Abstract: Objectives. Cancer has been proposed as a cardiovascular risk factor. We aimed to assess the cardiovascular risk profile and coronary angiography (CA) findings of cancer patients and compare them to those of patients without cancer. Methods. A retrospective case-control analysis was conducted on randomly enrolled cancer and non-cancer patients from a high-volume cardio-oncology center and a tertiary cardiology center, respectively, who underwent CA from April 2008 to June 2018. Baseline demographics, laboratory findings, cancer status and treatment, and current and prior CA findings were collected by chart review. Coronary artery disease (CAD) burden was assessed with machine-learning (neural-network) guided propensity-score adjusted multivariable regression, controlling for known CAD confounders. Results. Of the 480 enrolled patients, a total of 240 (50%) had cancer. Fewer cancer vs non-cancer patients had clinically significant lesions on the left anterior descending artery (25.00% vs 39.17%, respectively; P<.01) and left circumflex artery (15.83% vs 30.00%, respectively; P<.001). Left main and right coronary artery disease prevalence was similar. Subjects with cancer were less likely to have multivessel CAD (odds ratio, 0.53; 95% confidence interval, 0.29-0.98; P=.04) and significant left circumflex artery lesions (odds ratio, 0.47; 95% confidence interval, 0.26-0.85; P=.01), independent of known CAD confounders. Conclusions. Patients with cancer have a lower burden of angiographically detected coronary atherosclerosis. Cancer patients are more likely than non-cancer patients to undergo CA for reasons other than suspicion of CAD. Further studies should prospectively analyze the impact of cancer on the development of CAD.
J INVASIVE CARDIOL 2019;31(1):21-26.
Key words: cardio-oncology, coronary angiography, coronary artery disease, machine learning, risk factors
Cardiovascular disease (CVD) and cancer are the two main causes of mortality worldwide,1 with CVD expected to impact 1 in 2 people globally within 15 years, while doubling its current cost of approximately 1 in 5 health-care dollars in the United States alone.2 The prevalence of cancer and cancer survivors continues to grow, given the steady march of therapeutic successes that shift the morbidity and mortality from malignancy to CVD.3,4 These diseases are synergistically driving a growing movement of cardio-oncology, which is mainly challenged by a new “double-hit” threat: the already widespread burden of CVD (accentuated by the complex pathophysiology of cancer) and the toxicity of anticancer therapies. Basic science research supports the contribution of cancer to the pathophysiology of ischemic CVD, considering its prothrombotic and proinflammatory mechanisms,5,6 promoting endothelial damage with subsequent platelet aggregation, vasospasm, and possible accelerated atherosclerosis, similar to certain autoimmune disorders.5,7,8 Anticancer therapies also contribute to cardiovascular morbidity through a wide spectrum of cardiotoxic manifestations,9 such as precipitating cardiac ischemia,10 heart failure,11,12 myocarditis,13 or stress cardiomyopathy.14
Evidence-based frameworks guiding optimal cardio-oncology care are gradually emerging,15,16 yet the scant evidence is generally devoid of large cohort studies or adequately powered randomized clinical trials. As a result, optimal medical and interventional cardiac strategies are less understood in the cardio-oncology population, particularly since many cardiovascular trials exclude cancer patients or do not adjust for malignancy status. Cancer patients often face specific issues that only rarely complicate the cardiovascular management of non-oncology cases, ie, thrombocytopenia, prior mediastinal radiation, or conditions that require prompt or unexpected dual-antiplatelet therapy discontinuation following percutaneous coronary intervention (PCI).17 Although initial reports suggest that outcomes following PCI are similar in cancer patients compared to the general population,18 a more accurate understanding of the cardiovascular profile of cardio-oncology patients is necessary for optimizing management.16,19,20
Our objective was to evaluate the coronary angiography (CA) findings and clinical characteristics of cancer patients and compare them to those of patients without cancer.
This retrospective case-control analysis randomly enrolled cancer and non-cancer patients from a high-volume cardio-oncology center and a tertiary cardiology center, respectively, who underwent CA from April 2008 to June 2018. Subjects were divided into two equally-sized groups, based on cancer status. Baseline demographics, laboratory findings, cancer status and treatment, current and prior CA findings, comorbidities, and cardiovascular management were collected by chart review. Institutional Review Board ethics approval was granted through the University of Texas MD Anderson Cancer Center and informed, written consent was obtained.
Descriptive statistics were conducted for the full sample, in addition to bivariable analysis performed according to cancer status (yes/no) with independent sample t-test comparing means and Wilcoxon rank-sum tests comparing medians for continuous variables, as appropriate, while Pearson’s chi-square test or Fisher’s exact test compared proportions for categorical variables, as appropriate.
Causal inference statistics were utilized to investigate cancer and coronary artery disease (CAD) burden by the number of vessels involved and separately by the degree of stenotic lesions by major coronary vessels according to existing clinical criteria. CAD was considered to be clinically significant if there was a stenosis of ≥70% in the left anterior descending (LAD), left circumflex (LCX), or right coronary artery (RCA), or if there was a stenosis of ≥50% of the left main (LM) coronary artery according to CA.21 Single-vessel disease was defined as clinically significant lesions in the LAD, LCX, or RCA. Double-vessel disease was defined as the presence of significant lesions in the LM or two other distinct coronary arteries. Triple-vessel disease was defined as either significant disease of the LM and RCA, or significant disease of the LAD, LCX, and RCA.
Doubly robust forward and backward stepwise regression was used to augment machine learning (backward-propagation generated neural networks), which in turn guided propensity-score adjusted multivariable regression, controlling for known CAD confounders and the likelihood of developing cancer. In the first phase, significant associations in bivariable analysis and those extracted from their published research or clinical significance were identified for potential inclusion in the final regression models. Second, those variables were assessed in stepwise forward and backward regression to support statistical and clinical decision making of final model inclusion. Third, the final models were built with the propensity score of developing cancer (confirming balance among the blocks and then adding the score with the variables that constructed it into the final regression model), with model performance assessed based on regression diagnostics: comparison to the neural network results by accuracy and root mean squared error (RMSE), Hosmer-Lemeshow’s goodness-of-fit test, area under the curve, Akaike’s and Schwarz’s Bayesian information criteria, and specification error and multicollinearity testing with tolerance, variance inflation factor, and correlation matrix. Penalized logistic regression was utilized for subanalysis by particular coronary vessel, given the rarer outcome. Subgroup analysis by coronary vessel and cancer type was additionally conducted. An academic physician, academic biostatistician, and data scientist reviewed the final models to verify they were consistent with sound clinical principles and statistical theory. All regression estimates are reported as fully adjusted results with 95% confidence intervals (CIs). A two-tailed P-value <.05 served as the threshold for statistical significance. STATA 14.2 (STATACorp) was utilized for statistical analyses, and Java 9 (Oracle) was used for machine-learning analyses.
Of the 480 enrolled patients, a total of 240 (50%) had a diagnosis of cancer (Table 1). Compared to patients without cancer, those with a cancer diagnosis were older and more likely to be female, with a history of hypertension, diabetes, and premature CAD in their families. Cancer patients also had higher serum triglycerides and lower high-density lipoprotein levels (P<.05). The most common types of cancer diagnosed in our group were solid cancers (68%), while only one-third had hematologic malignancies (Table 2). There were also 4 patients (1.7%) with both solid and hematologic cancers.
Cancer patients who underwent CA were significantly less likely to have chest pain (18.33% vs 41.67%; P<.001) or previous CAD (13.75% vs 23.75%; P<.01) (Table 3). Cancer patients had overall less severe coronary lesions than patients without cancer, being less likely to have significant lesions on the LAD (25.00% vs 39.17%; P<.01) and LCX (15.83% vs 30.00%; P<.001), but having comparable LM and RCA disease, with fewer referrals for coronary artery bypass grafting (0.83% vs 10.42%; P<.001). Cancer vs non-cancer patients had lower rates of prior PCI (15.48% vs 54.85%; P<.001) and fewer stents (17.08% vs 29.58%; P<.001).
In multivariable propensity-score adjusted regression, cancer significantly reduced the odds of multivessel CAD (OR, 0.53; 95% CI, 0.29-0.98; P=.04) (Table 4), independent of age, sex, dyslipidemia, peripheral artery disease, and the likelihood of developing cancer. This model had similar accuracy RMSE compared to the neural-network machine-learning algorithm (72.24% and 0.436 vs 66.46% and 0.464).
In subgroup analysis by vessel type, cancer vs non-cancer patients had comparable LM, LAD, and RCA disease odds, but lower odds of significant/complete vs none/non-significant LCX lesions (OR, 0.47; 95% CI, 0.26-0.85; P=.01), even when adjusting for age, body mass index, dyslipidemia, peripheral artery disease, type 2 diabetes mellitus, and triglycerides (Table 4).
To the best of our knowledge, this is the first adequately powered machine-learning augmented causal inference analysis of CA findings among cancer vs non-cancer patients. Our study suggests that cancer overall significantly and independently reduces the odds of diagnosing multivessel compared to single-vessel CAD on CA, specifically significant or complete stenosis in the LCX, but not other major coronary vessels. These findings provide a detailed look at the cardiovascular profile of oncology patients undergoing CA, while showing that causal inference analysis can be confirmed and replicated automatically through machine learning (and, therefore, may accelerate real-time findings on larger high-dimensional datasets).
Cancer and CAD are frequently co-diagnosed, primarily due to the increasing age of cancer survivors, shared risk factors, and the cardiotoxic effects of cancer therapy.7,22,23 This has led to the appearance of a specific group of complex, interventional onco-cardiology patients, in which standard guidelines do not necessarily apply.17,24 Better defining and characterizing this specific population is necessary in order to establish both optimal cardio-oncology management and future research goals. In our group, although cancer patients appeared to have a higher burden of cardiovascular risk factors than patients without cancer (ie, more obesity, hypertension, diabetes, family history of premature CAD), CAD was found less often and when it was present, it was less severe.
It has been reported that patients with certain cancers experience increased rates of ischemic-related cardiovascular events, but currently no longitudinal association has been determined between malignancy and the development of CAD.25 However, the direct impact of cardiotoxic cancer therapies on the development of CAD is well established.7,26 Cardiovascular outcomes are generally worse in patients with a history of chemotherapy or irradiation compared to those who have never received either,27 possibly due to more advanced disease. Our study did not find any association between CAD, coronary lesion location, and history of chemotherapy or radiotherapy. However, we did not analyze this issue within particular cancer diagnoses because of the limited sample of cases within each group. Specific anatomical features of CAD have been studied in relation to breast cancer,28 lung cancer,29 and lymphoma,27 particularly due to the administration of radiotherapy for these cancers, a known risk factor for CAD.30
The decreased odds of diagnosing multivessel (compared to single-vessel) and significant or complete coronary stenosis on CA in cancer patients compared to non-cancer patients does not mean that cancer is somehow protective of these conditions, which would be in contradiction to medical literature and proposed pathogenic mechanisms. Rather, these results highlight a different overall pretest probability of CAD on CA in cancer patients due to different indications compared to the general population of cardiovascular patients. In our experience, non-cancer patients are more likely to undergo CA for suspicion of CAD, while in cancer patients, it is often that CA is part of pretreatment protocol for the malignancy despite no (or low) clinical suspicion for CAD. This selection bias is one of our study’s primary limitations, as we can only present the “real-world” cardiovascular profile of cancer patients undergoing CA and odds for encountering specific coronary lesions on CA; no causative hypothesis can be assumed. Future studies should address this concern by adjusting for pretest probability for CAD. The study’s observational nature also represents a limitation. Furthermore, this study lacks time-series and health-system data to provide a detailed analysis of the time from cancer to treatment to CAD and what ancillary services and level of health-care utilization each patient received.
Certain elements contribute to this study’s strengths. The relatively large number of randomly selected patients ensures that the study is adequately powered. The use of machine-learning algorithms for statistical analysis is generally more accurate and has proven its utility in cardiovascular medicine research.31
Compared to patients without malignancy, we found that cancer patients are less likely to have multivessel (compared to single-vessel), significant, or obstructive lesions on CA, despite a higher prevalence of conventional cardiovascular risk factors. These results are probably due to the likelihood of performing CA in cancer patients for indications other than CAD suspicion in patients with otherwise low pretest probability for CAD. Further studies should prospectively analyze the longitudinal association between cancer and the evolution of coronary lesions. This insight into the cardiovascular risk profile of cancer patients based on CA data and overall clinical presentation contributes to the understanding of the particularities of the interventional cardio-oncology population and “sets the stage” for the development of specific guidelines and more targeted cardiovascular therapy.
1. GBD 2013 Mortality and Causes of Death Collaborators. Global, regional, and national age-sex specific all-cause and cause-specific mortality for 240 causes of death, 1990-2013: a systematic analysis for the Global Burden of Disease Study 2013. Lancet. 2015;385:117-171.
2. Heidenreich PA, Trogdon JG, Khavjou OA, et al. Forecasting the future of cardiovascular disease in the United States: a policy statement from the American Heart Association. Circulation. 2011;123:933-944.
3. Miller KD, Siegel RL, Lin CC, et al. Cancer treatment and survivorship statistics, 2016. CA Cancer J Clin. 2016;66:271-289.
4. Barac A, Murtagh G, Carver JR, et al. Cardiovascular health of patients with cancer and cancer survivors: a roadmap to the next level. J Am Coll Cardiol. 2015;65:2739-2746.
5. Demers M, Krause DS, Schatzberg D, et al. Cancers predispose neutrophils to release extracellular DNA traps that contribute to cancer-associated thrombosis. Proc Natl Acad Sci U S A. 2012;109:13076-13081.
6. Blann AD, Dunmore S. Arterial and venous thrombosis in cancer patients. Cardiol Res Pract. 2011;2011:394740.
7. Giza DE, Marmagkiolis K, Mouhayar E, Durand JB, Iliescu C. Management of CAD in patients with active cancer: the interventional cardiologists’ perspective. Curr Cardiol Rep. 2017;19:56.
8. Chen KC, Liao YC, Wang JY, Lin YC, Chen CH, Juo SH. Oxidized low-density lipoprotein is a common risk factor for cardiovascular diseases and gastroenterological cancers via epigenomical regulation of microRNA-210. Oncotarget. 2015;6:24105-24118.
9. Conway A, McCarthy AL, Lawrence P, Clark RA. The prevention, detection and management of cancer treatment-induced cardiotoxicity: a meta-review. BMC Cancer. 2015;15:366.
10. Herrmann J, Lerman A. An update on cardio-oncology. Trends Cardiovasc Med. 2014;24:285-295.
11. van Nimwegen FA, Ntentas G, Darby SC, et al. Risk of heart failure in survivors of Hodgkin lymphoma: effects of cardiac exposure to radiation and anthracyclines. Blood. 2017;129:2257-2265.
12. Cardinale D, Colombo A, Bacchiani G, et al. Early detection of anthracycline cardiotoxicity and improvement with heart failure therapy. Circulation. 2015;131:1981-1988.
13. Asnani A. Cardiotoxicity of immunotherapy: incidence, diagnosis, and management. Curr Oncol Rep. 2018;20:44.
14. Budnik M, Kucharz J, Wiechno P, et al. Chemotherapy-induced Takotsubo syndrome. Adv Exp Med Biol. 2018 Jun 9 (Epub ahead of print).
15. Giza DE, Boccalandro F, Lopez-Mattei J, et al. Ischemic heart disease: special considerations in cardio-oncology. Curr Treat Options Cardiovasc Med. 2017;19:37.
16. Iliescu CA, Grines CL, Herrmann J, et al. SCAI expert consensus statement: evaluation, management, and special considerations of cardio-oncology patients in the cardiac catheterization laboratory (endorsed by the Cardiological Society of India, and Sociedad Latino Americana de Cardiologia intervencionista). Catheter Cardiovasc Interv. 2016;87:E202-E223.
17. Liu VY, Agha AM, Lopez-Mattei J, et al. Interventional cardio-oncology: adding a new dimension to the cardio-oncology field. Front Cardiovasc Med. 2018;5:48.
18. Hess CN, Roe MT, Clare RM, et al. Relationship between cancer and cardiovascular outcomes following percutaneous coronary intervention. J Am Heart Assoc. 2015;4(7).
19. Krone RJ. Managing coronary artery disease in the cancer patient. Prog Cardiovasc Dis. 2010;53:149-156.
20. Brown ML, Schaff HV, Sundt TM. Conduit choice for coronary artery bypass grafting after mediastinal radiation. J Thorac Cardiovasc Surg. 2008;136:1167-1171.
21. Patel MR, Calhoon JH, Dehmer GJ, et al. ACC/AATS/AHA/ASE/ASNC/SCAI/SCCT/STS 2017 appropriate use criteria for coronary revascularization in patients with stable ischemic heart disease: a report of the American College of Cardiology Appropriate Use Criteria Task Force, American Association for Thoracic Surgery, American Heart Association, American Society of Echocardiography, American Society of Nuclear Cardiology, Society for Cardiovascular Angiography and Interventions, Society of Cardiovascular Computed Tomography, and Society of Thoracic Surgeons. J Am Coll Cardiol. 2017;69:2212-2241.
22. Hecht HS, Henschke C, Yankelevitz D, Fuster V, Narula J. Combined detection of coronary artery disease and lung cancer. Eur Heart J. 2014;35:2792-2796.
23. Chan AO, Lam KF, Tong T, et al. Coexistence between colorectal cancer/adenoma and coronary artery disease: results from 1382 patients. Aliment Pharmacol Ther. 2006;24:535-539.
24. Zamorano JL, Lancellotti P, Rodriguez Munoz D, et al. 2016 ESC position paper on cancer treatments and cardiovascular toxicity developed under the auspices of the ESC Committee for Practice Guidelines: The Task Force for Cancer Treatments and Cardiovascular Toxicity of the European Society of Cardiology (ESC). Eur Heart J. 2016;37:2768-2801.
25. Whitlock MC, Yeboah J, Burke GL, et al. Cancer and its association with the development of coronary artery calcification: an assessment from the multi-ethnic study of atherosclerosis. J Am Heart Assoc. 2015;4(11).
26. Herrmann J, Yang EH, Iliescu CA, et al. Vascular toxicities of cancer therapies: the old and the new — an evolving avenue. Circulation. 2016;133:1272-1289.
27. van Rosendael AR, Daniels LA, Dimitriu-Leen AC, et al. Different manifestation of irradiation induced coronary artery disease detected with coronary computed tomography compared with matched non-irradiated controls. Radiother Oncol. 2017;125:55-61.
28. Nilsson G, Holmberg L, Garmo H, et al. Distribution of coronary artery stenosis after radiation for breast cancer. J Clin Oncol. 2012;30:380-386.
29. Zhang F, Chen J, Yang ZG, Liu H. Preoperative assessment of coronary arteries by cardiac computed tomography in patients with lung cancer. Thorac Cardiovasc Surg. 2018 Aug 16 (Epub ahead of print).
30. Spetz J, Moslehi J, Sarosiek K. Radiation-induced cardiovascular toxicity: mechanisms, prevention, and treatment. Curr Treat Options Cardiovasc Med. 2018;20:31.
31. Johnson KW, Torres Soto J, Glicksberg BS, et al. Artificial intelligence in cardiology. J Am Coll Cardiol. 2018;71:2668-2679.
*Joint first authors.
From the 1Department of Cardiology, The University of Texas MD Anderson Cancer Center, Houston, Texas; 2Department of Internal Medicine, McGovern Medical School at The University of Texas Health Science Center at Houston, Houston, Texas; and 3Department of Cardiology, Elias Emergency University Hospital, Carol Davila University of Medicine and Pharmacy, Bucharest, Romania.
Disclosure: The authors have completed and returned the ICMJE Form for Disclosure of Potential Conflicts of Interest. The authors report no conflicts of interest regarding the content herein.
Manuscript submitted September 3, 2018 and accepted September 11, 2018.
Address for correspondence: Cezar Iliescu, MD, FACC, FSCAI, Associate Professor, Department of Cardiology, The University of Texas MD Anderson Cancer Center, 1400 Pressler Street, Unit 1451, Houston, TX 77030. Email: email@example.com