morbidity linkage: a feasibility study comparing morbidity in two pharmacy based exposure cohorts. J Epidemiol Community Health. 1992;46: Paas GR and Veenhuizen KC. Research on the validity of the LMR. Utrecht; Prismant: Reitsma JB, Kardaun JW, Gevers E, de Bruin A, van der Wal J, Bonsel GJ. [Possibilities for anonymous follow-up studies of patients in Dutch national medical registrations using the Municipal Population Register: a pilot study]. Ned Tijdschr Geneeskd. 2003;147: Last accessed Mackenbach JP, Van Duyne WM, Kelson MC. Certification and coding of two underlying causes of death in The Netherlands and other countries of the European Community. J Epidemiol Community Health. 1987;41: Brenner DJ, Hall EJ. Computed tomography--an increasing source of radiation exposure. N Engl J Med. 2007;357: Budoff MJ, Shaw LJ, Liu ST et al. Long-term prognosis associated with coronary calcification: observations from a registry of 25,253 patients. J Am Coll Cardiol. 2007;49: Wong N, Gransar H, Shaw LJ et al. Thoracic Aortic Calcium Versus Coronary Artery Calcium for the Prediction of Coronary Heart Disease and Cardiovascular Disease Events. J Am Coll Cardiol Img. 2009;2: Murray CJ, Lopez AD. Global mortality, disability, and the contribution of risk factors: Global Burden of Disease Study. Lancet. 1997;349:

67 Use of unrequested information from routine care, diagnostic chest CT 19. Assmann G, Cullen P, Schulte H. Simple scoring scheme for calculating the risk of acute coronary events based on the 10-year follow-up of the prospective cardiovascular Munster (PROCAM) study. Circulation. 2002;105: Conroy RM, Pyorala K, Fitzgerald AP et al. Estimation of ten-year risk of fatal cardiovascular disease in Europe: the SCORE project. Eur Heart J. 2003;24: Wilson PW, D Agostino RB, Levy D, Belanger AM, Silbershatz H, Kannel WB. Prediction of coronary heart disease using risk factor categories. Circulation. 1998;97: Greenland P, LaBree L, Azen SP, Doherty TM, Detrano RC. Coronary artery calcium score combined with Framingham score for risk prediction in asymptomatic individuals. JAMA. 2004;291: Ridker PM, Paynter NP, Rifai N, Gaziano JM, Cook NR. C-reactive protein and parental history improve global cardiovascular risk prediction: the Reynolds Risk Score for men. Circulation. 2008;118: , 4p. 24. Grundy SM, Cleeman JI, Merz CN et al. Implications of recent clinical trials for the National Cholesterol Education Program Adult Treatment Panel III guidelines. Circulation. 2004;110: Naghavi M, Libby P, Falk E et al. From vulnerable plaque to vulnerable patient: a call for new definitions and risk assessment strategies: Part II. Circulation. 2003;108: Lloyd-Jones D, Adams R, Carnethon M et al. Heart disease and stroke statistics update: a report from the American Heart Association Statistics Committee and Stroke Statistics Subcommittee. Circulation. 2009;119:

68

69 Part 2 Calcium scoring as part of non-gated, low-dose CT screening for lung cancer in asymptomatic heavy smokers Chapter 5 Coronary artery calcium scoring in low-dose ungated computed tomography screening for lung cancer: interscan agreement. P.C.A. Jacobs, I. Isgum, M.J.A. Gondrie, W.P.Th.M. Mali, B. van Ginneken, M. Prokop, Y. van der Graaf. Accepted for AJR. Chapter 6 Coronary artery calcium can predict all-cause mortality and cardiovascular events on low-dose computed tomography screening for lung cancer. P.C.A. Jacobs, M.J.A. Gondrie, Y. van der Graaf, K.J.M. Janssen, H.J. de Koning, I.Isgum, B. van Ginneken, M. Prokop, R.J. van Klaveren, M. Oudkerk, W.P.Th.M. Mali. Submitted. Chapter 7 Comparing coronary artery calcium and thoracic aortic calcium for prediction of all-cause mortality and cardiovascular events on low-dose non-gated computed tomography. P.C.A. Jacobs, M. Prokop, Y. van der Graaf, M.J.A. Gondrie, K.J.M. Janssen, H.J. de Koning, I. Isgum, R.J. van Klaveren, M. Oudkerk, B. van Ginneken, W.P.Th.M. Mali. Accepted for Atherosclerosis. Chapter 8 Coronary artery calcium screening as part of low-dose computed tomography screening for lung cancer improves cardiovascular risk prediction in men. P.C.A. Jacobs, M.J.A. Gondrie, H.J. de Koning, K.J.M. Janssen, S. J. Otto, R.J. van Klaveren, W.P.Th.M. Mali, M. Oudkerk, M. Prokop, Y. van der Graaf. Submitted. Chapter 9 Smoking cessation and risk of cardiovascular disease among participants of a lung cancer screening trial P.C.A. Jacobs, Y. van der Graaf, M.J.A. Gondrie, M. Oudkerk, H. J. de Koning, R.J. van Klaveren, W.P.Th.M. Mali. Submitted.

70

71 CHAPTER 5 Coronary artery calcium scoring in low-dose ungated computed tomography screening for lung cancer: interscan agreement

72 Chapter 5 Abstract Objective Previous report have compared the detection of coronary artery calcium (CAC) with lowdose ungated MDCT performed for lung cancer screening with dedicated cardiac CT. We evaluated the interscan agreement of CAC scores on two consecutive low-dose ungated MDCT scans. Materials and Methods Subjects were 584 participants from the screen arm of a lung cancer screening trial that underwent two low-dose ungated MDCT within 4 months (mean 3.1 ± 0.6) from their baseline CT. Agatston score, volume score, and calcium mass score were measured by two observers. Interscan agreement for the stratification of participants into four Agatston score risk categories (0, 1-100, , >400) was assessed with κappa values. Interscan variability and 95% repeatability limits were calculated for all three calcium measures and compared with repeated measures ANOVA. Results An Agatston score > 0 was detected in 443 (76%) baseline CT scans. Interscan agreement of the 4 risk categories was good (κ = 0.67). Seventy-five percent (n=440) of participants had Agatston scores on scan 1 and scan 2 in the same risk category; 99% (n=578) had scores differing by a maximum of one category. Furthermore, mean interscan variability ranged form 61% for calcium volume score to 71% for Agatston score (P<0.01). A limitation of this study is that no comparison of CAC scores between the low-dose ungated CT and the gold standard ECG-gated CT was performed. Conclusion CVD risk stratification with low-dose ungated MDCT is feasible with good interscan agreement for the stratification of participants into Agatston score risk categories. High mean interscan variability precludes the use of this technique for monitoring CAC scores in individual patient settings. 72

73 Calcium scoring as part of non-gated, low-dose CT screening for lung cancer in asymptomatic heavy smokers Introduction Since prolonged and heavy smoking is a strong etiologic factor in the development of lung cancer and cardiovascular disease (CVD), simultaneous screening for both diseases in a high risk population of heavy (ex)smokers is an attractive strategy for maximizing the beneficial effects on survival in lung cancer screening programmes and minimizing the radiation dose for individuals in the screening population. Presence and extent of coronary artery calcification (CAC) detected on electron-beam CT (EBCT) or ECG-gated multidetector-row CT (MDCT) have been shown to independently predict CVD events 1 ; by contrast, screening for lung cancer is performed with low-dose ungated MDCT scanning protocols. Although the clinical importance of CAC detected with both ungated MDCT 2 and low-dose CT 3,4 has been established in the past, factors negatively influencing variability in dedicated calcium-scoring CT scans - such as image noise 5, intra- and interobserver variability 6, motion artefacts 7 or partial volume effects 8 - have always been deemed to have too big of an impact in ungated and low dose MDCT scans to consider it a reasonable alternative. However, two recent studies comparing calcium measurements in low-dose ungated MDCT with results form ECG-gated MDCT concluded that comparable results could be obtained 9,10. To be truly useful as a screening tool for cardiovascular disease, low-dose ungated MDCT scanning should also show sufficient interscan agreement. Even in dedicated calcium scoring, substantial interscan variability between repeat scans of the same person exist 11. However, if calcium scoring in low-dose ungated MDCT is used as a marker to predict CVD events, not the absolute or relative interscan variability of calcium scores should be our main concern, but the correct stratification of patients into CVD risk categories, such as the system based on the Agatston score described by Rumberger et al 12. This review suggested clinically applicable cut-offs in the continuous Agatston score (0 = low risk, = mild risk, = moderate risk, and >400 = high risk) corresponding to increasing levels of risk for coronary events. In analogy to the Framingham risk score, patients in the highest risk category (>400) should receive aggressive preventive treatment of CVD risk factors and this should possibly be extended to patients in the intermediate risk category ( ) as well. Therefore, the main purpose of this study was to determine the interscan agreement of Agatston score risk categories in repeat low-dose ungated MDCT to evaluate the usefulness of this technique for the prediction of CVD events in participants from a lung cancer screening trial. Materials & Methods Approval was obtained by the institutional review committees of all participating study sites. Informed consent was obtained from all participants. 73

74 Chapter 5 Participants Using population registries, subjects between 50 and 75 years of age were recruited to participate in a lung cancer screening trial. Inclusion criteria were: (a) participants had to be current or former smokers who quit <10 years ago, with a smoking history of >15 cigarettes/day for >25years or >10cigarettes/day for >30 years; (b) they should be able to climb 2 flights of stairs. A detailed description of patient selection for the study has been published 13. Between April 2004 and July 2006, a baseline CT scan was performed in participants randomized to the screen group in four participating medical centers. For the present study, a subgroup of 584 participants (497 men, mean age 59.9±5.8 years) from one of the 4 participating centers was identified who had a repeat scan within 4 months from their baseline CT scan. Indication for a repeat CT scan was detection of a pulmonary nodule mm 3 at baseline CT 14. The mean period between baseline and repeat scan was 3.1 months ± 0.6 (SD); range months. Since we can only expect limited progression of CAC in this short time interval 15, this offers a good opportunity to investigate the interscan variability of CAC measurements. Low-Dose chest CT protocol All baseline and repeat CT scans performed at the participating study site were conducted with a 16-slice MDCT scanner (Mx8000 IDT; Philips Medical Systems, Cleveland, Ohio). Scanning was performed in a spiral mode with 16x0.75mm collimation and pitch The scanning parameters have been described more extensively elsewhere 14. Transverse images with 1.0mm section thickness and 0.7mm increment were acquired from the level of the lung bases to the lung apices. The smallest field of view was chosen to include the outer rib margins at the widest dimension of the chest. Participants were scanned at suspended maximal inspiration after instructions about breath-holding. Total scan duration was within a single breath-hold (approximately 10 seconds). No electrocardiographic triggering was performed; no contrast agent was administered. The raw data were reconstructed into 3.1mm overlapping sections (with 1.4mm increment) for the analysis of coronary artery calcium. Low-dose exposure settings were applied based on body weight: 30 mas at a tube voltage of 120 kvp for subjects 80kg, and 30 mas at a tube voltage of 140 kvp for subjects >80kg. Radiation exposure (CTDIvol) was 2.2 and 3.5 mgy, accordingly. CAC assessment All coronary calcium scoring was performed with software described in Isgum et al 16. Two observers with two and three years of experience in cardiac CT independently read all 1168 CT scans blinded to age, sex and patient name. A pair of scans from a single participant 74

75 Calcium scoring as part of non-gated, low-dose CT screening for lung cancer in asymptomatic heavy smokers was read with a time interval of at least 2 weeks by one observer to eliminate the effect of interobserver variability. Scan pairs were divided equally among the two readers. Prior to the start of the study, both readers completed a training set of 50 randomly selected CT scans from the NELSON study database to assess interobserver variability (Intraclass correlation R = 0.97). Calcium scoring software written in C++ was used to define calcified plaque as all regions of 3 adjoining voxels (0.7 mm 3 ) with attenuation above 130 H. An investigator manually identified a point in each calcified lesion. Subsequently, three-dimensional component labeling using 26-connectivity was automatically performed to mark all connected voxels as calcification. Care was taken not to include noncoronary calcifications (eg, valve calcifications) or hyperattenuating foci due to image noise 16,17. Total Agatston score (AS), calcium volume score (VS) and calciumhydroxyapatite mass score (MS) were outputted by the software program as outlined by Ulzheimer 18. Separate scores were calculated for the left main (LM), left anterior descending (LAD), left circumflex (LCX), and right (RCA) coronary arteries. Agatston scores were further categorized into four groups (0, 1-100, , and >400) to be used for CVD risk stratification as outlined by Rumberger 12. Statistical analysis Intraclass correlation coefficients (ICC), kappa values (κ), and variabilities were used to express interobserver and interscan agreement of CAC scores. First, the presence of coronary calcium (Agatston score > 0) on scan 1 versus the presence of calcium on scan 2 was assessed for all participants (n=584). Agreement for the presence (yes/no) of calcium was calculated using κ statistics. All participants were additionally stratified into a CVD risk category for scan 1 and for scan 2 based on the Agatston score (0 (low risk), (moderate risk), (intermediate risk), and >400 (high risk)) 12. Agreement in risk category stratification was also performed using κ statistics. Relative interscan variability of two CAC scores was calculated as: absolute [score 1- score 2]/mean (score 1 + score 2) x 100%. In participants with an Agatston score > 0 in at least one scan (n=461), we used the regression method for nonuniform differences to establish 95% repeatability limits for all three measures 19. In this method, the absolute interscan difference (D) of each calcium measure is linearly modelled against the mean calcium measures (M). All models were adjusted for the total amount of calcium, as measured according to the mean (natural log transformed) calcium score. We chose to force these models through the origin since the interscan difference is zero when the mean calcium measure is zero 20. The resulting regression line is: D = b 1 M [equation 1], where b 1 is the slope of the line. We then modelled the absolute values of the unstandardized residuals ( R ) from the previous regression models against the mean calcium measures (M): R = c 0 + c 1 M [equation 2], where c 0 is the intercept and c 1 is the slope of the regression line. The 95% repeatability interval can be calculated by combining these two regression equations: 75

76 Chapter 5 b 1 M ± 2.46(c 0 + c 1 M) [equation 3]. By assuming that the systematic difference between the two scans (=b 1 M) equals zero, this equation can be written as: 2.46(c 0 + c 1 M) [equation 4]. In this method, we have substituted 1.96 with 2.46 since it is assumed that the absolute value of the residuals follow a half normal distribution. Therefore it is necessary to multiply 1.96 with π/2 19. Absolute interscan difference was plotted against mean calcium score using the Bland and Altman approach. With equation 4, we can calculate the upper and lower 95% repeatability limits for the absolute interscan difference of a mean calcium score value. A repeated measures analysis of variance was used to assess differences in relative interscan difference between the three different calcium measures. Bonferroni tests were used for post hoc comparison. Results A total of 584 participants (497 men, mean age 59.9±5.8 years) were included in this study. Figure 1 shows the frequency distribution of CAC on the baseline scan of all 584 participants subdivided into Agatston score-based risk categories. A total of 443 participants (76%) had any CAC (Agatston score >0) detected on at the baseline CT scan. Median Agatston score at baseline and repeat CT were 80.3 (range: ) and 78.1 (range: ), respectively. Agreement between absolute Agatston scores in two repeat low-dose ungated CT scans showed a intraclass R = Figure 1 Frequency histogram shows distribution of Agatston scores form baseline CT scans in 584 subjects. Y-axis is percentage of subjects per category of Agatston score. 76

77 Calcium scoring as part of non-gated, low-dose CT screening for lung cancer in asymptomatic heavy smokers Overall agreement between scans for the presence (yes/no) of any CAC was 92% [535 of 584 participants]; only 49 pairs of scans (8%) were discordant, resulting in a good kappa value of Figure 2 shows a bar chart of the distribution of absolute Agatston scores in those 49 participants with CAC in only one scan (discordant pairs). 71.4% of these participants (n=35) had an Agatston score (AS) of less than 10 which is neglectable and is likely caused by minor motion artefacts. Only 2 participants (4.1%) showed Agatston scores that were substantially out of range (maximum AS: 278.7) caused by major motion artefacts. Figure 2 Bar charts shows distribution of Agatston score among participants (n=49) with Coronary Artery Calcium (CAC) in only one of two scans (discordant pairs). Table 1 shows the interscan agreement of Agatston scores of two repeat low-dose CT scans in the same patient for the four categories used in CVD risk stratification. In approximately 3 out of 4 participants (440 [75.3%] of 584) no shift in Agatston score risk category occurred between scan 1 and scan 2. The unweighted kappa statistic showed good agreement (κ = 0.67) 21. Of those participants in which a shift in risk category did occur, only in 8 participants (1.4%) a shift of more than one risk category was observed. Table 1 Risk Stratification of Subjects by Agatston Score Risk Categories from Baseline and Repeat Low-Dose Ungated MDCT Baseline Low-Dose Ungated MDCT Repeat Low-Dose Ungated MDCT >400 Total > Total

78 Chapter 5 Figure 3 shows standard Bland-Altman plots where the means of calcium scores from scan 1 and scan 2 are plotted against the adjusted absolute interscan difference. As to be expected, all three plots show a nonuniform relation between the extent of calcium and the measurement error, ie the interscan variability increases as the total amount of calcium increases. Standard 95% confidence intervals (calculated as the mean difference ± 1.96 x SD) do not correspond well with this type of relation. Therefore, 95% confidence limits were calculated using nonuniform regression analysis and are represented in the plots by dashed lines. The slopes of these lines (r) can be interpreted as a measure of reproducibility. Figure 3 Bland-Altman plots showing absolute interscan difference plotted against the means of three calcium measures with CAC-adjusted 95% repeatability limits. A, Agatston score. B, calcium volume score. C, calcium mass score. 78

79 Calcium scoring as part of non-gated, low-dose CT screening for lung cancer in asymptomatic heavy smokers The steeper the slope of this line (higher value for r), the less reproducible is the calcium measure. The vertical distance between the dashed lines indicates the measurement error (95% CI) for a given mean calcium score value, ie 95% of the time the absolute inter-scan difference for a given calcium score will fall within these limits. In this study, the calcium volume score (r = ) showed best reproducibility compared with Agatston score (r = ) and calcium mass score (r = ). Table 2 compares the relative interscan variability for all three calcium measures in subjects that showed any CAC on at least one of two CT scans (n=461). For the Agatston score: intraclass R = 0.94, mean variability = 71%, and median variability = 51%; for the calcium volume score: intraclass R = 0.95; mean variability = 61%, and median variability = 36%; and for the calcium mass score: intraclass R = 0.94; mean variability = 65%, and median variability = 43%. A statistically significant difference (P <.01) was found between pairwise comparison of the means of variability of all three calcium measures. Table 2 Interscan variability of Coronary Artery Calcium (CAC) In participants with an Agatston Score >0 in At Least One Scan Scored with Three Different Calcium Measures on Low-Dose Ungated MDCT (n = 461) Value Agatston score Volume score Mass score Intraclass R Mean variability (%) Median variability (%) Note Post hoc comparisons after a repeated measures ANOVA showed statistically significant (P <.01) differences in mean variability between all three calcium measures. Discussion We report that using a low-dose ungated MDCT technique for detection and quantification of coronary artery calcium in the setting of a lung cancer screening trial gives a good interscan agreement in the assignment of participants into Agatston score categories for CVD risk stratification. These results support the idea that CAC scoring as part of low-dose ungated MDCT can be a useful tool to assess the CVD risk of lung cancer screening participants. Our results provide incremental evidence to the conclusion from two previous studies 9,10 reporting on this issue. Both studies compared the accuracy of detecting and categorizing CAC between a low-dose ungated MDCT protocol and ECG-gated CT (the gold standard for CAC scoring). Using a 40-MDCT unit Kim et al reported a concordance 83% (n=106) in stratifying participants into the same risk category on low-dose CT compared with ECG-gated CT and a maximum difference of only one category in the remaining 12 participants. Wu et al., using a 16-MDCT unit, even found a concordance of 93% (n=450) when comparing with the gold standard. 79

80 Chapter 5 Apart from high agreement with ECG-gated MDCT, it is important to reliably demonstrate high agreement between CAC scores in repeat CT scans of the same patient. Although mean variability between scores on both CT examinations was as high as 60-70%, we found that 75% of participants were categorized into the same risk category on two subsequent CT lowdose ungated MDCT scans (κ= 0.67). Taken together, these results support the idea that CAC scoring of absolute calcium scores in low-dose ungated MDCT may not be very accurate, but can be reliably used for CVD risk stratification. An important difference between these three studies is the racial make-up of the study populations. The two previous studies 9,10 were both conducted among self-referred, Asian populations resulting in a low prevalence of CAC compared with our study conducted among a predominantly white population. Since variability in scores (whether intertechnique or intratechnique) is strongly related to the total amount of calcium that can be scored in an individual, it is an important strength of this study that comparable results were demonstrated in a population with a far higher prevalence of CAC. An earlier study by Shemesh et al 22 determined CAC scores in participants of a lung cancer screening study using visual (semi-quantitative, grade 0-12) grading of CAC. The authors concluded that - using this tool - CAC scoring could be used for CVD risk stratification. Their choice to perform visual grading might stem from the fact that their study was performed with a 4-MDCT unit, resulting in less spatial and temporal resolution compared with a 16-MDCT unit. Semi-automated scoring has the advantage over visual grading of a better inter and intra-observer reproducibility. The subjective assessment (size, attenuation) of a vascular calcification is replaced by a software tool, identifying calcifications based on a clear attenuation threshold (130 H) and automatically calculating calcium scores. Apart from interscan agreement of Agatston risk categories, we investigated the ability of low-dose ungated MDCT to detect the presence (yes/no) of CAC between two repeat examinations. Previous reports have shown that the absence of any CAC, though not completely excluding the risk of coronary artery disease, is an important indicator of absence of significant (>50%) stenoses of the coronary arteries 23. Therefore, a high negative predictive value for interscan comparisons is desirable. The rate of discordant pairs (with calcium disappearing or suddenly appearing between two subsequent examinations) in our study was 8%. Rates of discordant pairs reported from previous interscan agreement studies using ECG-gating/triggering and varying tube currents are ranged form 0-6% respectively 3,17,24,25. The result from our study is only moderately worse when compared with these rates of discordant pairs and was obtained at a considerably reduced radiation dose to each patient. The higher noise levels in low-dose ungated MDCT compared with ECG-gated MDCT will in most cases be the cause of discordance 5. Applying a body weight-adapted scanning protocol in low-dose ungated MDCT could possibly result in improved noise levels 26. Even now, however, absence/ presence of CAC could reliably be established in 92% of subjects. 80

81 Calcium scoring as part of non-gated, low-dose CT screening for lung cancer in asymptomatic heavy smokers With respect to interscan variability, recent studies using ECG-gated MDCT scanning have reported interscan variability ranging from 12% to 32% 3,25,27 and EBCT studies from the late 90s showed variability s ranging from 13% to 51% 5,8,28-32 for both Agatston and calcium volume score algorithms. We have found interscan variability s of 60-70%. Therefore, we have to conclude that ECG-gated techniques remain the gold standard for obtaining accurate CAC scores in individual patient settings and monitoring of CAC over time, but low-dose ungated CT can be used for adequate CVD risk stratification in screening populations. Although percentage differences are the most common way to present variability, we like to stress that this measure actually is quite inappropriate to infer the true variability: only absolute differences in scores reflect the true variability (e.g. an absolute inter-scan difference of only 8 in Agatston score can be presented as a relative difference of as high as 200% (AS scan 1 = 0; AS scan 2 = 8) or as little as 2% (AS scan 1 = 400; AS score scan 2 = 408)). In this study, calcium volume and calcium mass score show significantly better interscan agreement than the Agatston score. This is in accordance with results from previous studies 33. Rather surprisingly, the volume score performs also significantly better than the mass score. A recent study by Hoffmann et al 34 suggested the superiority of the mass score. The theoretical advantage of mass above volume score is that it uses density information to correct for partial volume effects. However, in our ungated scans the influence of motion artefacts on the density of calcified plaques is far bigger than the influence of partial volume effects. We postulate, therefore, that the positive effect of the calcium mass score is outdone by the relatively poor quality of the scans. One limitation of our study is that we have calculated interscan variability between two scans with a mean interval of three months. To a certain extent, this interferes with a one-to-one comparison with other studies that have usually performed two scans only minutes apart. Normal progression of CAC is estimated at 14%-27% per year 15. Consequently, part of the variability observed in our study could be attributed to the real progression of CAC over the course of three months. Therefore, we expect that results from this study would have even been better if we had been able to perform two scans for all participants at baseline. Another limitation of our study is that in establishing the 95% confidence limits for different calcium measures, controlling for participant-specific covariates (BMI, CAC score) is a way to improve the generalizability of these results. However, we could not adjust calcium scores for body mass index (BMI). Previous work has shown that greater BMI is associated with lower inter-scan reproducibility, possibly caused by an increase in image noise 35. The use of a lowdose scan protocol in overweight people with low CAC scores is likely to interfere with the accurate detection of true calcifications. Even when it would have been possible to control for all participant-specific covariates, however, different scanner types from different vendors seem to limit the generalizability of 95% confidence limits derived from a single study even more profoundly

82 Chapter 5 CAC scoring with low-dose ungated MDCT as part of lung cancer screening can reliably be performed with good interscan agreement for the stratification of participants into CVD risk categories. This makes low-dose ungated MDCT a potentially valuable tool in the assessment of cardiovascular risk in large screening populations at a substantially reduced radiation dose. 82

83 Calcium scoring as part of non-gated, low-dose CT screening for lung cancer in asymptomatic heavy smokers References 1. Greenland P, Bonow RO, Brundage BH et al. ACCF/AHA 2007 clinical expert consensus document on coronary artery calcium scoring by computed tomography in global cardiovascular risk assessment and in evaluation of patients with chest pain: a report of the American College of Cardiology Foundation Clinical Expert Consensus Task Force (ACCF/AHA Writing Committee to Update the 2000 Expert Consensus Document on Electron Beam Computed Tomography) developed in collaboration with the Society of Atherosclerosis Imaging and Prevention and the Society of Cardiovascular Computed Tomography. J Am Coll Cardiol. 2007;49: Schmermund A, Erbel R, Silber S. Age and gender distribution of coronary artery calcium measured by four-slice computed tomography in 2,030 persons with no symptoms of coronary artery disease. Am J Cardiol. 2002;90: Horiguchi J, Yamamoto H, Hirai N et al. Variability of repeated coronary artery calcium measurements on low-dose ECG-gated 16-MDCT. AJR Am J Roentgenol. 2006;187:W1-W6. 4. Shemesh J, Evron R, Koren-Morag N et al. Coronary artery calcium measurement with multi-detector row CT and low radiation dose: comparison between 55 and 165 mas. Radiology. 2005;236: Bielak LF, Kaufmann RB, Moll PP, McCollough CH, Schwartz RS, Sheedy PF. Small lesions in the heart identified at electron beam CT: calcification or noise? Radiology. 1994;192: Kaufmann RB, Sheedy PF, Breen JF et al. Detection of heart calcification with electron beam CT: interobserver and intraobserver reliability for scoring quantification. Radiology. 1994;190: Horiguchi J, Fukuda H, Yamamoto H et al. The impact of motion artifacts on the reproducibility of repeated coronary artery calcium measurements. Eur Radiol. 2007;17: Kajinami K, Seki H, Takekoshi N, Mabuchi H. Quantification of coronary artery calcification using ultrafast computed tomography: reproducibility of measurements. Coron Artery Dis. 1993;4: Kim SM, Chung MJ, Lee KS, Choe YH, Yi CA, Choe BK. Coronary calcium screening using low-dose lung cancer screening: effectiveness of MDCT with retrospective reconstruction. AJR Am J Roentgenol. 2008;190: Wu MT, Yang P, Huang YL et al. Coronary arterial calcification on low-dose ungated MDCT for lung cancer screening: concordance study with dedicated cardiac CT. AJR Am J Roentgenol. 2008;190: Chen J, Krumholz HM. How useful is computed tomography for screening for coronary artery disease? Screening for coronary artery disease with electron-beam computed tomography is not useful. Circulation. 2006;113: Rumberger JA, Brundage BH, Rader DJ, Kondos G. Electron beam computed tomographic coronary calcium scanning: a review and guidelines for use in asymptomatic persons. Mayo Clin Proc. 1999;74: van Iersel CA, de Koning HJ, Draisma G et al. Risk-based selection from the general population in a screening trial: selection criteria, recruitment and power for the Dutch-Belgian randomised lung cancer multi-slice CT screening trial (NELSON). Int J Cancer. 2007;120: Xu DM, Gietema H, de Koning H et al. Nodule management protocol of the NELSON randomised lung cancer screening trial. Lung Cancer. 2006;54: Maher JE, Bielak LF, Raz JA, Sheedy PF, Schwartz RS, Peyser PA. Progression of coronary artery calcification: a pilot study. Mayo Clin Proc. 1999;74:

84 Chapter Isgum I, Rutten A, Prokop M, van GB. Detection of coronary calcifications from computed tomography scans for automated risk assessment of coronary artery disease. Med Phys. 2007;34: Rutten A, Isgum I, Prokop M. Coronary calcification: effect of small variation of scan starting position on Agatston, volume, and mass scores. Radiology. 2008;246: Ulzheimer S, Kalender WA. Assessment of calcium scoring performance in cardiac computed tomography. Eur Radiol. 2003;13: Bland JM, Altman DG. Measuring agreement in method comparison studies. Stat Methods Med Res. 1999;8: Detrano RC, Anderson M, Nelson J et al. Coronary calcium measurements: effect of CT scanner type and calcium measure on rescan reproducibility--mesa study. Radiology. 2005;236: Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics. 1977;33: Shemesh J, Henschke CI, Farooqi A et al. Frequency of coronary artery calcification on low-dose computed tomography screening for lung cancer. Clin Imaging. 2006;30: Rubinshtein R, Gaspar T, Halon DA, Goldstein J, Peled N, Lewis BS. Prevalence and extent of obstructive coronary artery disease in patients with zero or low calcium score undergoing 64-slice cardiac multidetector computed tomography for evaluation of a chest pain syndrome. Am J Cardiol. 2007;99: Chung H, McClelland RL, Katz R, Carr JJ, Budoff MJ. Repeatability limits for measurement of coronary artery calcified plaque with cardiac CT in the Multi-Ethnic Study of Atherosclerosis. AJR Am J Roentgenol. 2008;190:W87-W Takahashi N, Bae KT. Quantification of coronary artery calcium with multi-detector row CT: assessing interscan variability with different tube currents pilot study. Radiology. 2003;228: Mahnken AH, Wildberger JE, Simon J et al. Detection of coronary calcifications: feasibility of dose reduction with a body weight-adapted examination protocol. AJR Am J Roentgenol. 2003;181: Detrano R, Kang X, Mahaisavariya P et al. Accuracy of quantifying coronary hydroxyapatite with electron beam tomography. Invest Radiol. 1994;29: Callister TQ, Cooil B, Raya SP, Lippolis NJ, Russo DJ, Raggi P. Coronary artery disease: improved reproducibility of calcium scoring with an electron-beam CT volumetric method. Radiology. 1998;208: Devries S, Wolfkiel C, Shah V, Chomka E, Rich S. Reproducibility of the measurement of coronary calcium with ultrafast computed tomography. Am J Cardiol. 1995;75: Mao S, Bakhsheshi H, Lu B, Liu SC, Oudiz RJ, Budoff MJ. Effect of electrocardiogram triggering on reproducibility of coronary artery calcium scoring. Radiology. 2001;220: Mohlenkamp S, Behrenbeck TR, Pump H et al. Reproducibility of two coronary calcium quantification algorithms in patients with different degrees of calcification. Int J Cardiovasc Imaging. 2001;17: Wang S, Detrano RC, Secci A et al. Detection of coronary calcification with electron-beam computed tomography: evaluation of interexamination reproducibility and comparison of three image-acquisition protocols. Am Heart J. 1996;132: Hong C, Bae KT, Pilgram TK. Coronary artery calcium: accuracy and reproducibility of measurements with multidetector row CT--assessment of effects of different thresholds and quantification methods. Radiology. 2003;227:

85 Calcium scoring as part of non-gated, low-dose CT screening for lung cancer in asymptomatic heavy smokers 34. Hoffmann U, Siebert U, Bull-Stewart A et al. Evidence for lower variability of coronary artery calcium mineral mass measurements by multi-detector computed tomography in a community-based cohort--consequences for progression studies. Eur J Radiol. 2006;57: Sevrukov A, Pratap A, Doss C, Jelnin V, Hoff JA, Kondos GT. Electron beam tomography imaging of coronary calcium: the effect of body mass index on radiologic noise. J Comput Assist Tomogr. 2002;26:

86

87 CHAPTER 6 Coronary artery calcium can predict all-cause mortality and cardiovascular events on low-dose computed tomography screening for lung cancer

88 Chapter 6 Abstract Aims Performing CAC screening as part of low-dose lung cancer screening has been proposed as an efficient strategy to detect people with high cardiovascular risk and improve outcomes of primary prevention. This study aims to investigate whether coronary artery calcium (CAC) measured on low-dose computed tomography in a population of heavy (ex) smokers, is an independent predictor of all-cause mortality and cardiac events. Methods and results We used a case-cohort study among 958 subjects 50 years of age within the screen group of a randomized controlled lung cancer screening trial.we used Cox proportional-hazard models to compute hazard ratios adjusted for traditional cardiovascular risk factors to predict all-cause mortality and cardiovascular events. During a median follow-up of 21.5 months, 56 deaths and 127 cardiovascular events occurred. Compared with a CAC score of 0, multivariate-adjusted hazard ratios for all-cause mortality for CAC scores of 1-100, , and >1000 were 3.00 (95% CI, ), 6.13 (95%, CI ), and (95% CI, ). Multivariate-adjusted hazard ratios for coronary events were 1.38 (95% CI, ), 3.04 (95% CI, ), and 7.77 (95% CI, ). Of subjects in the highest risk category (CAC>1000), 43% did not receive optimal preventive treatment of CVD risk factors. Conclusion This study shows that low-dose CT can be used for the detection of CAC as an independent predictor of all-cause mortality and cardiovascular events. Applying this strategy to baseline lung cancer screening, 1 in 12 participants screened could benefit from starting preventive CVD risk factor treatment. 88

89 Coronary artery calcium can predict all-cause mortality and cardiovascular events Introduction Tobacco use causes both lung cancer and cardiovascular disease (CVD) and is the most important preventable cause of death worldwide 1. Early diagnosis is considered an important goal for both diseases. The effectiveness of CT screening to reduce mortality from lung cancer is currently being investigated 2. Coronary artery calcium (CAC) screening has been proposed as an attractive addition to screening for lung cancer 3. Adding CAC screening at baseline could lead to improved detection of high risk individuals and, consequently, the improved primary prevention of cardiovascular events through optimized medical treatment of cardiovascular risk factors. Furthermore, it is likely to offer benefits in terms of efficiency, radiation dose reduction and cost-effectiveness. Dedicated CAC screening is usually performed with high radiation dose and ECG-synchronization is considered a necessary part of the scanning protocol to reduce motion artifacts. A recent report, however, has demonstrated the feasibility of accurate CAC screening using non-gated, low-dose computed tomography (CT) compared with dedicated cardiac CT 4. The predictive properties of CAC above and beyond traditional risk factors for CVD have been thoroughly described in a number of large prospective cohort studies using electronbeam tomography (EBT) In spite of this, the utility of screening for CAC as a predictor of cardiovascular events remains contested 13. One of several factors that has been shown to influence the predictive value of CAC is the variability of underlying risk profiles across different populations studied 14. This calls for separate evaluation of the usefulness of CAC screening in different types of study populations. Currently, no outcome data are available on the association of CAC with all-cause mortality or risk of future cardiovascular events from a high-risk, asymptomatic lung cancer screening population using non-gated, low-dose CT. Within the NELSON study, a randomized controlled population-based lung cancer screening trial among heavy former or current smokers 50 years of age, we investigated whether CAC is an independent predictor of all-cause mortality, fatal and non-fatal cardiovascular events and fatal and non-fatal coronary events. Furthermore, we investigated the possible effect that this approach could have on the management of CVD risk factors within this screening population and its capacity to improve primary CVD prevention. Methods Patient selection The NELSON Study is a randomized controlled population-based trial comprising men and women 50 years of age. Its overall aim is to investigate the beneficial effects of screening for lung cancer with low-dose CT. In , in four distinct geographical areas (147 municipalities) all men born between living in 101 municipalities, and all 89

90 Chapter 6 women born between living in the remaining 46 municipalities were invited by mail to participate in this study. A more detailed description of patient slecetion and data collection has been described elsewhere (2). Every participant had a history of 15 pack years of smoking. From 2004 to 2006, baseline CT scans were performed in 7557 participants randomly allocated to the screen group. All analyses in the present study are derived from the screen group of the NELSON Study. The Medical Ethics Committees of all four participating hospitals approved the NELSON Study protocol, and written informed consent was obtained from all participants. Study design We used a case-cohort design 15 in which a random control sample is drawn from the baseline screen group of the NELSON trial at the beginning of the study (so-called subcohort, n=925). Cases are defined as all participants from the baseline screen group experiencing an outcome of interest (all-cause death or cardiovascular events) during follow-up. All cases (n=226) were ascertained through linkage with the national death registry and the national registry of hospital discharge diagnoses. The choice of the sample fraction (~12%) was calculated to correspond to approximately 4 controls per case detected. The major advantage of this design is that it enables the performance of survival analyses that produce valid estimates for the total study population without the need to perform time-consuming CAC scoring in all study participants. Through linkage with the national registry of hospital discharge diagnoses, we first excluded participants with a known history of cardiovascular disease (cases, n=72; subcohort, n=97). In this registry, all diagnoses are coded according to the International Classification of Diseases, 9th revision (ICD-9-CM). One research physician selected and excluded all subjects with a cardiovascular discharge diagnosis prior to the start of this study (January 2004) (ICD-9 codes , 428, , 440, 441, 443 and 444). Then, we excluded those participants who had missing baseline CT scans (cases, n=4; subcohort, n=17), or a baseline CT scan performed after follow-up had ended (cases, n=0; subcohort, n=3). This results in a final study cohort for this study of 958 subjects (cases, n= 150; subcohort, n=808). Baseline CAC scores were measured in all 958 subjects. Low-dose CT methods Baseline low-dose CT scans were conducted with a 16-slice MDCT scanner (Mx8000 IDT; Philips Medical Systems, Cleveland, Ohio in two participating hospitals; Sensation-16, Siemems AG, Forchheim, Germany in the third hospital). The scanning protocol has been described elsewhere 16. Briefly, the following parameters were applied: 16x0.75mm collimation; pitch ; caudocranial scan direction; smallest field of view to include the outer rib margins. This way, transverse images with 1.0mm section thickness and 0.7mm increment were acquired 90

91 Coronary artery calcium can predict all-cause mortality and cardiovascular events from the level of the lung bases to the lung apices. No electrocardiographic triggering was performed; no contrast agent was administered. Low-dose exposure settings were applied based on body weight: 30 mas at a tube voltage of 120 kvp for subjects 80kg and 140 kvp for subjects >80kg. This corresponds to an effective radiation dose of mSv All 958 CT scans were equally divided between two observers with two and three years of experience in reading cardiac CT who subsequently performed calcium scoring of the coronary arteries. The readers were blinded to other participant data. Prior to the start of the study, inter-observer variability of coronary calcium scoring was measured in a subset of 50 baseline scans not included in this study (κappa=0.72). To reduce image noise and to use data comparable to previously published studies 17 all scans were reconstructed to 3.1mm thick slices with an increment of 1.4 mm by averaging four neighboring slices. Calcium scoring was performed in these reconstructed images using software described in Isgum and Rutten 18,19 : all regions of 3 adjoining voxels (0.7 mm 3 ) with attenuation above 130 HU 20 were shown with a colored overlay. An investigator identified a point in each calcified lesion. Subsequently, three-dimensional component labeling using 26-connectivity was automatically performed 21 to mark all connected voxels as calcification. Care was taken not to include noncoronary calcifications (e.g. valve calcifications) or hyperattenuating foci due to image noise. Agatston scores were computed as outlined in Ulzheimer 22,23. Apart from being considered as a continuous measure, the Agatston score for coronary calcium was categorized into 4 welldefined risk categories: CAC= 0 (very low risk, reference category); CAC >0-100 (low risk); CAC > (moderate to high risk); CAC >1000 (very high risk). Classification of end points All participants in the screen group of the NELSON Study (n=7557) were linked with the national death registry and the national registry of hospital discharge diagnoses. This database linkage was performed on the basis of birth date, sex and postal code with a validated probabilistic method All-cause mortality was chosen as the primary end point for this study. Through linkage with the national death registry for the years a total of 56 deaths were detected. Secondary end points were defined as (1) a composite CVD end point consisting of cardiovascular deaths and all nonfatal cardiovascular hospital admissions and (2) a composite coronary heart disease (CHD) end point consisting of all fatal myocardial infarctions and nonfatal CHD admissions. To retrieve information on cardiovascular hospital admissions, all participants from the screen group were linked with the national registry of hospital discharge diagnoses for the years One research physician selected all cardiovascular discharge diagnoses and classified them as coronary heart disease (CHD) (codes ), or other CVD hospitalizations including peripheral arterial occlusive disease (codes 440, ), aortic aneurysm/dissection (code 441), cerebrovascular disease (codes ), heart failure (code 91

92 Chapter 6 428), and non-rheumatic valvular disease (code 424). All other codes included in the ICD-9- CM as diseases of the circulatory system were not included as valid end points. Through this linkage a total of 94 nonfatal cardiovascular events could be identified. Follow-up started after the baseline CT scan. Follow-up time differed for primary and secondary end points, because of the differential availability of the two registries used. For allcause mortality, follow-up was complete until January 1, 2007 (median: 21.5 months); for both secondary end points, until January 1, 2006 (median: 9.5 and 10.0). For all participants who experienced an event, follow-up ended at the date of diagnosis or death. Participants with a CVD hospital admission prior to death were counted as all-cause mortality. In participants with multiple cardiovascular hospital admissions during follow-up, the first hospital discharge diagnosis was used as end point. Assessment of covariates At baseline, all participants from the NELSON Study were asked to return a questionnaire containing information on prior and current smoking behavior. For those subjects drawn into the final study cohort of this study, a research physician collected information from their general practioners (GPs) using a standardized questionnaire. The obtained information included the current use of drugs, specifically the use of antihypertensive drugs (defined as diuretics, ACE inhibitors, angiotensin II receptor antagonists, β-blockers and/or calcium channel blockers); lipid-lowering drugs; oral hypoglycemic agents; insulin; and other drugs prescribed for cardiovascular purposes (most commonly antiplatelet drugs (e,g, acetylsalicylic acid, clopidogrel); systolic and diastolic blood pressure (BP); and nonfasting blood glucose, HbA1c, total cholesterol, LDL cholesterol and HDL cholesterol levels. The overall response rate was 70%. For all covariates obtained through the GP, missing values were imputed using regression methods implemented in SPSS software (SPSS 14.0, Chicago, Illinois) 28. We defined diabetes mellitus as a nonfasting glucose level 11.1 mmol/l and/or the use of oral hypoglycemic agents or insulin. Hypertension was defined as a diastolic BP >90mmHg, systolic BP >140mmHg and/or the use antihypertensive drugs. Hypercholesterolaemia was defined as a total cholesterol level >5.0 mmol/l, an LDL level >3.0 mmol/l and/or the use of lipid-lowering drugs. Statistical methods Baseline characteristics were summarized for the subcohort and the three different casegroups separately. Categorical variables were compared with a χ 2 statistic; continuous variables with t-tests. Means and SDs were computed for normally distributed variables; medians and interquartile ranges for variables with skewed distributions. Annualized event rates for all three end points in the full cohort were calculated by weighting the number of subjects per CAC risk category in the subcohort by the inverse of the sampling fraction (n subcohort * 1/0.107). 92

93 Coronary artery calcium can predict all-cause mortality and cardiovascular events The association of CAC with all-cause mortality, the composite CVD end point and coronary heart disease was evaluated with Cox proportional hazard analyses with modification of the standard errors based on robust variance estimates. We used the method according to Prentice in which all subcohort members are equally weighted. Cases outside the subcohort are not weighed before failure and at failure receive the same weight as members of the subcohort 15. This method has been shown to resemble estimates from a full-cohort analysis most accurately 29. We classified participants using a 4-level risk stratification based on the Agatston scoring (AS) algorithm 23 and used the lowest level (AS=0) as the reference category. First, Cox proportional hazard models for all three end points were performed for Agatston score risk categories unadjusted for covariates (model 1). In model 2, we adjusted for age and sex. In model 3, current smoking and history of hypertension, diabetes and hypercholesterolaemia were added. We tested the interaction of CAC risk categories with smoking status by entering this interaction term in the models. Unadjusted and risk-adjusted Cox proportional hazard survival curves were drawn to show the effects of CAC per risk category. All analyses were performed with the statistical software package SPSS (SPSS for Windows, release 14.0, Chicago, Illinois) and the cch(survival) package (standard available in R software, version 6.2). Results Table 1 shows the baseline characteristics of the subcohort and the all-cause mortality, cardiovascular disease and coronary heart disease cases. At the time of baseline screening, mean age of the members in the subcohort was 59.5±5.6 years. Current cigarette smoking was recorded for 56%, diabetes for 7%, hypercholesterolaemia for 75%, hypertension for 64%. CAC = 0 was detected in 24% of participants: CAC scores of 1-100, , and >1000 were detected in 29%, 30%, and 17% respectively (median CAC score: 74). As presented in Table 2, there is a strong association between increasing categories of CAC and the annualized event rate for all three end points. The mortality for all participants increased in a linear fashion from 0.08% to 0.2%, 0.6% and 1.1% with increasing CAC categories of 0, 1-100, and >1000, respectively (p < 0.001). 93

94 Chapter 6 Table 1 Baseline Characteristics for Subjects in the Subcohort, All-Cause Mortality, Fatal/Nonfatal Cardiovascular (CVD) Events, and Fatal/Nonfatal Coronary (CHD) Events Case-Groups Variable Subcohort (n=808) All-cause mortality a (n=56) CVD end point a (n=127) CHD end point a (n=61) Age, y 59.5± ± ± ±5.6 Men, % Hypertension, % Hypercholesterolaemia, % Diabetes, % Current smokers, % CAC score (AS) b > CAC score (AS) continuous c 74(591) 685(1828) 769(2063) 1055(2017) Categorical variables are expressed as percentage. a Median follow-up per case-group: 21.5 months for all-cause mortality; 9.5 months for CVD events; 10.0 months for CHD events b AS, Agatston score. c Median (interquartile range) Table 2 Annualized Event Rates for All-Cause Mortality, Fatal/Nonfatal Cardiovascular (CVD) Events and Fatal/ Nonfatal Coronary (CHD) Events According to CAC Risk Categories a Characteristic All-cause mortality CVD end point b CHD end point b CAC risk category c (2) 0.7 (10) 0.3 (4) (8) 1.5 (27) 0.4 (8) (22) 1.7 (32) 1.0 (18) > (24) 6.1 (58) 3.2 (31) Data are percentages. Data in parentheses are numbers of patients. a Median follow-up per case-group: 21.5 months for all-cause mortality; 9.5 months for CVD events; 9.8 months for CHD events. b CVD end point (n=127) consists of 10 fatal (myocardial infarction (MI) (n=5), stroke (n=3), aortic aneurysm (AA) (n=1), and peripheral arterial occlusive disease (PAOD) (n=1)) and 117 nonfatal events (MI (n=13), angina pectoris (n=43), aortic valve stenosis (n=24), stroke (n=14), AA (n=12), and PAOD (n=11)). Of these 127 events, all fatal/nonfatal MI (n=18) and angina pectoris (n=43) events were included in the CHD end point (n=61). c To estimate risks for the full cohort, the number of participants in the denumerator needs to be estimated by weighting the number of subjects per CAC risk category in the subcohort by the inverse of the sampling fraction (n subcohort * 1/0.107) 94

95 Coronary artery calcium can predict all-cause mortality and cardiovascular events During a median follow-up of 21.5 months (range: days), 56 subjects died. CAC score was associated with risk of all-cause mortality (Table 3). Compared with the reference category (CAC=0) crude hazard ratios for CAC scores of 1-100, , and >1000 were 3.03, 8.26, and 16.17, respectively. After adjustment for age and sex the strength of the association was attenuated. Compared with a CAC score of 0, hazard ratios for scores of 1-100, , and >1000 were 2.82, 5.96, and Adjustment for additional cardiovascular risk factors did not materially change the association of CAC with all-cause mortality. During a median follow-up 10.0 months (range: days) 127 incident fatal/non-fatal CVD events of which 61 were fatal/non-fatal coronary events (see Table 2 for specification of end points). CAC showed a graded association with the risk of all fatal/non-fatal CVD events and with only coronary events. In case of fatal/non-fatal coronary events, crude hazard ratios for scores of 1-100, , and >1000 compared with a CAC score of 0 were 1.42, 3.13, and 9.97, respectively (Table 3). After adjustment for all cardiovascular risk factors, the corresponding hazard ratios were 1.38, 3.04, and Attenuation of these hazard ratios was mainly caused by adjustment for hypertension. Compared with all-cause mortality, adjustment for age did not materially influence the association for these two end points. The models for all CVD events showed a similar pattern, although the strength of the association was slightly less than for coronary events alone (Table 3). No interaction was present between CAC and current smoking (p for interaction = 0.9) in relation to all three end points. Most probably this is due to the fact that the entire study population consisted of heavy (ex) smokers. To illustrate the possible benefits in terms of improved primary prevention when adding CAC screening to lung cancer screening, we summarized the percentage of subjects treated and not treated with antihypertensive drugs and statins stratified into CAC categories (Table 4). In subjects with an intermediate-to-high risk of CVD events (CAC score= ), 55% of participants were not treated with antihypertensive drugs and 64% were not treated with statins. In subjects at very high risk (CAC score >1000), still 43% and 48% of people were not treated with antihypertensive drugs and statins, respectively. 95

96 Chapter 6 Table 3 Hazard Ratios for Events According to CAC Risk Categories HR (95% CI) Model 1 Model 2 Model 3 All-cause mortality (n=56) CAC risk categories ( ) 2.82 ( ) 3.00 ( ) ( ) 5.96 ( ) 6.13 ( ) > ( ) ( ) ( ) CVD end point (n=127) CAC risk categories ( ) 1.88 ( ) 1.76 ( ) ( ) 2.09 ( ) 1.93 ( ) > ( ) 6.88 ( ) 5.33 ( ) CHD end point (n=61) CAC risk categories ( ) 1.49 ( ) 1.38 ( ) ( ) 3.45 ( ) 3.04 ( ) > ( ) ( ) 7.77 ( ) Model 1 shows crude hazard ratios; model 2 was adjusted for age and sex; model 3 was adjusted for age, sex, smoking, hypertension, hypercholesterolaemia, and diabetes. Table 4 Proportion of Patients Treated/Untreated with Antihypertensive Drugs or Statins in the Subcohort per CAC Category (n=808) Antihypertensive drugs Statins Not Treated Treated Not Treated Treated CAC risk category 0 (n=197) 70 (137) 30 (60) 76 (149) 24 (48) (n=231) 67 (154) 33 (77) 74 (170) 26 (61) (n=239) 55 (131) 45 (108) 64 (154) 36 (85) >1000 (n=141) 43 (60) a 57 (81) 48 (68) b 52 (73) Data are percentages. Data in parentheses are numbers of patients. a Of the 60 participants with a CAC score >1000 not treated with antihypertensive drugs 26 participants (43%) actually had hypertension (diastolic BP >90 mmhg or systolic BP > 140 mmhg). b Of the 68 participants with a CAC score >1000 not treated with statins 42 (62%) actually had elevated lipid levels (Total cholesterol > 5.0 mmol/l or LDL > 3.0 mmol/l). 96

97 Coronary artery calcium can predict all-cause mortality and cardiovascular events Discussion The present study shows that CAC is an independent predictor of all-cause mortality, fatal/ nonfatal cardiovascular events and fatal/nonfatal coronary events in a lung cancer screening population. These associations are independent of traditional cardiovascular risk factors. Before interpreting the results of this study, issues with respect to different types of cardiac imaging techniques should be considered. Dedicated CAC screening has traditionally been performed with EBT imaging 30. The comparable accuracy and reproducibility of detecting CAC with normal-dose, ECG-gated multislice CT of the heart (MDCT) versus EBT had already been established 31,32, but until now performing CAC screening on low-dose, non-gated CT was considered too inadequate to be clinically useful. A recent report, however, has shown that this technique is valid to be used for CAC screening with an accuracy of detecting CAC for up to 90% compared with dedicated MDCT and correct risk stratification in almost 83% of participants 4. Based on these results, we believe that CAC scoring using a non-gated, low-dose CT protocol can be employed as a clinically useful tool to detect people at risk and improve risk stratification of participants in a lung cancer screening program. Study strengths This is the first study that shows the predictive value of CAC in a cohort of heavy (ex) smokers participating in a lung cancer screening trial using low-dose, non-gated CT. One previous study similarly reported good predictive value of CAC in a large cohort of (ex) smokers using higher-dose, ECG-gated EBT imaging 33. However, their patients were derived from a cohort of asymptomatic individuals referred for evaluation of cardiac risk by their GP. Comparing baseline characteristics, this has resulted in a younger and healthier population than in our study. As a consequence our population has higher CAC scores and is therefore at higher risk of all-cause mortality and CVD events. Consequently, we report higher risk ratios than those reported by Shaw et al 33. We feel that the importance of our results is threefold. First, this study shows that CAC, when measured on non-gated, low-dose CT as part of a lung cancer screening program in heavy smokers, can predict all-cause mortality and cardiovascular events in a comparable way as when using a normal dose, gated CT protocol. Subjects in the highest risk category had an almost sixfold risk of coronary events compared with those without any detectable CAC. These results are comparable with several different prospective cohort studies previously described 5-12,33. In general, these studies demonstrated strong and independent associations between CAC and poor clinical outcome with e.g. risk-adjusted hazard ratios of 9.4 for all-cause mortality for participants with a score>1000 (HR=10.9 in present study) 6. Younger patients (mean age, 43 years) were analyzed in the Prospective Army Coronary Calcium Project and they reported an 11-fold increased risk for hard coronary events (HR=7.8 in present study)

98 Chapter 6 We think that these minor differences can be explained by the different underlying risk profiles of various study populations. One study that specifically selected participants with a high-risk profile ( 2 coronary risk factors; mean age, 66 years), and therefore more closely resembling our population, reported a much weaker association (RR=2.3 for CAC score above the median) 34. Our results are somewhere in between those described above, reflecting in one respect the high-risk profile of our participants, but on the other hand that mean age in our study is approximately 7 years lower than in the former example 34. As outlined above, these studies invariably used EBT imaging protocols for CAC scoring with, typically, an effective radiation dose of between 1.0 and 1.3 msv; effective radiation dose of calcium scoring with MDCT the technique nowadays most often used in clinical practice ranges between 1.5 and 6.2 msv 35. Our study demonstrates that comparable risk estimates can be obtained with low-dose CT, typically involving an effective radiation dose of only msv. We consider the potential reduction of radiation dose as an important strength of this study. Second, this is the first study to demonstrate the very high proportion of asymptomatic heavy (ex) smokers with excessively high (CAC score >1000) CAC scores. By comparison, the two largest cohorts of asymptomatic people followed-up so far have reported a prevalence of CAC scores >1000 of 3 to 4% 6,11 compared with 17% of participants in our population. Conversely, although our study population consists of people with a minimum of 15 pack years of smoking, still 24% of participants have no detectable calcium (CAC score=0) corresponding with a low risk. Third, this study shows that approximately 55% of study participants with intermediate-tohigh risk (CAC score ) and 43% of participants with very high risk (>1000) of CVD events do not receive optimal treatment of CVD risk factors. Apart from the fact that around 50% of these subjects actually had elevated BP and/or lipid levels and should have been treated anyway, all these subjects are likely to benefit from risk factor treatment because of their extremely high (>1000) CAC score. They currently do not receive it for reasons unknown. This indicates that extra information from CAC scores could have great impact in changing the management of CVD risk factors in these particular subjects. For example, when performing CAC scoring as part of baseline lung cancer screening in 1000 participants without a history of coronary heart disease, one would detect 175 subjects with a CAC score >1000 (massively increased risk); and of these 175 subjects, 84 subjects (1 in 12 of all participants screened) could benefit from starting secondary preventive medical treatment of cardiovascular risk factors by initiating antihypertensive and/or lipid-lowering drugs that they did not receive before. A scenario analysis and cost-effectiveness study will be needed to calculate the exact contribution of incorporating CAC screening into lung cancer screening. 98

99 Coronary artery calcium can predict all-cause mortality and cardiovascular events Study limitations Several possible limitations of our study need to be considered. This study has a rather limited follow-up, especially for cardiovascular and coronary events, but despite this fact a substantial number of events did occur. In our cohort only 17% of participants were female. This could limit the generalizability of our results to women. However, CAC has previously been shown to be a comparable predictor of CVD risk in men and women 9. In the present study most baseline information has been obtained through GPs. A response rate by GPs of 70% may be a limitation, yet we did not observe any differential pattern of missing values for cases and non-cases. Imputation of missing covariate information, taking into account the 100% completeness of follow-up for all end points, has been shown superior to a complete case analysis 28,36,37. Although risk stratification of subjects has been shown accurate using our lowdose, non-gated CT protocol making it a valid tool for identification of high-risk CT screening participants, inter-scan variability of absolute Agatston scores in individual patients is still considerable. For this reason, the low-dose CT scanning protocol will be optimized during the next screening round. Finally, assessment of a history of cardiovascular disease prior to the start of this study was unavailable for years prior to Linkage of patient characteristics with the registry of hospital discharge diagnoses was not possible before that date. This will have resulted in an underestimation of participants with a history of cardiovascular disease. We believe that, given the mean age of our population in 1995 and the fact that the vast majority of CVD events takes place >45 years of age, that this underestimation will not have influenced the results of our study substantially. Conclusion Our results demonstrate that CAC is a strong and independent predictor of all-cause mortality and cardiovascular events in a population of heavy (ex) smokers participating in lung cancer screening using a low-dose CT protocol. Our results further suggest that adding CAC scoring to lung cancer screening could be a clinically useful tool for detection of people at risk for cardiovascular disease and to improve primary prevention of CVD events by optimizing risk factor management in these patients. 99

100 Chapter 6 References 1. Gerhardsson de Verdier M. The big three concept: a way to tackle the health care crisis? Proc Am Thorac Soc. 2008;5: van Iersel CA, de Koning HJ, Draisma G et al. Risk-based selection from the general population in a screening trial: selection criteria, recruitment and power for the Dutch-Belgian randomised lung cancer multi-slice CT screening trial (NELSON). Int J Cancer. 2007;120: Takasu J, Budoff MJ, O Brien KD et al. Relationship between coronary artery and descending thoracic aortic calcification as detected by computed tomography: The Multi-Ethnic Study of Atherosclerosis. Atherosclerosis [Epub ahead of print]. 4. Kim SM, Chung MJ, Lee KS, Choe YH, Yi CA, Choe BK. Coronary calcium screening using low-dose lung cancer screening: effectiveness of MDCT with retrospective reconstruction. AJR Am J Roentgenol. 2008;190: Arad Y, Goodman KJ, Roth M, Newstein D, Guerci AD. Coronary calcification, coronary disease risk factors, C-reactive protein, and atherosclerotic cardiovascular disease events: the St. Francis Heart Study. J Am Coll Cardiol. 2005;46: Budoff MJ, Shaw LJ, Liu ST et al. Long-term prognosis associated with coronary calcification: observations from a registry of 25,253 patients. J Am Coll Cardiol. 2007;49: Greenland P, LaBree L, Azen SP, Doherty TM, Detrano RC. Coronary artery calcium score combined with Framingham score for risk prediction in asymptomatic individuals. JAMA. 2004;291: Kondos GT, Hoff JA, Sevrukov A et al. Electron-beam tomography coronary artery calcium and cardiac events: a 37-month follow-up of 5635 initially asymptomatic low- to intermediate-risk adults. Circulation. 2003;107: LaMonte MJ, FitzGerald SJ, Church TS et al. Coronary artery calcium score and coronary heart disease events in a large cohort of asymptomatic men and women. Am J Epidemiol. 2005;162: Raggi P, Callister TQ, Cooil B et al. Identification of patients at increased risk of first unheralded acute myocardial infarction by electron-beam computed tomography. Circulation. 2000;101: Shaw LJ, Raggi P, Schisterman E, Berman DS, Callister TQ. Prognostic value of cardiac risk factors and coronary artery calcium screening for all-cause mortality. Radiology. 2003;228: Taylor AJ, Bindeman J, Feuerstein I, Cao F, Brazaitis M, O Malley PG. Coronary calcium independently predicts incident premature coronary heart disease over measured cardiovascular risk factors: mean three-year outcomes in the Prospective Army Coronary Calcium (PACC) project. J Am Coll Cardiol. 2005;46: Greenland P, Lloyd-Jones D. Defining a rational approach to screening for cardiovascular risk in asymptomatic patients. J Am Coll Cardiol. 2008;52: Shaw LJ, O Rourke RA. The challenge of improving risk assessment in asymptomatic individuals: the additive prognostic value of electron beam tomography? J Am Coll Cardiol. 2000;36: Prentice RL. A case-cohort design for epidemiologic cohort studies and disease prevention trials. Biometrika. 1986;73: Xu DM, Gietema H, de Koning H et al. Nodule management protocol of the NELSON randomised lung cancer screening trial. Lung Cancer. 2006;54:

101 Coronary artery calcium can predict all-cause mortality and cardiovascular events 17. Detrano RC, Anderson M, Nelson J et al. Coronary calcium measurements: effect of CT scanner type and calcium measure on rescan reproducibility--mesa study. Radiology. 2005;236: Isgum I, Rutten A, Prokop M, van GB. Detection of coronary calcifications from computed tomography scans for automated risk assessment of coronary artery disease. Med Phys. 2007;34: Rutten A, Isgum I, Prokop M. Coronary calcification: effect of small variation of scan starting position on Agatston, volume, and mass scores. Radiology. 2008;246: Shemesh J, Evron R, Koren-Morag N et al. Coronary artery calcium measurement with multi-detector row CT and low radiation dose: comparison between 55 and 165 mas. Radiology. 2005;236: Jain AK. Fundamentals of digital image processing. Englewood Cliffs, NJ: Prentice-Hall, Inc; Ulzheimer S, Kalender WA. Assessment of calcium scoring performance in cardiac computed tomography. Eur Radiol. 2003;13: Agatston AS, Janowitz WR, Hildner FJ, Zusmer NR, Viamonte M, Jr., Detrano R. Quantification of coronary artery calcium using ultrafast computed tomography. J Am Coll Cardiol. 1990;15: De Bruin A, Kardaun JW, Gast A, Bruin E, van Sijl M, Verweij G. Record linkage of hospital discharge register with population register: experiences at Statistics Netherlands. Stat J UN Econ Comm Eur. 2004;21: Herings RM, Bakker A, Stricker BH, Nap G. Pharmaco-morbidity linkage: a feasibility study comparing morbidity in two pharmacy based exposure cohorts. J Epidemiol Community Health. 1992;46: Paas GR and Veenhuizen KC. Research on the validity of the LMR. Utrecht; Prismant: Reitsma JB, Kardaun JW, Gevers E, de Bruin A, van der WJ, Bonsel GJ. [Possibilities for anonymous follow-up studies of patients in Dutch national medical registrations using the Municipal Population Register: a pilot study]. Ned Tijdschr Geneeskd. 2003;147: Little RJA. Regression with missing X s: a review. J Am Stat Assoc. 1992;87: Onland-Moret NC, van der A D, van der Schouw YT et al. Analysis of case-cohort data: a comparison of different methods. J Clin Epidemiol. 2007;60: Schmermund A, Baumgart D, Erbel R. Coronary calcification by electron beam tomography: comparison with coronary risk factors and angiography. J Cardiovasc Risk. 2000;7: Becker CR, Kleffel T, Crispin A et al. Coronary artery calcium measurement: agreement of multirow detector and electron beam CT. AJR Am J Roentgenol. 2001;176: Stanford W, Thompson BH, Burns TL, Heery SD, Burr MC. Coronary artery calcium quantification at multi-detector row helical CT versus electron-beam CT. Radiology. 2004;230: Shaw LJ, Raggi P, Callister TQ, Berman DS. Prognostic value of coronary artery calcium screening in asymptomatic smokers and non-smokers. Eur Heart J. 2006;27: Detrano RC, Wong ND, Doherty TM et al. Coronary calcium does not accurately predict near-term future coronary events in high-risk adults. Circulation. 1999;99: Hunold P, Vogt FM, Schmermund A et al. Radiation exposure during cardiac CT: effective doses at multi-detector row CT and electron-beam CT. Radiology. 2003;226:

102 Chapter van der Heijden GJ, Donders AR, Stijnen T, Moons KG. Imputation of missing values is superior to complete case analysis and the missing-indicator method in multivariable diagnostic research: a clinical example. J Clin Epidemiol. 2006;59: Rubin DB, Schenker N. Multiple imputation in health-care databases: an overview and some applications. Stat Med. 1991;10:

103

104

105 CHAPTER 7 Comparing coronary artery calcium and thoracic aortic calcium for prediction of all-cause mortality and cardiovascular events on low-dose non-gated computed tomography

106 Chapter 7 Abstract Background Coronary artery calcium (CAC) and thoracic aorta calcium (TAC) can be detected simultaneously on low-dose, non-gated computed tomography (CT) scans. CAC has been shown to predict cardiovascular (CVD) and coronary (CHD) events. A comparable association between TAC and CVD events has yet to be established, but TAC could be a more reproducible alternative to CAC in low-dose, non-gated CT. This study compared CAC and TAC as independent predictors of all-cause mortality and cardiovascular events in a population of heavy smokers using low-dose, non-gated CT. Methods Within the NELSON study, a population-based lung cancer screening trial, the CT screen group consisted of 7557 heavy smokers aged years. Using a case-cohort study design, CAC and TAC scores were calculated in a total of 958 asymptomatic subjects who were followed up for all-cause death, and CVD, CHD and non-cardiac events (stroke, aortic aneurysm, peripheral arterial occlusive disease). We used Cox proportional-hazard regression to compute hazard ratios (HR) with adjustment for traditional cardiovascular risk factors. Results A close association between the prevalence of TAC and increasing levels of CAC was established (p<0.001). Increasing CAC and TAC risk categories were associated with all-cause mortality (p for trend=0.01 and 0.001, respectively) and CVD events (p for trend <0.001 and 0.03, respectively). Compared with the lowest quartile (reference category), multivariateadjusted HRs across categories of CAC were higher (all cause-mortality, HR:9.13 for highest quartile; CVD events, HR:4.46 for highest quartile) than of TAC scores (HR:5.45 and HR:2.25, respectively). However, TAC is associated with non-coronary events (HR:4.69 for highest quartile, p for trend=0.01) and CAC was not (HR:3.06 for highest quartile, p for trend=0.40). Conclusions CAC was found to be a stronger predictor than TAC of all-cause mortality and CVD events in a high risk population of heavy smokers scored on low-dose, non-gated CT. TAC, however, is stronger associated with non-cardiac events than CAC and could prove to be a preferred marker for these events. 106

107 Comparing coronary artery calcium and thoracic aortic calcium for prediction of all-cause mortality Introduction In the past, several studies using both ultrasound and plain radiography have shown an association between calcified plaques in the thoracic aorta (TAC) and cardiovascular and cerebrovascular events 1-4. CAC, measured by computed tomography (CT), has proven to be a strong and independent predictor of coronary events and all-cause mortality 5,6. Non-enhanced computed tomography (CT) can simultaneously detect both CAC and TAC. Several studies using non-enhanced CT have demonstrated a close association of TAC and CAC Based on these results, TAC has recently been suggested as an independent predictor of cardiovascular disease, but only two follow-up studies have been reported on this topic 12,13. The intuitive advantage of using TAC instead of CAC lies primarily in the fact that CAC measurement may be hampered by motion artifacts of the beating heart while this will hardly affect TAC measurements. As has recently been suggested 9, TAC could prove to be a substitute of CAC for prediction of cardiovascular events in non-gated CT. Since the heart and thoracic aorta are both depicted on low-dose CT scans for lung cancer screening, and lung cancer and cardiovascular disease share an increased risk with the prolonged use of tobacco, CAC and TAC measurements could be employed to expand the scope of the screening effort and include estimation of cardiovascular risk of screening subjects as well. Adding these measurements at baseline could lead to improved detection of high-risk individuals and, consequently, improved primary prevention of CVD events through optimized preventive treatment of cardiovascular risk factors. So far, no large follow-up studies have investigated the relationship between TAC and CAC in a cohort of asymptomatic smokers. In this study, we investigated whether TAC, as measured on low-dose, non-gated CT, can be used as an independent predictor of all-cause mortality and cardiovascular events compared with CAC. As a secondary analysis, we compared the role of CAC and TAC for the prediction of coronary and non-cardiac events separately. Methods Study population The NELSON Study is a randomized controlled population-based trial comprising men and women aged years. Its overall aim is to investigate the beneficial effects of screening for lung cancer with low-dose CT. In , in three regions in the Netherlands and one region in Belgium all men born between living in 101 distinct municipalities, and all women born between living in the remaining 46 municipalities were invited by mail to participate in this study. Every participant had a history of 15 pack years of smoking. From 2004 to 2006, baseline CT scans were performed in 7557 participants randomly allocated to the screen group. The Medical Ethics Committees of all four participating 107

108 Chapter 7 hospitals approved the NELSON study protocol, and written informed consent was obtained from all participants. A more detailed description of patient selection and data collection has been published elsewhere 14. For the present study we used a case-cohort design 15 in which cases are defined as all participants from the screen group experiencing an outcome of interest (all-cause death or cardiovascular events) during follow-up. A random sample of 925 participants was drawn from the baseline screen group (so-called subcohort). All cases (n=226) were obtained through linkage with the national death registry and the national registry of hospital discharge diagnoses. The choice of the sample fraction was calculated to correspond to approximately 4 controls per case detected. Through linkage with the national registry of hospital discharge diagnoses, we first excluded participants with a known history of hospitalization for cardiovascular disease (cases, n=72; subcohort, n=97). This database linkage was performed for the years Prior to 1995 this information could not be retrieved, since database linkage is only possible from 1995 onwards. Then, we excluded those participants who had missing baseline CT scans (cases, n=4; subcohort, n=17), or a baseline CT scan performed after follow-up had ended (cases, n=0; subcohort, n=3). This results in a final study cohort for this study of 958 subjects (additional cases, n= 150; subcohort, n=808). Baseline CAC and TAC scores were measured in all 958 subjects. CT scan acquisition and image analysis Baseline low-dose CT scans were conducted with a 16-slice MDCT scanner (Mx8000 IDT; Philips Medical Systems, Cleveland, Ohio in two participating hospitals; Sensation-16, Siemems AG, Forchheim, Germany in the third hospital). The scanning protocol consists of the following parameters were applied: 16x0.75mm collimation; pitch ; caudocranial scan direction; smallest field of view to include the outer rib margins. This way, transverse images with 1.0mm section thickness and 0.7mm increment were acquired from the level of the lung bases to the lung apices. No electrocardiographic triggering was performed; no contrast agent was administered. Low-dose exposure settings were applied based on body weight: 30 mas at a tube voltage of 120 kvp for subjects 80kg and 140 kvp for subjects >80kg. This corresponds to an effective radiation dose of msv. All 958 CT scans were equally divided between two observers with two and three years of experience in reading cardiac CT who subsequently performed calcium scoring of the coronary arteries and the thoracic aorta. The readers were blinded to other participant data. Prior to the start of the study, inter-observer variability of CAC scoring was measured in a subset of 50 baseline scans not included in this study (κappa=0.72). To reduce image noise and to use data comparable to previously published studies 16 all scans were reconstructed to 3.1mm thick slices with an increment of 1.4mm by averaging four neighboring slices. Calcium scoring was performed in these reconstructed images using a technique described in Isgum 17 : 108

109 Comparing coronary artery calcium and thoracic aortic calcium for prediction of all-cause mortality all regions of 3 adjoining voxels (0.7 mm 3 ) with attenuation above 130 HU 18 were shown with a colored overlay. An investigator identified a point in each calcified lesion. Subsequently, three-dimensional component labeling using 26-connectivity was automatically performed 19 to mark all connected voxels as calcification. Agatston scores were computed as outlined in Ulzheimer 20,21. TAC was measured in the ascending aorta, aortic arch and descending aorta inferiorly to the upper limit of the eleventh thoracic vertebra (Figure 1). Figure 1 Example of low-dose, non-gated CT image showing calcified plaques in the ascending and descending thoracic aorta, and the left main (LM) and left anterior descending (LAD) coronary arteries. In this image an automatic overlay indicates all areas of >3 voxels with an attenuation >130 HU to facilitate calcium scoring. Classification of end points All participants in the screen group of the NELSON study (n=7557) were linked with the national death registry and the national registry of hospital discharge diagnoses. This database linkage was performed on the basis of birth date, sex and postal code with a validated probabilistic method Through linkage with the national death registry for the years a total of 56 all-cause deaths were detected. Other end points were defined as (1) a composite CVD end point consisting of cardiovascular deaths and all nonfatal cardiovascular hospital admissions; that was subdivided in (2) a coronary (CHD) end point consisting of all fatal myocardial infarctions and nonfatal CHD admissions and (3) a composite end point of non-cardiac events consisting 109

110 Chapter 7 of all fatal and nonfatal cases of cerebrovascular disease, peripheral arterial occlusive disease (PAOD), and aortic aneurysms. To retrieve information on cardiovascular hospital admissions, all participants from the screen group were linked with the national registry of hospital discharge diagnoses for the years In this registry, all diagnoses are coded according to the International Classification of Diseases, 9th revision (ICD-9-CM). One research physician selected all cardiovascular discharge diagnoses and classified them as heart failure (code 428) and coronary heart disease (CHD) (codes ), or other CVD hospitalizations including PAOD (codes 440, ), aortic aneurysm (code 441), cerebrovascular disease (codes ), and non-rheumatic valvular disease (code 424). All other codes included in the ICD-9- CM as diseases of the circulatory system were not included as valid end points. Through this linkage a total of 94 nonfatal cardiovascular events could be identified. Follow-up started after the baseline CT scan. Follow-up time differed for all-cause mortality and the other two end points, because of the differential availability of the two registries used. For all-cause mortality, follow-up was complete until January 1, 2007 (median: 21.5 months); for both other end points, until January 1, 2006 (median: 9.5 and 10.0 months). For all participants who experienced an event, follow-up ended at the date of diagnosis or death. In participants with multiple cardiovascular hospital admissions during follow-up, the first hospital discharge diagnosis was used as end point. Assessment of covariates At baseline, all participants from the NELSON Study were asked to return a questionnaire containing information on prior and current smoking behavior. For subjects in this casecohort substudy, a research physician collected information from their general practioners (GPs) using a standardized questionnaire. The obtained information included the current use of drugs, specifically the use of antihypertsenive drugs (defined as diuretics, ACE inhibitors, angiotensin II receptor antagonists, β-blockers and/or calcium channel blockers); lipidlowering drugs; oral hypoglycemic agents; and insulin; systolic and diastolic blood pressure (BP); and nonfasting blood glucose, HbA1c, total cholesterol, LDL cholesterol and HDL cholesterol levels. The overall response rate was 70%. For all covariates obtained through the GP, missing values were imputed using regression methods implemented in SPSS software (SPSS 14.0, Chicago, Illinois) 26. We defined diabetes mellitus as a nonfasting glucose level 11.1 mmol/l and/or the use of oral hypoglycemic agents or insulin. Hypertension was defined as a diastolic BP >90mmHg, systolic BP >140mmHg and/or the use antihypertensive drugs. Hypercholesterolaemia was defined as a total cholesterol level >5.0 mmol/l, an LDL level >3.0 mmol/l and/or the use of lipid-lowering drugs. 110

111 Comparing coronary artery calcium and thoracic aortic calcium for prediction of all-cause mortality Statistical methods Baseline characteristics were summarized for the subcohort and the three different casegroups separately. Categorical variables were compared with a χ 2 statistic; continuous variables with the Mann-Whitney U test. Unadjusted annualized event rates for all-cause mortality and CVD events were calculated per CAC and TAC quartile in the subcohort. In the subcohort, the association between continuous measures of TAC and CAC was investigated with Spearman s rank correlation. Prevalence of CAC and TAC across age groups was compared for men and women separately. Association between the distribution of TAC and CAC risk categories was investigated using a χ 2 test. Since no accepted thresholds exist on cut-offs for TAC risk categories, we chose to divide both TAC and CAC scores into quartiles to maximize comparability. The association of TAC and CAC with all-cause mortality and the composite CVD, CHD and non-coronary end point was evaluated with Cox proportional hazard analyses. To account for the case-cohort design, modification of the standard errors was based on robust variance estimates. We used the method according to Prentice in which all subcohort members are equally weighted. Cases outside the subcohort are not weighted before failure and at failure receive the same weight as members of the subcohort 14. This method has been shown to resemble most closely estimates from a full-cohort analysis 27. Cox proportional hazard analyses for all four end points were performed for increasing TAC and CAC categories and adjusted for cardiovascular risk factors (age, sex, current smoking, hypertension, diabetes and hypercholesterolaemia). The lowest quartiles of TAC and CAC scores were used as reference categories. In the test for trend analysis, quartiles of TAC and CAC were replaced with continuous TAC and CAC scores. Statistical analyses were performed with SPSS 14.0 software for Windows (SPSS Inc., Chicago, IL) and R 6.2. Results Table 1 shows the baseline characteristics of a representative baseline sample (subcohort) and all four event-groups. The subcohort included 808 subjects (671 men, 137 women; mean age, 60±6). Compared with the subcohort, subjects in all four event-groups were more often men (p<0.0001) and more were classified as having diabetes (7% versus 11-18%; (p=0.001). In subjects from the composite CVD, CHD and non-cardiac event-groups, hypertension was more frequent compared with the subcohort (p<0.0001). Subjects in all four event-groups had higher CAC and TAC scores (p<0.0001) compared with subjects from the subcohort. Highest median CAC scores were recorded in the CHD event-group, whereas highest median TAC scores were found for the all-cause mortality and non-cardiac event-groups (Table 1). 111

112 Chapter 7 Table 1 Baseline characteristics Variables a Subcohort All-cause mortality CVD end point CHD end point Non-CHD end point (n=808) (n=56) (n=127) (n=61) (n=38) Age, y 59.5± ± ± ± ±5.5 Men, % Hypertension, % Hypercholesterolaemia, % Diabetes, % Current smokers, % TAC score (AS) b 536(1837) 3411(6820) 1548(3358) 1236(2997) 2604(5375) CAC score (AS) b 74(591) 685(1828) 769(2063) 1055(2017) 268(1716) AS: Agatston score; CAC: Coronary artery calcium; SD: standard deviation; TAC: Thoracic aortic calcium. a Expressed as percentage or mean±sd. b Median (interquartile range) In the subcohort, cut-offs for quartiles of CAC score were 0-1 (1 st quartile), 1-74 (2 nd quartile), (3 rd quartile), and >592 (4 th quartile); cut-offs for TAC score quartiles were 0-99 (1 st quartile), (2 nd quartile), (3 rd quartile), and >1937 (4 th quartile). Seventyfour percent (n=599) of subjects had both TAC>0 and CAC>0 scores; 23% (n=177) had only TAC; 1% (n=12) had only CAC; and only 2% (n=20) of subjects had no detectable CAC or TAC. The distribution of TAC and CAC scores in risk categories were closely associated (p<0.0001) (Figure 2). A strong correlation existed between the continuous CAC and TAC scores (Spearman s r: 0.49, p<0.001). Figure 2 Distribution of quartiles of thoracic aortic calcium (TAC) score across quartiles of coronary artery calcium (CAC) score. Increasing levels of CAC scores are closely associated with increasing levels of TAC score (p<0.001). 112

113 Comparing coronary artery calcium and thoracic aortic calcium for prediction of all-cause mortality The prevalence of both CAC and TAC increases by age group in men and women (Figure 3A and 3B). A striking difference between sexes is the higher prevalence of CAC compared with TAC in men aged <65, whereas in women TAC is consistently more prevalent than CAC in all age categories. Throughout all age categories CAC is more prevalent in men compared with women, but TAC is more prevalent in women than in men. Figure 3 Prevalence of calcifications in the coronary arteries with an Agatston score > 1 st quartile (AS >1), and calcfications in the thoracic aorta with an Agatston score > 1 st quartile (AS>99) by age category. Panel A, for men (n=670). The χ 2 test indicated a trend across age categories for CAC (P<0.001) and TAC (P<0.001). Panel B, for women (n=138). The χ 2 test indicated a trend across age categories for CAC (P=0.001) and TAC (P=0.03). 113

114 Chapter 7 Unadjusted annualized event rates for all-cause mortality with increasing quartiles of CAC and TAC score were 0%, 0.28%, 1.17%, and 0.88% and 0%, 0.57%, 0.29%, and 1.54%, respectively. For total CVD events the corresponding event rates with increasing quartiles of CAC and TAC scores were 2.35%, 1.87%, 2.02%, and 4.11% and 2.11%, 2.56%, 1.99%, and 3.79%, respectively. During a median follow-up 21.5 months (range: days), 56 subjects died. Both CAC and TAC scores were associated with risk of all-cause mortality (p for trend=0.01 and 0.001, respectively) (Table 2). Compared with the first quartile of TAC, risk factor-adjusted hazard ratios (HR) for the second, third, and fourth quartile were 1.22, 2.12, and 5.45, respectively. Only in case of subjects in the highest quartile did this association reach statistical significance. Compared with risk factor-adjusted hazard ratios for quartiles of CAC (HR: 3.70, 5.75, and 9.13 for the second, third and fourth risk category, respectively), hazard ratios for TAC were consistently lower. During a median follow-up 9.5 months (range: days) and 10.0 months (range: days) incident fatal and non-fatal cardiovascular events occurred in 127 subjects. CAC and TAC again showed a graded association with the risk of fatal and non-fatal CVD events (p for trend <0.001 and 0.03, respectively). Risk-factor adjusted hazard ratios for the second, third, and fourth quartile compared with the first quartile of TAC were 0.87, 1.51, and 2.25, respectively (Table 2). Only in case of subjects in the highest quartile did this association reach statistical significance. As with all-cause mortality, risk factor-adjusted hazard ratios for CAC risk categories (HR: 1.74, 1.88, and 4.46 for the second, third and fourth category respectively) were consistently higher compared with the same hazard ratios for TAC. By breaking down the composite CVD end point into coronary cases (61 events) and noncardiac cases (including stroke, aortic aneurysm, and PAOD) (38 events), we performed a secondary analysis (Table 3) investigating the specific patterns of association between TAC/ CAC and cardiovascular events. Only CAC was found to be associated with coronary events. Compared with the first quartile of CAC, multivariate-adjusted hazard ratios for the second, third, and fourth quartile were 1.62, 2.12, and 6.86, respectively (p for trend <0.001 compared with p for trend=0.33 for TAC). Conversely, only TAC was associated with the composite end point of non-cardiac events. After adjustment for cardiovascular risk factors, hazard ratios for the second, third, and fourth quartile were 1.14, 1.39, and 4.69, respectively (p for trend=0.02). CAC was not associated with this end point (p for trend=0.40) (Table 3). 114

115 Comparing coronary artery calcium and thoracic aortic calcium for prediction of all-cause mortality Table 2 Multivariate-adjusted hazard ratios for all-cause mortality and cardiovascular (CVD) events according to quartiles of coronary artery calcium (CAC) and thoracic aortic calcium (TAC) Hazard Ratios (95% CI) Quartiles of CAC Quartiles of TAC All-cause mortality (n=56) Quartile of CAC (AS) Quartile of TAC (AS) 1 (0-1) (0-99) (1-74) 3.70 ( ) 2 ( ) 1.22 ( ) 3 (75-592) 5.75 ( ) 3 ( ) 2.12 ( ) 4 (>592) 9.13 ( ) 4 (>1937) 5.45 ( ) P for trend 0.01 P for trend CVD end point (n=127) Quartile of CAC (AS) Quartile of TAC (AS) 1 (0-1) (0-99) (1-74) 1.74 ( ) 2 ( ) 0.87 ( ) 3 (75-592) 1.88 ( ) 3 ( ) 1.51 ( ) 4 (>592) 4.46 ( ) 4 (>1937) 2.25 ( ) P for trend <0.001 P for trend 0.03 All models adjusted for age, sex, current smoking, hypertension, hypercholesterolaemia, and diabetes. AS: Agatston Score. Cut-off values of absolute Agatston scores for quartiles of CAC and TAC in brackets. Table 3 Multivariate-adjusted hazard ratios for coronary (CHD) and non-cardiac (Non-CHD) events according to quartiles of coronary artery calcium (CAC) and thoracic aorta calcium (TAC). Hazard Ratios (95% Cl) Quartiles of CAC Quartiles of TAC CHD end point (n=61) Quartile of CAC (AS) Quartile of TAC (AS) 1 (0-1) (0-99) (1-74) 1.62 ( ) 2 ( ) 0.74 ( ) 3 (75-592) 2.12 ( ) 3 ( ) 1.55 ( ) 4 (>592) 6.86 ( ) 4 (>1937) 1.48 ( ) P for trend <0.001 P for trend 0.33 Non-CHD end point (n=38) Quartile of CAC (AS) Quartile of TAC (AS) 1 (0-1) (0-99) (1-74) 2.56 ( ) 2 ( ) 1.14 ( ) 3 (75-592) 2.08 ( ) 3 ( ) 1.39 ( ) 4 (>592) 3.06 ( ) 4 (>1937) 4.69 ( ) P for trend 0.40 P for trend 0.01 All models adjusted for age, sex, current smoking, hypertension, hypercholesterolaemia, and diabetes. AS: Agatston Score. Cut-off values of absolute Agatston scores for quartiles of CAC and TAC in brackets. 115

116 Chapter 7 Discussion CAC has previously been found to be independently associated with all-cause mortality and cardiovascular events 5,6. In the present study, TAC was closely associated with CAC. Although both CAC and TAC were associated with all-cause mortality and CVD events in this population of heavy smokers, risk factor-adjusted hazard ratios were consistently higher for CAC compared with TAC. Furthermore, only CAC was associated with coronary events, whereas TAC and not CAC was found to be associated with non-cardiac events (stroke, aortic aneurysm, PAOD). A number of previous studies reported on the positive association of TAC and CAC In these studies a wide array of different measurement techniques for detecting calcified plaques in the thoracic aorta has generally been used, ranging form plain radiography to ECG-gated multislice CT. Furthermore, these studies were conducted in very different populations, ranging from symptomatic CVD patients to population-based cohorts. Also, the part of the aorta evaluated for plaques ranged from the parts of the descending segment depicted on a cardiac CT scan to the full stretch of the intrathoracic aorta. Therefore, a direct comparison between the results reported by these studies should be undertaken with caution. Our study was performed within the NELSON study, a large randomized controlled population-based trial in subjects aged years. We used low-dose, non-gated CT to simultaneously detect plaques in the coronary arteries and all three segments of the thoracic aorta. This technique has shown to be valid for detecting CAC with an accuracy of up to 90% compared with dedicated cardiac CT 28. Furthermore, since motion artifacts do occasionally cause blurring of calcified plaques and consequently intra-individual variability of CAC scores, we grouped all absolute CAC and TAC scores into quartiles and all analyses for this study were performed on a quartile-by-quartile basis thus reducing the impact of this effect on our results. These observations all contribute to the overall higher prevalence of TAC in our study compared with two previously published population-based CT cohorts 9,10. However, in a number of key aspects our results are consistent with earlier results. In all studies TAC was more prevalent than CAC among women of all ages, whereas CAC occurred more than TAC in men under age 65. It has been suggested that this might result from the interplay between osteoporotic and atherosclerotic pathophysiological mechanisms 29,30. Our study likewise demonstrated a more pronounced increase in TAC with increasing age compared with CAC. To our knowledge, only two studies have been reported on the association between TAC and all-cause mortality or cardiovascular events. In a study among 361 stable angina pectoris patients (mean age, 62) hazard ratios of 2.84 and 2.70 were demonstrated for the risk of total events (defined as fatal/nonfatal CVD events, non-cardiac fatalities, and peripheral revascularization) and CVD events in patients with TAC>0 compared with patients with TAC=0 12. However, this is a smaller study in symptomatic angina patients, and since TAC was 116

117 Comparing coronary artery calcium and thoracic aortic calcium for prediction of all-cause mortality only visually assessed as present/absent only very crude estimates could be presented. To be useful in clinical practice, screening for TAC should improve existing risk stratification tools, and should consequently be targeted at asymptomatic subjects since symptomatic CVD patients already fall by the very nature of their condition in the highest risk category. Furthermore, TAC is shown to be very prevalent and small quantities of TAC are not necessarily associated with an increased risk of adverse events; so the choice of any TAC as a cut-off point is probably not the best. Using quartiles of semi-automatically scored TAC, we have demonstrated a graded association of TAC with all-cause mortality and CVD events, providing a more accurate tool for risk stratification. In another study, a direct comparison between CAC and TAC showed a strong relation between CAC risk categories and CVD/CHD events in physician referred or self-referred patients, but no such relation was found for TAC 12. Although results in our study similarly show that association with CVD and CHD events is stronger for CAC than for TAC, our results do show an association between TAC and CVD events. This may be caused by a higher prevalence of TAC in our study. The higher prevalence of TAC can in part be explained by our population (heavy smokers), but more likely will be caused by differences in the CT protocols used. In our low-dose chest CT protocol, we evaluated the whole intrathoracic aorta for presence of TAC in stead of only those parts of the aorta visible on a dedicated cardiac CT protocol. Excluding TAC plaques in the aortic arch is likely to influence the association with incident stroke cases which are an important part of the composite CVD end point. We therefore feel that the distribution of subjects in their study would have been radically different had they included all TAC plaques and, consequently, would have resulted in different associations. Furthermore, the authors have applied the same cut-off values for TAC risk categories as for CAC risk categories since there are no accepted thresholds for TAC score at present. In our opinion, the caliber of the thoracic aorta is incomparable with the combined caliber of the coronary arteries making a one-on-one comparison using the same cut-off values for both scores inappropriate. In the absence of well-established cut-offs we feel that using quartiles of CAC/TAC is the best option when comparing them. Choosing quartiles as cut-off, is also the reason why event/hazard rates for CAC quartiles from this study should not be compared one-on-one with event/hazard rates form previous studies. Furthermore, the lack of a stepwise increase in event rates for the combined CVD end point can be explained by the fact that event rates unlike hazard rates were presented unadjusted for traditional CVD risk factors. Given the high prevalence of these risk factors among our population even participants with low or intermediate CAC/TAC scores had unadjusted event rates of ~2% per year. Although hampered by a limited number of events, we found in our secondary analysis that TAC was associated with non-cardiac events, but not with coronary events. This pattern 117

118 Chapter 7 is completely reversed in the case of CAC, being more strongly associated with coronary events and showing no association with non-cardiac events. Based on these observations, we hypothesize that TAC - although at a certain level strongly correlated with the extent of CAC - is perhaps more than CAC a measure of generalized atherosclerosis and therefore possibly better suited as a predictor of non-cardiac CVD events. Longer follow-up with more noncardiac CVD events are needed to confirm this hypothesis. From a clinical perspective, imaging of atherosclerotic disease has generally been described as having added value in a population at intermediate risk of future CVD events 31. Therefore, a possible limitation of this study could be the high-risk profile of this population of heavy smokers. For this reason, we calculated the Framingham Risk Score 32 in this population and found that approximately 50% is at intermediate risk (10-year risk 10%-20%) and only 35% at high risk (10-year risk >20%) which emphasize the potential clinical implications of CAC/ TAC imaging as part of lung cancer screening. Limited number of events caused some of the associations not to reach statistical significance. Taking into account the consistent and statistically significant trends across quartiles, we believe that with more events these associations would have become statistically significant. Furthermore, baseline covariate information was obtained through the GPs and was not 100% complete. A response rate of 70%, however, should not introduce a large bias and imputation of missing covariate information has been shown superior to a complete case analysis 26,33. Finally, exclusion of subjects with a history of cardiovascular disease was performed through database linkage. Since this information was unavailable prior to 1995, this could have resulted in false inclusion of a number of subjects with a positive history of CVD. We believe, however, this number to be small - taking into account the mean age of our population prior to and think this will not have influenced the results of this study. In conclusion, our results provide further evidence that TAC and CAC are closely associated markers of cardiovascular disease. This study in a high-risk population of heavy smokers has demonstrated that CAC is a stronger predictor than TAC for all-cause mortality, CVD events, and coronary events. However, TAC appears to be a stronger predictor for non-cardiac CVD events (stroke, aortic aneurysm, PAOD). Future studies with more non-cardiac events are needed to gain a more profound insight into these effects. 118

119 Comparing coronary artery calcium and thoracic aortic calcium for prediction of all-cause mortality References 1. Agmon Y, Khandheria BK, Meissner I et al. Relation of coronary artery disease and cerebrovascular disease with atherosclerosis of the thoracic aorta in the general population. Am J Cardiol. 2002;89: Fazio GP, Redberg RF, Winslow T, Schiller NB. Transesophageal echocardiographically detected atherosclerotic aortic plaque is a marker for coronary artery disease. J Am Coll Cardiol. 1993;21: Iribarren C, Sidney S, Sternfeld B, Browner WS. Calcification of the aortic arch: risk factors and association with coronary heart disease, stroke, and peripheral vascular disease. JAMA. 2000;283: Witteman JC, Kok FJ, van Saase JL, Valkenburg HA. Aortic calcification as a predictor of cardiovascular mortality. Lancet. 1986;2: Budoff MJ, Shaw LJ, Liu ST et al. Long-term prognosis associated with coronary calcification: observations from a registry of 25,253 patients. J Am Coll Cardiol. 2007;49: Shaw LJ, Raggi P, Schisterman E, Berman DS, Callister TQ. Prognostic value of cardiac risk factors and coronary artery calcium screening for all-cause mortality. Radiology. 2003;228: Adler Y, Fisman EZ, Shemesh J et al. Spiral computed tomography evidence of close correlation between coronary and thoracic aorta calcifications. Atherosclerosis. 2004;176: Oei HH, Vliegenthart R, Hak AE et al. The association between coronary calcification assessed by electron beam computed tomography and measures of extracoronary atherosclerosis: the Rotterdam Coronary Calcification Study. J Am Coll Cardiol. 2002;39: Takasu J, Budoff MJ, O Brien KD et al. Relationship between coronary artery and descending thoracic aortic calcification as detected by computed tomography: The Multi-Ethnic Study of Atherosclerosis. Atherosclerosis [Epub ahead of print]. 10. Wong ND, Sciammarella M, Arad Y et al. Relation of thoracic aortic and aortic valve calcium to coronary artery calcium and risk assessment. Am J Cardiol. 2003;92: Rivera JJ, Nasir K, Katz R et al. Relationship of thoracic aortic calcium to coronary calcium and its progression (from the Multi-Ethnic Study of Atherosclerosis [MESA]). Am J Cardiol. 2009;103: Eisen A, Tenenbaum A, Koren-Morag N et al. Calcification of the thoracic aorta as detected by spiral computed tomography among stable angina pectoris patients: association with cardiovascular events and death. Circulation. 2008;118: Wong N, Gransar H, Shaw LJ et al. Thoracic Aortic Calcium Versus Coronary Artery Calcium for the Prediction of Coronary Heart Disease and Cardiovascular Disease Events. J Am Coll Cardiol Img. 9 A.D.;2: van Iersel CA, de Koning HJ, Draisma G et al. Risk-based selection from the general population in a screening trial: selection criteria, recruitment and power for the Dutch-Belgian randomised lung cancer multi-slice CT screening trial (NELSON). Int J Cancer. 2007;120: Prentice RL. A case-cohort design for epidemiologic cohort studies and disease prevention trials. Biometrika. 1986;73: Detrano RC, Anderson M, Nelson J et al. Coronary calcium measurements: effect of CT scanner type and calcium measure on rescan reproducibility--mesa study. Radiology. 2005;236:

120 Chapter Isgum I, Rutten A, Prokop M, van Ginneken B. Detection of coronary calcifications from computed tomography scans for automated risk assessment of coronary artery disease. Med Phys. 2007;34: Shemesh J, Evron R, Koren-Morag N et al. Coronary artery calcium measurement with multi-detector row CT and low radiation dose: comparison between 55 and 165 mas. Radiology. 2005;236: Jain AK. Fundamentals of digital image processing. Englewood Cliffs, NJ: Prentice-Hall, Inc; Agatston AS, Janowitz WR, Hildner FJ, Zusmer NR, Viamonte M, Jr., Detrano R. Quantification of coronary artery calcium using ultrafast computed tomography. J Am Coll Cardiol. 1990;15: Ulzheimer S, Kalender WA. Assessment of calcium scoring performance in cardiac computed tomography. Eur Radiol. 2003;13: De Bruin A, Kardaun JW, Gast A, Bruin E, van Sijl M, Verweij G. Record linkage of hospital discharge register with population register: experiences at Statistics Netherlands. Stat J UN Econ Comm Eur. 2004;21: Herings RM, Bakker A, Stricker BH, Nap G. Pharmaco-morbidity linkage: a feasibility study comparing morbidity in two pharmacy based exposure cohorts. J Epidemiol Community Health. 1992;46: Paas GR and Veenhuizen KC. Research on the validity of the LMR. Utrecht; Prismant: Reitsma JB, Kardaun JW, Gevers E, de Bruin A, van der WJ, Bonsel GJ. [Possibilities for anonymous follow-up studies of patients in Dutch national medical registrations using the Municipal Population Register: a pilot study]. Ned Tijdschr Geneeskd. 2003;147: Little RJA. Regression with missing X s: a review. J Am Stat Assoc. 1992;87: Onland-Moret NC, van der A D, van der Schouw YT et al. Analysis of case-cohort data: a comparison of different methods. J Clin Epidemiol. 2007;60: Kim SM, Chung MJ, Lee KS, Choe YH, Yi CA, Choe BK. Coronary calcium screening using low-dose lung cancer screening: effectiveness of MDCT with retrospective reconstruction. AJR Am J Roentgenol. 2008;190: Carr JJ, Register TC, Hsu FC et al. Calcified atherosclerotic plaque and bone mineral density in type 2 diabetes: the diabetes heart study. Bone. 2008;42: Dhore CR, Cleutjens JP, Lutgens E et al. Differential expression of bone matrix regulatory proteins in human atherosclerotic plaques. Arterioscler Thromb Vasc Biol. 2001;21: Greenland P, LaBree L, Azen SP, Doherty TM, Detrano RC. Coronary artery calcium score combined with Framingham score for risk prediction in asymptomatic individuals. JAMA. 2004;291: Wilson PW, D Agostino RB, Levy D, Belanger AM, Silbershatz H, Kannel WB. Prediction of coronary heart disease using risk factor categories. Circulation. 1998;97: Rubin DB, Schenker N. Multiple imputation in health-care databases: an overview and some applications. Stat Med. 1991;10:

121

122

123 CHAPTER 8 Coronary artery calcium screening as part of low-dose computed tomography screening for lung cancer improves cardiovascular risk prediction in men

124 Chapter 8 Abstract Context Tobacco use causes both lung cancer and cardiovascular disease. Early detection of both diseases is possible using low-dose computed tomography (CT). Objective To determine whether coronary artery calcium (CAC) scoring on low-dose CT as part of a lung cancer screening trial improves cardiovascular risk prediction beyond a risk prediction model containing traditional risk factors only. Design, Setting, and Participants This study was performed within male participants without a history of cardiovascular disease (CVD) from the screen group of the NELSON study, a randomized controlled lung cancer screening trial in subjects 50 years or older and with a history of at least 15 pack-years of smoking. All subjects were followed up for a median of 10.1 months (range: ) (January 2006) for incident fatal and nonfatal CVD events determined through database linkage with the national death registry and national registry of hospital discharge diagnoses. Using a casecohort design, CAC scoring was performed and risk factors were ascertained in a random sample (n=637) of male participants as well as in all incident cases (n=120). Main Outcome Measure We compared risk prediction for CVD events using a model containing traditional risk factors vs a traditional model plus CAC. Results Risk prediction improved after adding CAC to the traditional risk factor model (model discrimination: C index 0.68 vs. 0.73, P<0.001). Based on the model with only traditional risk factors, 50% of subjects would be classified at intermediate risk (10-year risk 5%-20%). Adding CAC to this model, reclassifies 33% of subjects at intermediate risk into higher- or lower-risk categories with improved accuracy (clinical net reclassification index (CNRI) 43.4%, P<0.001). To demonstrate the possible gains in terms of optimal targeting of primary prevention, we found that 19% of subjects at high-risk did not receive any type of preventive risk factor treatment. 124

125 Coronary artery calcium screening as part of low-dose computed tomography screening for lung cancer Conclusion This study demonstrates that risk prediction for CVD events improves with the addition of a CAC as scored on low-dose CT - to a traditional model, reclassifying 33% of subjects at intermediate risk into higher- or lower-risk categories. Implementing this strategy could benefit 1 in 12 subjects screened in terms of improved detection and optimized targeting of preventive risk factor treatment. 125

126 Chapter 8 Introduction Throughout the world several clinical trials are currently being undertaken among heavy (former) smokers to investigate the effectiveness of screening for lung cancer with lowdose computed tomography (CT) 1-6. Since smoking plays an important causal role in the development of lung cancer and cardiovascular disease, adding the coronary artery calcium (CAC) score to baseline screening for lung cancer has been suggested 7. This could be a costeffective way to improve detection of people at high risk for cardiovascular events in this vulnerable population, that were previously not recognized as such by their GPs, and to initiate preventive therapies in these patients. Moreover, in this highly selected population of heavy smokers the addition of CAC to existing global risk scores could improve model discrimination and, accordingly, result in a more accurate targeting of preventive therapies. This is because existing global risk scores, such as the Framingham Risk Score (FRS) 8,9, do not always allow proper discrimination between those who will or will not experience an event when used in different types of population 10. Especially in a highly selected population of heavy smokers one would not expect such a global risk prediction model to perform optimally. Although CAC is independently associated with future cardiovascular events, existing global risk prediction models do not include CAC as variable in their model. Adding the CAC score to risk prediction models with well-established cardiovascular risk factors could result in an improved cardiovascular risk prediction model for use in this population. We collected information on traditional cardiovascular risk factors and measured the CAC score at baseline low-dose CT in a cohort of 743 initially asymptomatic male participants of a lung cancer CT screening trial followed up for a median period of 10.1 months for the development of fatal and nonfatal cardiovascular events. We compared the accuracy of risk prediction between two models consisting of traditional cardiovascular risk factors with and without the CAC score in terms of model calibration, discrimination and reclassification. We also quantified the number of participants at high risk not receiving any type of preventive therapy at baseline CT to investigate the possible benefits from instituting this type of screening. Methods Design and study cohort The study population consisted of male participants from the NELSON Study, a randomized controlled population-based lung cancer screening trial in subjects 50 years and older and with at least 15 pack years of smoking 6. Women were excluded from this substudy given the small percentage of women in the trial (17%). From 2004 to 2006, baseline CT examinations were performed in all 7557 participants in the screen group of this trial. We used a case-cohort design using all available incident fatal and nonfatal CVD cases at the time (n=120), compared 126

127 Coronary artery calcium screening as part of low-dose computed tomography screening for lung cancer with a random sample (subcohort, n=637) of the NELSON screen group at baseline. Sampling was restricted to those men for which a baseline CT scan could be retrieved. Subjects with a history of cardiovascular disease were excluded. Ascertainment of a history of cardiovascular disease was performed through linkage with the national registry of hospital discharge diagnoses. All subjects with a cardiovascular discharge diagnosis prior to the start of this study (January 2004) were excluded (ICD-9-CM codes (ischaemic heart disease), 424 (nonrheumatic valvular disease), 428 (heart failure), (cerebrovascular disease), 440, 441, 443 and 444 (peripheral arterial disease). Participants gave written informed consent. The Medical Ethics Committees of all four participating hospitals approved this study. Data collection and CT measurements Information on patient characteristics (age; sex; smoking) was obtained through patient questionnaires at the start of the study. Traditional risk factors for cardiovascular events (blood pressure; total cholesterol, LDL and HDL cholesterol levels; nonfasting glucose levels; and information on the use of antihypertensive, lipid-lowering, and oral hypoglycemic medication, and insulin) were obtained by sending questionnaires to their general practitioners (GPs). The overall response rate was 70%. For all risk factors obtained through the GP, missing values were imputed using regression methods implemented in SPSS software (SPSS 14.0, Chicago, Illinois) 11. We defined diabetes mellitus as a nonfasting glucose level 200 mg/dl and/or the use of oral hypoglycemic agents or insulin. Hypertension was defined as a diastolic BP >90 mmhg, systolic BP >140 mmhg and/or the use antihypertensive drugs. Hypercholesterolaemia was defined as a total cholesterol level >200 mg/dl, an LDL level >115mg/dL and/or the use of lipid-lowering drugs. Baseline CT scans were performed using a low-dose, non-gated acquisition protocol reconstructed to 3.1mm thick slices (for details see 12 ), which has a high accuracy 13 compared with dedicated cardiac CT. Two observers blinded to all other participant data performed calcium scoring using software described in Isgum and Rutten 14,15, and calcium scores were calculated using the Agatston scoring algorithm as outlined in 16. Follow-up procedures and end point definition A composite CVD end point consisting of all cardiovascular deaths and all nonfatal cardiovascular hospital admissions was chosen as the primary end point for this study. To retrieve information on cardiovascular deaths and hospital admissions, all participants in the screen group of the NELSON Study (n=7557) were linked with the national death registry (for the years ) and the national registry of hospital discharge diagnoses (for the years ). Information on discharge diagnoses for the year 2006 was not yet available. This database linkage was performed on the basis of birth date, sex and postal code with a validated probabilistic method

128 Chapter 8 In these registries diagnoses are coded according to the International Classification of Diseases (ICD), 9 th (discharge diagnoses) and 10 th (cause of death) revision. One research physician selected all cases of cardiovascular death and cardiovascular discharge diagnoses and classified them as coronary heart disease (CHD) (ICD-9 codes ), peripheral arterial disease (codes 440, ), aortic aneurysm (code 441), cerebrovascular disease (codes ), heart failure (code 428) and non-rheumatic valvular disease (code 424). All other codes included in the ICD-9 as diseases of the circulatory system were not included as valid end points. Follow-up started after the baseline CT scan and lasted up to 21 months (median, 10.1 months). For all participants who experienced an event, follow-up ended at the date of diagnosis or death. In participants with multiple cardiovascular hospital admissions during follow-up, the first hospital discharge diagnosis was used as end point. Prediction models Keeping close to existing risk prediction models for the development of any CVD event, a prespecified model (model 1), consisting of all covariates used in the most recently updated Framingham Risk Score (FRS) for use in primary care 8, was fit in all subjects using Cox proportional hazard analysis. We accounted for the case-cohort design by modification of the standard errors using the method by Prentice 21. We chose to adapt the model taking into account not only the use of antihypertensive drugs, but also of lipid-lowering drugs by including hypertension and hypercholesterolaemia as dichotomous variables. For example, in the FRS, natural log of total cholesterol is included without reference to subjects using lipid-lowering drugs. In our model 1, this coefficient is substituted by a dichotomous variable for hypercholesterolaemia defined as a total cholesterol level >200 mg/dl, an LDL level >115mg/dL and/or the use of lipid-lowering drugs. A comparable variable for hypertension was included in model 1. No stepwise selection procedures were used for variable selection since all these covariates are well-established risk factors for cardiovascular disease 22,23. As in the FRS, continuous variables (age, HDL-C, and CAC) were transformed using the natural log. Statistical analysis To assess the added value of CAC, global model fit, and model discrimination, calibration and reclassification were compared for model 1 (including age, hypertension, HDL-C, hypercholesterolaemia, smoking status, and diabetes) and model 2 (model 1 plus CAC). To evaluate global model fit, model χ 2 statistics were directly compared with a likelihood ratio test. Predictive accuracy of both models was compared by checking their discrimination and calibration. Discrimination refers to the ability of the model to distinguish between patients with and without a CVD event. Discriminative ability was evaluated using the C index 24. A higher value indicates a higher probability that a CVD event is correctly assigned to those subjects actually having experienced a CVD event compared with those who have not. 128

129 Coronary artery calcium screening as part of low-dose computed tomography screening for lung cancer Calibration refers to the agreement between the predicted risks and the observed frequencies. As a measure of model calibration, observed and predicted probabilities were compared in a calibration table. To correct for overfitting, the regression coefficients were shrunk using bootstrapping techniques 24,25. To appraise the improvement in risk prediction when adding CAC to an existing model in a clinically relevant way, we constructed event-specific reclassification tables 26. All subjects were classified into simplified 1-year predicted risk categories of <0.5%, 0.5-<1.0%, 1.0- <2.0%, and 2.0% (based on 10-year risk cut-offs of 5%, 10% and 20% divided by 10 27,28 ). We then calculated the proportion of subjects reclassified into higher- or lower-risk categories using model 2 instead of model 1. These cross tabulations were performed separately for subjects with and without events, since only upward movement of subjects with events and downward movement for subjects without events can be considered as improvement. This improvement can be quantified into the net reclassification index (NRI) and tested for significance 26. An adaptation of the NRI, the clinical NRI (CNRI), is a measure reflecting the correct upward/downward movement of subjects at intermediate risk since for these subjects an upward or downward shift in risk category will have clinical implications 29. To address whether the simplified assumption of a linear development in absolute risk over a 10-year period, and consequently the choice of cut-off points in the reclassification table, had any impact on our findings, we performed a sensitivity analysis using the mathematically exact cut-off points in the reclassification table (<0.53, 0.53-<1.05, 1.05-<2.21, 2.21). The exact cut-off points were calculated using the formula: S(1-year)= S(10-year)^(1/10) (S=survival probability). Finally, we included a nomogram based on regression coefficients of model 2 after shrinkage in order to calculate a patient s individual risk category. All analyses were conducted with SPSS (SPSS for Windows, version 14.0, Chicago, Ill) and R statistical software version 6.2. Results As a random sample of the full cohort, baseline characteristics of all men in the subcohort are shown in Table 1. During a total follow-up of 21 months (median: 10.1 months), 120 subjects experienced either a nonfatal cardiovascular event (n=112) or CVD death (n=8). Prediction Models The prespecified model (model 1) included 6 predictors that are well-established cardiovascular risk factors; age, hypercholesterolaemia, HDL-C, hypertension, diabetes, and current smoking. The regression coefficients, standard errors and confidence intervals from the multivariate Cox regression model for these 6 predictors with (model 2) and without (model 1) CAC are presented in Table

130 Chapter 8 Table 1 Baseline Characteristics of Subcohort Risk Factors Subcohort (n = 637) Age, (mean) (SD), y 59 (5) Cholesterol, mean (SD), mg/dl Total 203 (44) LDL-C (39.5) HDL-C 50.9 (19.4) Lipid-lowering therapy, No. (%) 206 (32) Blood pressure, mean (SD), mm Hg Systolic 137 (19) Diastolic 81 (10) Antihypertensive therapy, No. (%) 255 (40) Smoking status, No. (%) Current Past 368 (58) 269 (42) Diabetes, No. (%) 54 (9) CAC score, median (IQR) 95 (652) CAC, coronary artery calcium; HDL-C, high-density lipoprotein cholesterol; IQR, interquartile range; LDL-C, low-density lipoprotein cholesterol; SD, standard deviation. SI conversion factors: to convert cholesterol from mg/dl to mmol/l, multiply by Table 2 Multivariate Cox Regression Analyses Predicting Fatal and Nonfatal Cardiovascular Events in Model 1 Based on Covariates From the Framingham Risk Score and Model 2 Based on Model 1 plus Coronary Artery Calcium Scores Model 1 Model 2 ß (SE) 95% CI ß (SE) 95% CI Natural logarithm Age (1.155) 0.540, (1.247) , Hypercholesterolaemia (0.262) , (0.270) , Natural logarithm HDL-C (0.324) , (0.342) , Hypertension (0.289) 0.643, (0.294) 0.548, Diabetes (0.317) , (0.330) , Current smoking (0.220) , (0.226) , Natural logarithm CAC (0.048) 0.119, CAC, coronary artery calcium; CI, confidence interval; CVD, cardiovascular disease; HDL-C, high-density lipoprotein cholesterol; SE, standard error. All regression coefficients are corrected for overfitting by bootstrap methods. Shrinkage factor model 1: 0.91; Shrinkage factor model 2: 0.94) Table 3 summarizes two statistics of model performance comparing global model fit and model discrimination between model 1 and model 2. The χ 2 value for model 2 (χ 2 =94.6, df=6) was larger than a model based on covariates in the Framingham Risk Score (model 1) (χ 2 =63.5, df=5) suggesting a better overall model fit (P value <.001). In terms of discrimination, 130

131 Coronary artery calcium screening as part of low-dose computed tomography screening for lung cancer model 2 had a larger C index (0.73, 95% CI ) than model 1 (0.68, 95% CI ) (P value <.001) indicating better discrimination when CAC was added to a model based on Framingham covariates. Table 4 is a calibration table demonstrating good model calibration for the model including CAC with increasing frequency of observed events in increasing predicted risk categories. Table 3 Summary Statistics of Model Performance Comparing Model Fit and Discrimination of CVD Events in a Model Based on Covariates From the Framingham Risk Score (Model 1) and a Model Based on Framingham Covariates Plus Coronary Artery Calcium Scores (Model 2) Model 1 Model 2 P Value* Global fit Model χ <.001 Discrimination C index (95% CI) 0.68 ( ) 0.73 ( ) <.001 Abbreviations: CI, confidence interval; CVD, cardiovascular disease. * Likelihood ratio test for model χ 2. Higher C index indicates better discrimination. Table 4 Calibration Table Presenting Number of Subjects With and Without Observed Events According to Predicted Risk Categories Based on a Model of Framingham Risk Score Covariates Plus Coronary Artery Calcium (Model 2) Predicted Risk Categories * N full cohort Observed Events (%) <0.5% (0.4%) 0.5%-<1.0% (0.6%) 1.5%-< (1.2%) (3.5%) * Number of subjects in full cohort (n=7557) calculated by multiplying the N subcohort (n=623) with the inverse sampling fraction (1/α) per risk category (α= ). Reclassification As has been put forward recently, the added value of a new predictor can sometimes not be deduced from a change in C index 26. Adapting an existing risk prediction model by adding a new predictor leads to reclassification of absolute risk in individual subjects, and it is this shift of subjects across risk categories which is arguably the most important aspect when evaluating the performance of a new prediction model. Ideally, all subjects with a CVD event should be reclassified upward and subjects without an event downward so that targeting of preventive risk factor treatment can be optimized. This is what can be termed correct reclassification. Table 5 shows the distribution of subjects with and without events initially classified to 1-year CVD risk categories of <0.5%, 0.5-<1.0%, 1.0-<2.0%, and 2.0% based on 131

132 Chapter 8 Framingham covariates (model 1) that are being reclassified to higher- or lower risk categories after adding CAC (model 2). Model 2 reclassified a total of 44.1% (328/743) of all subjects; taking events status of individual subjects into account, 63.4% (208/328) of reclassified subjects were reclassified correctly. The NRI calculated from Table 4 was 18.1% (P value <.001). Focusing on subjects shifting from intermediate risk categories ( <2.0%) estimated from model 1 to low (<0.5%) or high ( 2.0) risk categories after adding CAC, a total of 33.7% (118/350) of all subjects was reclassified; of which 65.2% (77/118) were reclassified correctly. The corresponding CNRI was 43.4% (P value <.001). To address the impact of choice of cut-off values for our results, we repeated the above analyses using exact cut-off values for 1-year risks (generating risk groups of <0.53%, 0.53%- 1.05%, 1.05%-2.21%, and 2.21%). In this example, 43.9% (326/743) of subjects were reclassified into higher or lower-risk categories with comparable accuracy as in our primary analysis (NRI=20.0%, CNRI=45.6%, P value <.001). (Reclassification table not presented). Table 5 Risk Reclassification Among Subjects With CVD Events and Without CVD Events Comparing a Model Based on Traditional Framingham Risk Factors (Model 1) to the Same Model With CAC Added 1-Year Risks from Model 1 + CAC 1-Year Risk Categories (Model 1), % <0.5% <1.0% <2.0% 2.0% Total Correct Reclassified* No. of subjects with CVD Event <0.5% / <1.0% / <2.0% /23 2.0% /12 Total No. of subjects without CVD Event <0.5% / <1.0% / <2.0% / % /67 Total Abbreviations: CAC, coronary artery calcium; CVD, cardiovascular disease. * Number of subjects reclassified correctly/total number of subjects reclassified. Subjects not reclassified after adding CAC appear in bold. Translating these Results to Clinical Practice Based on the estimates from the improved risk prediction model where we added CAC to traditional risk factors (Table 5), 19% (51/269) of subjects at high risk (>2%) for future cardiovascular events did not receive any type of preventive therapy (defined as either 132

133 Coronary artery calcium screening as part of low-dose computed tomography screening for lung cancer antihypertensive drugs or lipid-lowering drugs); furthermore, 50% (169/338) of subjects at intermediate risk ( <2.0%) did not receive treatment with these drugs. Various national guidelines promote the institution of preventive therapy even for subjects at intermediate risk 27,28. Figure 1 summarizes the possible benefits of performing CAC scoring in a given population of 900 males screened with low-dose CT. Of all participants 50% (n=450) would be at intermediate risk for future CVD events using a prediction model consisting of traditional risk factors. Adding CAC would reclassify one third (n=150) of participants at intermediate risk of which 50% (n=75) would be reclassified into the high risk category. Moreover, of all 390 participants assigned to the high risk category (either by the traditional model or by the new model), 19% (n=74) of participants did not receive preventive risk factor treatment. This amounts to 1 in 12 male participants screened that could possibly benefit from this strategy of improved detection and classification (Figure 1). 133

134 Chapter 8 In Figure 2 we present a nomogram based on regression coefficients of the model with CAC that can be used in clinical practice. The bottom line indicates a subject s risk category. Discussion In this study conducted among 743 asymptomatic male participants of a lung cancer CT screening trial with a median follow-up of 10.1 months we found that adding CAC scoring to baseline low-dose CT resulted in improved risk prediction of future cardiovascular events compared with a risk prediction model based on traditional CVD risk factors alone. Currently, approximately 50% of subjects in this screening population would be classified into the intermediate risk category. A risk prediction model including traditional risk factors (age, current smoking, diabetes, hypercholesterolaemia, HDL-C, and hypertension) plus CAC reclassifies one third of these subjects at intermediate risk into higher- or lower-risk categories. 134

135 Coronary artery calcium screening as part of low-dose computed tomography screening for lung cancer Taking a typical cut-off point of a 1-year risk estimate of 2% or higher for starting preventive treatment of risk factors, reclassification by using this algorithm would increase the number of people receiving preventive therapy by 8%. Apart from improved risk prediction compared with a traditional model, we believe that performing CAC scoring on baseline low-dose CT in participants of a lung cancer screening study is a very elegant way of detecting people at high-risk for future CVD events. Our results show that approximately 20% of subjects in the high-risk category were previously not recognized as such by their health care providers and, consequently, did not receive any preventive treatment of cardiovascular risk factors (ie, no treatment with antihypertensive or lipid-lowering drugs). Together, improved detection and reclassification will result in optimized targeting of preventive therapy to those subjects that will benefit most from risk factor treatment. These results provide more support to the notion that CAC has clinically important added value in risk prediction beyond traditional risk factors alone. Several previous studies have demonstrated this prognostic value for future coronary events in asymptomatic populationbased cohorts Our report is unique in several ways compared with these previous studies. These reports recruited their subjects from population-based cohorts in order to demonstrate the possible benefits of CAC screening in large groups of asymptomatic low-to-intermediate risk people. The added value of CAC in a selected population with a higher risk profile (ie, heavy (former) smokers) has been questioned 37 since these people are generally considered to gain little form reclassification when adding a CAC score to a risk that is already high when estimated with traditional risk factors. In contrast, our results prove that risk prediction can be improved by adding CAC in high-risk population. Furthermore, scoring of CAC in our study is performed with a low-dose CT protocol ( msv) minimizing the radiation dose that typically affects people who are subjected to CAC screening with standard electronbeam CT ( msv) or multidetector-row CT ( msv) protocols 38. Successful dose reduction in cardiac CT has been reported in several studies Briefly, low-dose CT is able to correctly stratify people into appropriate CVD risk categories at the expense of precision in the absolute CAC score 13. This study is the first to demonstrate that improved risk prediction with CAC is possible using a low-dose, non-gated CT protocol. If lung cancer screening programs will be implemented, CAC scores at baseline screening can be obtained without having to expose participants to any additional radiation dose. Another important strength of this study is that it includes noncoronary CVD end points. Earlier reports on the prognostic value of CAC typically are limited to coronary events 34-36,43. From an etiologic point of view, coronary heart disease is one of many atherosclerotic cardiovascular disease entities that are all interrelated and responsive to the same regimen of preventive treatment. We therefore adhere to the idea that a new algorithm should focus on predicting any major CVD event 8,

136 Chapter 8 Despite these strengths, limitations of this study need to be discussed. In the present report we have limited our analyses to male participants of lung cancer screening, thus possibly reducing the generalizability of our results to female smokers. However, CAC has previously been shown to be a comparable predictor in both sexes 45 and all traditional risk factors included in the model are a standard part of existing risk prediction algorithms for men and women 9,46,47. Furthermore, our data on cardiovascular risk factors were obtained through GP questionnaires. However, we believe that in the absence of direct measurements more accurate information on these risk factors can be obtained when provided by health care professionals than by patient self-report. A response rate by GPs of 70% may be a limitation, yet we did not observe any differential pattern of missing values for cases and non-cases. Imputation of missing covariates has been shown superior to a complete case analysis 11,48. This study has limited follow-up. However, we believe that we present robust estimates given the substantial number of events that have been observed in this population in even a short follow-up period. Previous studies on the prognostic value of CAC reporting longer follow-up times have fewer events 33,34,36,43. We chose simplified risk cut-offs for 1-year risk by dividing accepted 10-year risk thresholds by ten. We note that this assumption has been made previously including international guidelines on primary prevention of CVD events 27,28. Results from our sensitivity analysis - using exact cut-off values for 1-year predicted risk show comparable estimates as in our primary analysis. Finally, external model validation in new patients needs to be performed before a prediction model can be used in clinical practice. Also it should be noted that these results do not automatically imply improved clinical outcomes. Only an impact study, randomizing physicians to the traditional or the new algorithm when deciding on preventive therapies, will answer that question. In conclusion, adding CAC scoring to baseline low-dose CT screening for lung cancer in a population of asymptomatic, intermediate-to-high risk heavy (former) smokers provides a combination of detection of previously unrecognized subjects at high-risk and accurate reclassification of subjects at intermediate risk into higher- or lower-risk categories. Application of this strategy could benefit 1 in every 12 subjects screened in terms of improved detection and targeting of preventive risk factor treatment. 136

137 Coronary artery calcium screening as part of low-dose computed tomography screening for lung cancer References 1. Blanchon T, Brechot JM, Grenier PA et al. Baseline results of the Depiscan study: a French randomized pilot trial of lung cancer screening comparing low dose CT scan (LDCT) and chest X-ray (CXR). Lung Cancer. 2007;58: Henschke CI, Yankelevitz DF, Libby DM, Pasmantier MW, Smith JP, Miettinen OS. Survival of patients with stage I lung cancer detected on CT screening. N Engl J Med. 2006;355: Infante M, Lutman FR, Cavuto S et al. Lung cancer screening with spiral CT: baseline results of the randomized DANTE trial. Lung Cancer. 2008;59: Lopes PA, Picozzi G, Mascalchi M et al. Design, recruitment and baseline results of the ITALUNG trial for lung cancer screening with low-dose CT. Lung Cancer. 2009;64: Roberts HC, Patsios D, Paul NS et al. Lung cancer screening with low-dose computed tomography: Canadian experience. Can Assoc Radiol J. 2007;58: van Iersel CA, de Koning HJ, Draisma G et al. Risk-based selection from the general population in a screening trial: selection criteria, recruitment and power for the Dutch-Belgian randomised lung cancer multi-slice CT screening trial (NELSON). Int J Cancer. 2007;120: Takasu J, Budoff MJ, O Brien KD et al. Relationship between coronary artery and descending thoracic aortic calcification as detected by computed tomography: The Multi-Ethnic Study of Atherosclerosis. Atherosclerosis [Epub ahead of print]. 8. D Agostino RB, Sr., Vasan RS, Pencina MJ et al. General cardiovascular risk profile for use in primary care: the Framingham Heart Study. Circulation. 2008;117: Wilson PW, D Agostino RB, Levy D, Belanger AM, Silbershatz H, Kannel WB. Prediction of coronary heart disease using risk factor categories. Circulation. 1998;97: US Diverse Populations Collaborative Group. Prediction of mortality from coronary heart disease among diverse populations: is there a common predictive function? Heart. 2002;88: Little RJA. Regression with missing X s: a review. J Am Stat Assoc. 1992;87: Xu DM, Gietema H, de Koning H et al. Nodule management protocol of the NELSON randomised lung cancer screening trial. Lung Cancer. 2006;54: Kim SM, Chung MJ, Lee KS, Choe YH, Yi CA, Choe BK. Coronary calcium screening using low-dose lung cancer screening: effectiveness of MDCT with retrospective reconstruction. AJR Am J Roentgenol. 2008;190: Isgum I, Rutten A, Prokop M, van Ginneken B. Detection of coronary calcifications from computed tomography scans for automated risk assessment of coronary artery disease. Med Phys. 2007;34: Rutten A, Isgum I, Prokop M. Coronary calcification: effect of small variation of scan starting position on Agatston, volume, and mass scores. Radiology. 2008;246: Ulzheimer S, Kalender WA. Assessment of calcium scoring performance in cardiac computed tomography. Eur Radiol. 2003;13: De Bruin A, Kardaun JW, Gast A, Bruin E, van Sijl M, Verweij G. Record linkage of hospital discharge register with population register: experiences at Statistics Netherlands. Stat J UN Econ Comm Eur. 2004;21:

138 Chapter Herings RM, Bakker A, Stricker BH, Nap G. Pharmaco-morbidity linkage: a feasibility study comparing morbidity in two pharmacy based exposure cohorts. J Epidemiol Community Health. 1992;46: Paas GR and Veenhuizen KC. Research on the validity of the LMR. Utrecht; Prismant: Reitsma JB, Kardaun JW, Gevers E, de Bruin A, van der Wal J, Bonsel GJ. [Possibilities for anonymous follow-up studies of patients in Dutch national medical registrations using the Municipal Population Register: a pilot study]. Ned Tijdschr Geneeskd. 2003;147: Prentice RL. A case-cohort design for epidemiologic cohort studies and disease prevention trials. Biometrika. 1986;73: Harrell FE, Jr. Regression Modelling Strategies with Applications to Linear Models, Logistic Regression, and Survival Analysis. New York, NY: Springer; Steyerberg EW. Clinical prediction Models: A Practical Approach To Development, Validation, and Updating. New York, NY: Springer; Harrell FE, Jr., Lee KL, Mark DB. Multivariable prognostic models: issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors. Stat Med. 1996;15: Steyerberg EW, Harrell FE, Jr., Borsboom GJ, Eijkemans MJ, Vergouwe Y, Habbema JD. Internal validation of predictive models: efficiency of some procedures for logistic regression analysis. J Clin Epidemiol. 2001;54: Pencina MJ, D Agostino RB, Sr., D Agostino RB, Jr., Vasan RS. Evaluating the added predictive ability of a new marker: from area under the ROC curve to reclassification and beyond. Stat Med. 2008;27: De Backer G, Ambrosioni E, Borch-Johnsen K et al. European guidelines on cardiovascular disease prevention in clinical practice. Third Joint Task Force of European and Other Societies on Cardiovascular Disease Prevention in Clinical Practice. Eur Heart J. 2003;24: Pearson TA, Blair SN, Daniels SR et al. AHA Guidelines for Primary Prevention of Cardiovascular Disease and Stroke: 2002 Update: Consensus Panel Guide to Comprehensive Risk Reduction for Adult Patients Without Coronary or Other Atherosclerotic Vascular Diseases. American Heart Association Science Advisory and Coordinating Committee. Circulation. 2002;106: Cook NR. Comments on Evaluating the added predictive ability of a new marker: From area under the ROC curve to reclassification and beyond by M. J. Pencina et al., Statistics in Medicine (DOI: /sim.2929). Stat Med. 2008;27: Graham I, Atar D, Borch-Johnsen K et al. European guidelines on cardiovascular disease prevention in clinical practice: executive summary. Eur Heart J. 2007;28: Grundy SM, Cleeman JI, Merz CN et al. Implications of recent clinical trials for the National Cholesterol Education Program Adult Treatment Panel III guidelines. Circulation. 2004;110: McPherson R, Frolich J, Fodor G, Genest J. Canadian Cardiovascular Society position statement: recommendations for the diagnosis and treatment of dyslipidemia and prevention of cardiovascular disease. Can J Cardiol. 2006;22: Arad Y, Goodman KJ, Roth M, Newstein D, Guerci AD. Coronary calcification, coronary disease risk factors, C-reactive protein, and atherosclerotic cardiovascular disease events: the St. Francis Heart Study. J Am Coll Cardiol. 2005;46: Greenland P, LaBree L, Azen SP, Doherty TM, Detrano RC. Coronary artery calcium score combined with Framingham score for risk prediction in asymptomatic individuals. JAMA. 2004;291:

139 Coronary artery calcium screening as part of low-dose computed tomography screening for lung cancer 35. Kondos GT, Hoff JA, Sevrukov A et al. Electron-beam tomography coronary artery calcium and cardiac events: a 37-month follow-up of 5635 initially asymptomatic low- to intermediate-risk adults. Circulation. 2003;107: Raggi P, Callister TQ, Cooil B et al. Identification of patients at increased risk of first unheralded acute myocardial infarction by electron-beam computed tomography. Circulation. 2000;101: Detrano RC, Wong ND, Doherty TM et al. Coronary calcium does not accurately predict near-term future coronary events in high-risk adults. Circulation. 1999;99: Hunold P, Vogt FM, Schmermund A et al. Radiation exposure during cardiac CT: effective doses at multi-detector row CT and electron-beam CT. Radiology. 2003;226: Jakobs TF, Becker CR, Ohnesorge B et al. Multislice helical CT of the heart with retrospective ECG gating: reduction of radiation exposure by ECG-controlled tube current modulation. Eur Radiol. 2002;12: Mahnken AH, Wildberger JE, Simon J et al. Detection of coronary calcifications: feasibility of dose reduction with a body weight-adapted examination protocol. AJR Am J Roentgenol. 2003;181: Shemesh J, Evron R, Koren-Morag N et al. Coronary artery calcium measurement with multi-detector row CT and low radiation dose: comparison between 55 and 165 mas. Radiology. 2005;236: Takahashi N, Bae KT. Quantification of coronary artery calcium with multi-detector row CT: assessing interscan variability with different tube currents pilot study. Radiology. 2003;228: Taylor AJ, Bindeman J, Feuerstein I, Cao F, Brazaitis M, O Malley PG. Coronary calcium independently predicts incident premature coronary heart disease over measured cardiovascular risk factors: mean three-year outcomes in the Prospective Army Coronary Calcium (PACC) project. J Am Coll Cardiol. 2005;46: Grundy SM. The changing face of cardiovascular risk. J Am Coll Cardiol. 2005;46: LaMonte MJ, FitzGerald SJ, Church TS et al. Coronary artery calcium score and coronary heart disease events in a large cohort of asymptomatic men and women. Am J Epidemiol. 2005;162: Assmann G, Cullen P, Schulte H. Simple scoring scheme for calculating the risk of acute coronary events based on the 10-year follow-up of the prospective cardiovascular Munster (PROCAM) study. Circulation. 2002;105: Conroy RM, Pyorala K, Fitzgerald AP et al. Estimation of ten-year risk of fatal cardiovascular disease in Europe: the SCORE project. Eur Heart J. 2003;24: Rubin DB, Schenker N. Multiple imputation in health-care databases: an overview and some applications. Stat Med. 1991;10:

140

141 CHAPTER 9 Smoking cessation and risk of cardiovascular disease among participants of a lung cancer screening trial

142 Chapter 9 Abstract Background Smoking is the leading preventable cause of death worldwide. The effect of (time since) smoking cessation on the process of atherosclerotic plaque formation and on the risk of a cardiovascular disease (CVD) event has not been firmly established in a population of heavy (former) smokers participating in a lung cancer screening trial. Objective To investigate the effect of smoking cessation on the coronary artery calcium (CAC) and the thoracic aorta calcium (TAC) score and on risk of future CVD events among participants of a lung cancer screening study. Methods Using a case-cohort design, this study was conducted among 910 participants (769 men) from the CT screening group of the NELSON study, a population-based, randomized, controlled trial among current or former smokers (mean number of pack-years, 40.8) aged 50 to 75 years. Past and current smoking behaviour was investigated with a baseline questionnaire. Incident fatal and nonfatal CVD events were used as main outcome measure. Results A total of 127 incident CVD events (11 fatal and 116 nonfatal cases) occurred during a median of 9.5 months of follow-up (range: 0-21 months). Among former smokers, we found a multivariate-adjusted linear relationship between a longer duration of smoking cessation and lower CAC scores (P=0.048 for trend); however, stopping smoking did not have an effect on the TAC score (P=0.83 for trend). Compared with current smokers, the multivariate-adjusted hazard ratio (HR) for CVD events for former smokers was 0.65 (95% confidence interval [CI], ). The protective effects of smoking cessation seem to occur almost immediately on quitting: a multivariate-adjusted HR of CVD events for those who had stopped smoking <1 year of 0.62 (95% CI, ). Further reduction of excess risk for those who had stopped smoking for a longer period of time (>1 -<10 years) could not be demonstrated. Conclusions Longer time since stopping smoking is associated with lower CAC (but not TAC) scores among former heavy smokers. This is reflected in a significant reduction of excess risk of CVD events among former smokers compared with current smokers. The protective effects of smoking cessation seem to occur already within 1 year after quitting. 142

143 Smoking cessation and risk of cardiovascular disease among participants of a lung cancer screening trial Introduction Coronary artery calcium (CAC) is a well-established risk factor for cardiovascular disease (CVD) events and all-cause mortality among smokers and non-smokers 1,2. Smoking cessation is associated with a reduction of risk for various CVD events (including myocardial infarction and stroke) as well as with a significant decrease in a variety of biochemical markers of CVD risk 3-6. However, the effect of smoking cessation on the process of atherosclerotic plaque formation/degradation (both in the coronary arteries (CAC) and in the thoracic aorta (TAC)) has not been extensively studied. Furthermore, the expected reduction of risk for CVD events has not been thoroughly investigated within a population with a history of very heavy smoking (>40 pack-years). You could hypothesize whether the beneficial effects of smoking cessation remain in people that have been exposed this thoroughly. Therefore, we analyzed a) the relationship of smoking cessation with the CAC and TAC scores and b) its effect on the risk of CVD events among (former) heavy smokers participating in a lung cancer screening trial. Materials and Methods Study population The NELSON Study is a randomized controlled population-based lung cancer screening trial comprising men and women aged 50 to 75 years, residing in 1 of 157 municipalities, that completed a mailed questionnaire regarding medical and smoking history (at least 15 pack-years of smoking. From 2004 to 2006, baseline low-dose CT scans were performed in participants randomly allocated to the screen group. Through linkage with the national registry of hospital discharge diagnoses, we first excluded participants with a known history of hospitalization for cardiovascular disease. After exclusion, we used a case-cohort design for this analysis including all incident CVD cases plus a random sample from the baseline screen group. Smoking history and other CVD risk factors Information on exposure to tobacco was collected through a baseline questionnaire. Dose was calculated for all participants as number of pack-years [number of cigarettes smoked per day x number of years smoked/20]. Participants were classified as current or former smokers. Former smokers were further subdivided into those who quitted <1 year, >1-<3 years, >3-<5 years, and >5-<10 years before baseline. Information on other known CVD risk factors was obtained from general practitioners (GPs). This included the current use of antihypertensive drugs, lipid-lowering drugs, oral hypoglycemic agents or insulin; systolic and diastolic blood pressure (BP); and nonfasting blood glucose, LDL and total cholesterol 143

144 Chapter 9 levels. The overall GP response rate was 70%. Missing values were imputed using regression methods implemented in SPSS software (SPSS 14.0, Chicago, Illinois) 7. We defined diabetes as a nonfasting plasma glucose level 11.1 mmol/l and/or the use of oral hypoglycemic agents or insulin. Hypertension was defined as a diastolic BP >90mmHg, systolic BP >140mmHg and/ or the use antihypertensive drugs. Hypercholesterolaemia was defined as a total cholesterol level >5.0 mmol/l, an LDL level >3.0 mmol/l and/or the use of lipid-lowering drugs. CAC/TAC scoring Baseline low-dose CT scans were conducted with a 16-slice MDCT scanner (Mx8000 IDT; Philips Medical Systems, Cleveland, Ohio in two participating hospitals; Sensation-16, Siemems AG, Forchheim, Germany in the third hospital). The scanning protocol consists of the following parameters were applied: 16x0.75mm collimation; pitch ; caudocranial scan direction; smallest field of view to include the outer rib margins. This way, transverse images with 1.0mm section thickness and 0.7mm increment were acquired from the level of the lung bases to the lung apices. No electrocardiographic triggering was performed; no contrast agent was administered. Low-dose exposure settings were applied based on body weight: 30 mas at a tube voltage of 120 kvp for subjects 80kg and 140 kvp for subjects >80kg. This corresponds to an effective radiation dose of msv. All scans were equally divided between two observers with two and three years of experience in reading cardiac CT who subsequently performed calcium scoring of the coronary arteries. The readers were blinded to other participant data. Prior to the start of the study, interobserver variability of CAC scoring was measured in a subset of 50 baseline scans not included in this study (ICC=0.97). To reduce image noise and to use data comparable to previously published studies 8 all scans were reconstructed to 3.1mm thick slices with an increment of 1.4mm by averaging four neighboring slices. All regions of 3 adjoining voxels (0.7 mm 3 ) with attenuation above 130 HU 9 were shown with a colored overlay. An investigator identified a point in each calcified lesion. Subsequently, three-dimensional component labeling using 26-connectivity was automatically performed 10 to mark all connected voxels as calcification. Calcium scores were computed using the Agatston algorithm 11. Ascertainment of CVD events All participants in the screen group of the NELSON study (n=7557) were linked with the national death registry and the national registry of hospital discharge diagnoses starting at the date of their baseline CT and with follow-up completed until January 1, 2006 (median follow-up: 10.0 months). This database linkage was performed on the basis of birth date, sex and postal code with a validated probabilistic method In these registries, all diagnoses are coded according to the International Classification of Diseases, 9th revision (ICD-9-CM). The primary end point in this analysis was total incident CVD events defined as fatal and 144

145 Smoking cessation and risk of cardiovascular disease among participants of a lung cancer screening trial nonfatal cases of coronary heart disease (CHD) (ICD-9 codes ), heart failure (code 428), cerebrovascular disease (codes ), peripheral arterial occlusive disease (PAOD) (codes 440, ), and aortic aneurysm/dissection (code 441). All other codes included in the ICD-9-CM as diseases of the circulatory system were not included as valid end points. Correct designation of causes of death has been established in a comparison study with patient medical records 16. Follow-up started after the baseline CT scan..for all participants, follow-up ended at the date of death, the occurrence of a nonfatal CVD event, or January 1, 2006, whichever came first. In participants with multiple cardiovascular hospital admissions during follow-up, the first hospital discharge diagnosis was used as end point. Statistical analysis Within the random sample, the crude association between median calcium scores according to time since smoking cessation was graphically displayed. We used linear regression analyses to quantitatively investigate the associations between absolute CAC/TAC scores and time since smoking cessation(as continuous variable, months). Natural log(calcium score +1) for TAC and CAC were used as outcome variables to account for the skewness of absolute CAC and TAC score distributions. Models were adjusted for age, sex, number of pack-years, diabetes, hypertension, and hypercholesterolaemia. The relationship between smoking cessation and incident CVD events was evaluated with Cox proportional hazard analyses. To account for the case-cohort design, modification of the standard errors was based on robust variance estimates. We used the method according to Prentice in which all subcohort members are equally weighted. Cases outside the subcohort are not weighted before failure and at failure receive the same weight as members of the subcohort 17. This method has been shown to resemble most closely estimates from a fullcohort analysis 18. Cox proportional hazard analyses were performed for increasing categories of time since quitting smoking and adjusted for CVD risk factors (age, sex, hypertension, diabetes and hypercholesterolaemia). Current smokers were used as reference category. To control for the fact that former smokers tend to have started smoking at an older age and to have smoked fewer cigarettes per day 3, we further adjusted the hazard ratios of CVD events with the number of pack-years. No additional adjustment for the presence and amount of CAC was performed because CAC needs to be considered an intermediate in the association between smoking and CVD events. Test for trend were conducted with using the median value for each category of smoking status as a continuous variable. Statistical analyses were performed with SPSS 14.0 software for Windows (SPSS Inc., Chicago, IL) and R software, version

146 Chapter 9 Results During a median of 9.5 months of follow-up (range 0-21 months), we documented 127 incident CVD events (11 fatal and 116 nonfatal cases). Table 1 shows the baseline characteristics of current and former smokers (according to the time since stopping smoking) within the random baseline sample of our study cohort. Former smokers were somewhat less healthy having a more frequent history of hypertension, hypercholesterolaemia and diabetes compared with people who continue to smoke. Table 1 Baseline Characteristics of Lung Cancer Screening Participants According to Smoking Status Characteristic Smoking Status Current Former Smokers, Years since Smoking Cessation Smokers <1 >1 - <3 >3 - <5 >5 - <10 Total No.* Age, y, mean Male sex, % Packyears, y, mean Hypertension, % Hypercholesterolaemia, % Diabetes, % * Based on a random baseline sample (subcohort, n=798) of participants from the screen group of this lung cancer CT screening trial. Pack-years are calculated as ((number of cigarettes smoked per day x number of years smoked)/20). Figure 1 presents median CAC and TAC scores according to smoking status. Three important observations can be drawn form this graph. First, a healthy-smoking effect can be observed with current smokers having a substantially lower CAC score compared with former smokers. Second, among former smokers an increasing time since quitting smoking is associated with a linear and stepwise decrease of median CAC scores. Third, this pattern was not observed for median TAC scores. No effect of smoking cessation on the amount of TAC was established. Table 2 shows these associations quantitatively. Linear regression analysis with adjustment for age, sex, pack-years, history of hypertension, hypercholesterolaemia, and diabetes showed that an increase of 1 month in duration of smoking cessation was associated with a lower CAC score (β = -0.01; P = for trend). Time since smoking cessation was not associated with TAC (P = 0.83 for trend). 146

147 Smoking cessation and risk of cardiovascular disease among participants of a lung cancer screening trial Figure 1 Median coronary artery calcium (CAC) and thoracic aorta calcium (TAC) scores according to duration of stopping smoking among a random sample of participants of a lung cancer screening study. Table 2 Association of Time since Smoking Cessation with Natural Log(CAC+1) and Natural Log(TAC+1) (n=798)* CAC TAC ß (95% CI) P Value for Trend ß (95% CI) P Value for Trend Time since smoking cessation ( ) ( ) 0.83 *Both models adjusted for age, sex, pack-years, history of hypertension, hypercholesterolaemia, and diabetes. ß represents the increase/decrease (95% confidence interval) in log(calcium score+1) for coronary artery calcium (CAC) and thoracic aorta calcium (TAC) per 1 month longer duration of smoking cessation. Table 3 shows the effect of smoking cessation on the risk of any CVD event. Compared with current smokers, the multivariate-adjusted hazard ratio (HR) for any CVD event among former smokers regardless of the time since smoking cessation was 0.65 (95% CI ). In secondary analyses, we divided former smokers into those who stopped smoking less than 1, 1 to 3, 3 to 5, 5 to 10 years before baseline (Table 4). The multivariate-adjusted HRs for any CVD event in all four categories of former smokers showed a similar, non-significant reduction of excess risk compared with current smokers. Therefore, this risk reduction was irrespective of the time since they had stopped smoking. Participants who had stopped smoking more than 5 years ago had a reduced risk (HR, 0.75 [95% CI, ]) more or less comparable with people who had stopped smoking during the last year (HR, 0.62 [95% CI, ]) (P = 0.42 for trend). Although not statistically significant in the present analysis, it seems that the protective effects of smoking cessation for the occurrence of any CVD event occur already within the first year after quitting. 147