Prevalence of Colorectal Neoplasm Among Patients with Newly Diagnosed Coronary Artery Disease

ORIGINAL CONTRIBUTION

Prevalence of Colorectal Neoplasm Among Patients With Newly Diagnosed Coronary Artery Disease

Annie On On Chan, MD, PhD Context Colorectal neoplasm and coronary artery disease (CAD) share similar risk Man Hong Jim, MD factors, and their co-occurrence may be associated. Kwok Fai Lam, PhD Objectives To investigate the prevalence of colorectal neoplasm in patients with CAD Jeffrey S. Morris, PhD in a cross-sectional study and to identify the predisposing factors for the association of the 2 diseases. David Chun Wah Siu, MD Design, Setting, and Participants Patients in Hong Kong, China, were recruited Teresa Tong, BSc for screening colonoscopy after undergoing coronary angiography for suspected CAD Fook Hong Ng, MD during November 2004 to June 2006. Presence of CAD (n=206) was defined as at least 50% diameter stenosis in any 1 of the major coronary arteries; otherwise, pa- Siu Yin Wong, MD tients were considered CAD-negative (n=208). An age- and sex-matched control group Wai Mo Hui, MD was recruited from the general population (n=207). Patients were excluded for use Chi Kuen Chan, MD of aspirin or statins, personal history of colonic disease, or colonoscopy in the past 10 years. Kam Chuen Lai, MD Main Outcome Measures The prevalence of colorectal neoplasm in CAD- Ting Kin Cheung, MD positive, CAD-negative, and general population participants was determined. Bivari- Pierre Chan, MD ate logistic regression was performed to study the association between colorectal neoplasm and CAD and to identify risk factors for the association of the 2 diseases after Grace Wong, MD adjusting for age and sex. Man Fung Yuen, MD, PhD Results The prevalence of colorectal neoplasm in the CAD-positive, CAD-negative, Yuk Kong Lau, MD and general population groups was 34.0%, 18.8%, and 20.8% (PϽ.001 by ␹2 test), prevalence of advanced lesions was 18.4%, 8.7%, and 5.8% (PϽ.001), and preva- Stephen Lee, MD lence of cancer was 4.4%, 0.5%, and 1.4% (P=.02), respectively. Fifty percent of the Ming Leung Szeto, MD cancers in CAD-positive participants were early stage. After adjusting for age and sex, Benjamin C. Y. Wong, MD, PhD an association still existed between colorectal neoplasm and presence of CAD (odds ratio [OR], 1.88; 95% confidence interval [CI], 1.25-2.70; P=.002) and between ad- Shiu Kum Lam, MD vanced lesions and presence of CAD (OR, 2.51; 95% CI, 1.43-4.35; P=.001). The metabolic syndrome (OR, 5.99; 95% CI, 1.43-27.94; P=.02) and history of smoking OLORECTAL CANCER IS THE (OR, 4.74; 95% CI, 1.38-18.92; P=.02) were independent factors for the association second most prevalent can- of advanced colonic lesions and CAD. 1 cer worldwide. There were Conclusions In this study population undergoing coronary angiography, the preva- about 1 million new cases and lence of colorectal neoplasm was greater in patients with CAD. The association be- 500C 000 deaths due to colorectal can- tween the presence of advanced colonic lesions and CAD was stronger in persons with cer in 2002.1 It has been estimated that the metabolic syndrome and a history of smoking. 1 in 20 healthy individuals will even- JAMA. 2007;298(12):1412-1419 www.jama.com tually develop colorectal cancer. Coro- nary artery disease (CAD) is the single Author Affiliations: Departments of Medicine (Drs A. pital (Drs G. Wong and Szeto), Hong Kong, China; O. O. Chan, Jim, Siu, Hui, C. K. Chan, Lai, Cheung, Department of Biostatistics, M. D. Anderson Cancer leading cause of death in the United P. Chan, Yuen, Lee, B. C. Y. Wong, and S. K. Lam and Center, Houston, Texas (Dr Morris). States and other industrialized coun- Ms Tong) and Statistics and Actuarial Science (Dr K. Corresponding Author: Annie On On Chan, MD, PhD, 2 F. Lam), University of Hong Kong, Department of Department of Medicine, University of Hong Kong, tries. We previously published a ret- Medicine, Ruttonjee Hospital (Drs Ng, S. Y. Wong, and Queen Mary Hospital, Pokfulam Road, Hong Kong, rospective study that reported a strong Lau), and Department of Medicine, Tuen Mun Hos- China ([email protected]).

Downloaded From: https://jamanetwork.com/ on 09/25/2021 COLORECTAL NEOPLASM IN NEWLY DIAGNOSED CORONARY ARTERY DISEASE

association between colorectal cancer/ tal, and Ruttonjee Hospital. These are re- They were age- and sex-matched to the adenoma and CAD, possibly due to the gionalhospitalsservingpatientswithheart CAD-positive group within 5 years on a sharing of common environmental risk disease in the Hong Kong West cluster. nearly case-to-case basis. This group did factors.3 The study was approved by the insti- not undergo coronary angiogram. The Colorectal cancer and CAD share tutional review boards of the Univer- same exclusion criteria were also ap- similar environmental risks factors, such sity of Hong Kong/Queen Mary Hospi- plied in this second control group. as diabetes mellitus; smoking; hyperlip- tal, Grantham Hospital, and Ruttonjee All individuals in the 3 groups were idemia; sedentary lifestyle; high-fat, low- Hospital, all in Hong Kong. All partici- invited to participate. Information on fiber diet; obesity; and hypertension.4-6 pants provided written informed consent. age, sex, history of smoking, diabetes The metabolic syndrome is being in- Consecutive patients with suspected mellitus, hypertension, family history creasingly recognized as a significant CAD (ie, those with angina or abnor- of colorectal cancer, and use and du- health hazard worldwide.7 It com- mal exercise stress test results) who pre- ration of aspirin and statins was re- prises a constellation of metabolic risk sented for the first time for coronary an- corded. Waist circumference was mea- factors, including most of the underly- giography were invited to participate in sured. Blood was drawn and fasting ing risk factors for both colorectal can- the study. Because we aimed at study- glucose level and complete lipid pro- cer and CAD: diabetes or impaired glu- ing the prevalence of colorectal neo- file were measured. The metabolic syn- cose tolerance, hypertriglyceridemia, low plasm in patients with newly diag- drome was defined as at least 3 of the high-density lipoprotein cholesterol nosed CAD, and to avoid the potential following criteria set forth by the Na- level, central obesity, and hyperten- protective effect of aspirin and statins on tional Cholesterol Education Pro- sion. Persons with the metabolic syn- colorectal neoplasm, patients with a his- gram’s modified Adult Treatment Panel drome have been reported to have in- tory of CAD for more than 1 year or who III (Asian Pacific Region criteria)13: creased risk of developing CAD.8,9 We had been taking aspirin or a statin for (1) abdominal obesity: waist circum- postulated that the metabolic syn- more than 1 year were excluded. In ad- ference of at least 36 in (91 cm) for men drome might also be an important risk dition, those presenting for coronary an- and 32 in (81 cm) for women; (2) low factor for the development of both co- giography for reasons other than sus- high-density lipoprotein cholesterol lorectal cancer and CAD. pected CAD (eg, congestive heart failure, level: 40 mg/dL (1.03 mmol/L) or lower Hong Kong is an industrialized re- cardiomyopathy) were also excluded. for men and 50 mg/dL (1.3 mmol/L) or gion with incidences of and mortality due Additional exclusion criteria were his- lower for women; (3) hypertriglyceri- to colorectal cancer and CAD similar to tory of colonic disease, such as colorec- demia: 150 mg/dL (1.7 mmol/L) or that in western countries.10-12 Although tal cancer, polyp, and inflammatory higher; (4) hypertension: blood pres- we observed an association between co- bowel disease, and history of colorectal sure 130/85 mm Hg or higher; and (5) lorectal cancer and CAD in our previ- surgery or colonoscopy within the pre- impaired glucose tolerance: fasting glu- ous study,3 we were not able to identify vious 10 years. cose 110 mg/dL (6.1 mmol/L) or higher. the risk factors involved because of its Consecutive patients with coronary retrospective nature.3 We thus de- angiogram were invited for colonos- Colonoscopy signed and conducted the current cross- copy regardless of hemoglobin status to Colonoscopy was scheduled within 8 sectional study, the primary aim of which avoid preselection in the CAD-positive weeks after assessing for eligibility or af- was to investigate the prevalence of co- group of patients with gastrointestinal ter revascularization for critical pa- lorectal cancer and adenoma (colorec- tract bleeding due to aspirin or clopido- tients. All patients received the same tal neoplasms) in patients with newly di- grel therapy. In accordance with Ameri- bowel preparation. Colonoscopies were agnosed CAD. A secondary aim was to can College of Cardiology/American repeated the next day for those with poor identify the underlying risk factors, af- Heart Association guidelines, patients bowel preparation. Patients taking clo- ter adjusting for age and sex, that pre- were defined as CAD-positive if at least pidogrel had their treatment changed to disposed to the 2 conditions. The re- 50% diameter stenosis in any 1 of the ma- subcutaneous heparin 2 days before colo- sults have important implications in jor coronary arteries was found on coro- noscopy. Endoscopists were blinded to prevention of both colorectal neoplasm nary angiography; otherwise, patients the CAD status of the patients. The with- and CAD, as well as for the screening were defined as CAD-negative. drawal time of the colonoscopy proce- strategy of colorectal cancer. Because age and sex matching was not dure was more than 6 min to minimize available in the CAD-negative group, a the chance of lesions being missed.14 In- METHODS second control group from the general complete examination was excluded Patient Recruitment population was recruited. They were from analysis. The study period encompassed Novem- healthy participants who were asymp- Advanced colonic lesion was de- ber 2004 to June 2006. Participating hos- tomatic other than having functional epi- fined as presence of cancer or adeno- pitals include the cardiac divisions of gastric pain but with a normal result on mas with villous component, with high- Queen Mary Hospital, Grantham Hospi- upper endoscopy performed in our unit. grade dysplasia, or 1 cm or larger.

Downloaded From: https://jamanetwork.com/ on 09/25/2021 COLORECTAL NEOPLASM IN NEWLY DIAGNOSED CORONARY ARTERY DISEASE

δ =1ϩ (γ −1){γ [P(Y =1)−P(W =1) Table 1. Key Risk Factors and Assessment of Heterogeneity of the CAD-Negative and i i i i i 2 ϩ 2ϩ General Population Groups ] −[P(Yi=1) P(Wi=1)] 2[P(Yi=1) ϩ CAD- General P(Wi=1)]} Negative Group Population Group Adjusted Risk Factors (n = 208) (n = 207) P Valuea and Male, No. (%) 132 (63.5) 152 (73.4) .17 P(Y =1,W =0)=P(Y =1)−P(Y =1,W =1), Age, mean (SD), y 62.9 (8) 62.9 (8) .97 i i i i i Family history of colorectal cancer, No. (%) 12 (5.8) 24 (11.6) .18 P(Yi=0,Wi=1)=P(Wi=1)−P(Yi=1,Wi=1), Ever smoking, No. (%) 37 (17.8) 45 (21.7) .94

Metabolic syndrome, No. (%) 38 (18.3) 20 (9.7) .08 P ( Y i =0,W i =0)=1−P ( Y i =1)−P ϩ Fasting glucose Ն110 mg/dL, No. (%) 42 (20.2) 42 (20.3) Ͼ.99 (Wi=1) P(Yi=1,Wi=1). Abbreviation: CAD, coronary artery disease. SI conversion: To convert fasting glucose to mmol/L, multiply by 0.555. Maximum likelihood estimates of the a Adjusted using the Holm18 method for multiple testing. regression coefficients, namely ␣,␤,and ␪, were computed using a tailor-made The size of a polyp was assessed by com- lorectal neoplasm, respectively. The mar- computer program in FORTRAN lan- paring it with an open colonoscope bi- ginal probabilities of Y and W were mod- guage.The95%confidenceintervals(CIs) opsy forcep (Bard, Murray Hill, NJ). The eled by the usual logistic regression were computed based on the asymptotic site of colorectal neoplasm was defined analysis with models ␹2 distributed properties of the log- as right-sided if proximal to splenic flex- α +α + +α likelihood ratio. The CAD-negative and ure. The size, site, histologic findings, and logit[P(Yi=1)]= 1 2x1i ... k+1xk,i general population groups were pooled number of polyps were recorded. and together as the control group in the bivariate logistic regression analysis because β +β + +β Statistical Analysis logit[P(Wi=1)]= 1 2x1i ... k+1xk,i theywerenotstatisticallysignificantlydif- The polyp prevalence in Hong Kong was for i=1, ..., n. ferentintermsofsexandagedistribution, assumed to be 24% (including hyperplas- The odds ratio [OR] is defined as family history of colorectal cancer, smok- 15 tic polyps). Assuming an unadjusted = = × = = ing history, metabolic syndrome, or dia- 3 γ = P(Yi 1,Wi 1) P(Yi 0,Wi 0) oddsratioof2.0, thesamplesizerequired i = = × = = betes (TABLE 1). Our aim was to (1) as- to achieve a 2-tailed ␣level of significance P(Yi 1,Wi 0) P(Yi 0,Wi 1) sess if any association existed between ␤ of .05 with a power of 1− =.9 was 207 This OR measures the association of CAD and colorectal neoplasm and (2) in the CAD-positive, CAD-negative, and CAD and colorectal neoplasm. An OR identify the risk factors that predispose general population groups. γ to both diseases simultaneously. of i=1 corresponds to no association ␹2 The Pearson test or the Fisher ex- between CAD and colorectal neo- There are a number of potential co- act test (for small prevalences) was car- γ Ͼ variates, but not all of them are useful plasm, and i 1 means that a patient ried out to perform simple compari- with CAD is also more likely to have predictors in the marginal probabili- sons of prevalence based on various colorectal neoplasm, with larger val- ties of Y or W or the ORs of the 2 events. demographic and clinical characteris- γ The effects of age and sex were in- ues of i indicating a stronger associa- tics among participants in the CAD- tion. While it is possible to model this cluded in the modeling of the mar- positive, CAD-negative, and general γ γ ginal probabilities P(Y=1) and P(W=1) OR i as a constant , forcing it to be population groups. F tests based on a the same for all participants in the since they are well known to be the risk 1-way analysis of variance model was population, we chose to also let it de- factors that predispose to the develop- used to compare the ages of the pa- pend on covariates, thus allowing pa- ment of both CAD and colonic le- tients in the 3 groups. Descriptive sta- tients with certain characteristics to sions. We did not consider other fac- tistics and simple analyses were car- have a stronger association between tors in the marginal models because our ried out using JMP software, release CAD and colorectal neoplasm than oth- primary objective was to investigate the 6.0.2 (SAS Institute Inc, Cary, North ers. This relationship is mitigated effect of factors on the association be- Carolina). through the equation tween CAD and colorectal neoplasm or A bivariate logistic regression analy- advanced lesions, and the inclusion of 16,17 γ θ +␪ + +θ sis was carried out to examine the as- log( i)= 1 2x1i ... k+1xk,i. too many simultaneous factors in the sociation between colorectal neoplasm With this specification, the joint prob- model could result in colinearity or and CAD, adjusted accordingly for po- other difficulties fitting the model. ␹ ␹ abilities of Y and W are given by tential risk factors 1,..., k. In the cur- In the first phase of our analysis, we = = = rent application, we let Y=1 denote pres- P(Yi 1,Wi 1) simply assumed that the OR γ was a con- γ − = + = + − δ 1/2 ence of CAD and W=1 denote presence ( i 1)[P(Yi 1) P(Wi 1)] 1 i stant that, thus, was not affected by any γ − of colorectal neoplasm, while Y=0 and 2( i 1) covariates. We tested the null hypoth- W=0 denote the absence of CAD and co- where esis of no association between CAD and

Downloaded From: https://jamanetwork.com/ on 09/25/2021 COLORECTAL NEOPLASM IN NEWLY DIAGNOSED CORONARY ARTERY DISEASE

colorectal neoplasm using a Wald test. FIGURE. The final numbers of partici- ing to the Holm adjustment, we con- In the second phase of the bivariate lo- pants for analysis were 206, 208, and cluded that the mean age and distribu- gistic regression analysis, we allowed 207 in the CAD-positive, CAD- tions of family history of colorectal the OR γ to depend on some explana- negative, and general population cancer were insignificantly different tory variable that may play an impor- groups, respectively. among the 3 groups, but there were dif- tant role in mitigating the association The baseline demographic and clini- ferences in the distributions of sex, of the 2 diseases. The factors of age, sex, cal characteristics in each group are smoking history, the metabolic syn- metabolic syndrome, smoking his- summarized in TABLE 2. The CAD- drome, and the individual factors of the tory, and family history of colon can- positive and general population group metabolic syndrome among the 3 cer were included in the regression were age- and sex-matched. Accord- groups (Table 2). Eighty-five percent γ model for the log OR, log( i); while the effects of age and sex were adjusted in the modeling of P(Yi=1) and P(Wi=1). Figure. Study Recruitment and Participant Flow A backward-elimination procedure was adopted to remove the most insignifi- 706 Individuals assessed for eligibility cant variable in the regression model γ for log( i) at each step until the P values for the variables remained in the 245 CAD-positive 234 CAD-negative 227 From general population working model were all less than .10. The P values reported are based on 39 Excluded 26 Excluded 20 Excluded 33 Refused colonoscopy 23 Refused colonoscopy 16 Refused colonoscopy 2-tailed alternatives in the cases in which 6 Colonoscopy failed 3 Colonoscopy failed 4 Colonoscopy failed comparisons were made between 2 2 Developed bradycardia (sharp bend or excessive 1 Uncooperative and hypotension looping of the colon) 3 Sharp bend or excessive groups and for the regression coefficients 1 Had oxygen desaturation looping of the colon of the log OR in the bivariate logistic re- 2 Uncooperative 18 1 Sharp bend or excessive gression analyses. The method of Holm looping of the colon was used to adjust the P values in mul-

tiple testing, where appropriate, particu- 206 Included in analysis 208 Included in analysis 207 Included in analysis larly on the demographic variables and prevalence of colonic lesions of the 3 CAD indicates coronary artery disease. groups. For example, when reporting K P values for K distinct tests, the Holm th Table 2. Baseline Demographic and Clinical Characteristics of the CAD-Positive, method is to compare the r smallest P CAD-Negative, and General Population Groups value(for r=1,...,K)amongthe KPval- CAD- CAD- General P Value ues with .05/(K−rϩ1), and the test re- Positive Negative Population Comparing Adjusted Group Group Group the 3 P Value sult is considered statistically significant Characteristics (n = 206) (n = 208) (n = 207) Groupsa (Rank)b after adjustment for the multiple tests if Age, mean (SD), y 64.7 (8) 62.9 (8) 62.9 (8) .04 .07 (9) th the r smallest P value is less than .05/ Male, No. (%) 161 (78.2) 132 (63.5) 152 (73.4) .003 .009 (8)c ϩ th (K−r 1). However, if the r smallest P Metabolic syndrome, No. (%) 78 (37.9) 38 (18.3) 20 (9.7) Ͻ.001 .001 (1)c value is the first that exceeds .05/ Waist circumference Ն91.4 cm for men 77 (37.6) 74 (35.6) 47 (22.8) .002 .01 (6)c (K−rϩ1), then the test results associated or Ն81.3 cm in for women, No. (%) with the (K−rϩ1) largest P values are Fasting glucose Ն110 mg/dL, 87 (42.2) 42 (20.2) 42 (20.3) Ͻ.001 .001 (2)c considered statistically nonsignificant ac- No. (%) Ն c cording to the Holm method. To make Triglycerides 150 mg/dL, No. (%) 70 (34.0) 53 (25.5) 40 (19.3) .003 .01 (7) HDL-C Յ40 mg/dL for men or Յ50 78 (37.9) 61 (29.3) 28 (13.5) Ͻ.001 .001 (3)c the presentation simpler, we let the ad- pmg/dL for women, No. (%) ϩ justed P value be (K−r 1) times the Blood pressure Ն130/85 mm Hg, 88 (42.7) 54 (26.0) 28 (13.5) Ͻ.001 .001 (4)c original P value and simply compare the No. (%) adjusted P value with .05 to determine Ever smoking, No. (%) 78 (37.9) 37 (17.8) 45 (21.7) Ͻ.001 .001 (5)c whether a particular test result is statis- Family history of colorectal cancer, 12 (5.8) 12 (5.8) 24 (11.6) .04 .04 (10)d tically significant after adjustment. No. (%) Abbreviations: CAD, coronary artery disease; HDL-C, high-density lipoprotein cholesterol. SI conversions: To convert fasting glucose to mmol/L, multiply by 0.555; to convert triglycerides to mmol/L, multiply by RESULTS 0.0113; to convert HDL-C to mmol/L, multiply by 0.0259. a P values calculated from 1-way analysis of variance and/or ␹2 or Fisher exact tests. Baseline Characteristics b Adjusted P values using the Holm18 method for multiple testing, with the first entry being the adjusted P value and the rank in parentheses the rank of the associated original P value in ascending order from most to least significant. There were 706 participants assessed for c Statistically significant (PϽ.05) after adjustment for multiple tests using the Holm method. eligibility in the 3 groups. The re- d The variable of age (rank = 9) is not statistically significant at the PϽ.05 level and, hence, the family history of colorectal cancer (rank = 10) is not statistically significant using the Holm method. sponse rates are summarized in the

Downloaded From: https://jamanetwork.com/ on 09/25/2021 COLORECTAL NEOPLASM IN NEWLY DIAGNOSED CORONARY ARTERY DISEASE

Table 3. Type and Prevalence of Colonic Lesions in the CAD-Positive, CAD-Negative, and General Population Groups P Value No. (%) Comparing Adjusted P Value CAD-Positive P Value Comparing All 3 and General CAD-Positive CAD-Negative General Comparing Groups Using Population Groups Group Group Population Group the Prevalence Holm Method (Holm-Adjusted Type of Colonic Lesion (n = 206) (n = 208) (n = 207) in All 3 Groupsa (Rank)b P Value)c Endoscopic polyp 84 (40.8) 59 (28.4) 69 (33.3) .03 .055 (4) .12 (.24) Colorectal neoplasm 70 (34.0) 39 (18.8) 43 (20.8) Ͻ.001 .002 (2)d Ͻ.001 (.01) Hyperplastic polyp 17 (8.3) 10 (4.8) 15 (7.2) .33 .33 (5) .70 (.70) Advanced lesion 38 (18.4) 18 (8.7) 12 (5.8) Ͻ.001 .001 (1)d Ͻ.001 (.001) Colorectal cancer 9 (4.4) 1 (0.5) 3 (1.4) .02 .06 (3) .09 (.26) Abbreviation: CAD, coronary artery disease. a P values calculated from ␹2 or Fisher exact tests. b Adjusted P values using the Holm18 method for multiple testing, with the first entry being the adjusted P value and the rank in parentheses the rank of the associated original P value in ascending order from most to least significant. c P values in this column are for 2-tailed alternative hypotheses. d Statistically significant (PϽ.05) after adjustment for multiple tests using the Holm method.

of the patients in the CAD-positive lence of colorectal neoplasm (34.0% vs marginal probabilities for age and sex, group underwent revascularization and 20.8%; adjusted P=.01 by Pearson ␹2 we observed evidence of a strong asso- 10% underwent coronary artery by- test) and advanced lesions (18.4% vs ciation between colorectal neoplasm pass graft surgery; the remaining 5% 5.8%; adjusted PϽ.001 by Pearson ␹2 and CAD (OR, 1.88; 95% CI, 1.25- were treated noninvasively. None of the test) were still observed when compar- 2.70; 2-tailed P=.002; TABLE 4), and be- participants in the CAD-negative and ing only the CAD-positive and general tween advanced lesions and CAD (OR, general population groups underwent population groups, which were age- and 2.51; 95% CI, 1.43-4.35; 2-tailed revascularization or coronary artery by- sex-matched. P=.001; TABLE 5). If all explanatory pass graft surgery. The prevalence of the Of the 9 adenocarcinomas (4.4%) de- variables (age, sex, smoking status, fam- metabolic syndrome in the CAD- tected in the CAD-positive group, 8 ily history of colon cancer, and the positive patients in our study is simi- were asymptomatic. Five (55%) of the metabolic syndrome) were included in lar to that reported in the Asia/Pacific 9 cancers detected were at an early stage the estimate of the marginal probabili- 19,20 Ͻ region. (Dukes A or B) and were small ( 3 ties P(Yi=1) and P(Wi=1) and assum- cm). In contrast, only 1 cancer (0.5%) ing the OR did not depend on any ex- Prevalence and Characteristics and 3 cancers (1.4%) were detected in planatory variables, the constant OR for of Colorectal Neoplasms the CAD-negative and general popula- CAD and colorectal neoplasm is esti- The characteristics and prevalence of tion groups, respectively. All were mated to be 1.54 (95% CI, 1.04-2.30; colonic lesions in the 3 groups are sum- asymptomatic and 75% (3/4) were of 2-tailed P=.048). marized in TABLE 3 together with the small size (1-2 cm) and early stage In the second part of the bivariate lo- original and Holm-adjusted P values. (Dukes A). gistic regression analysis, we allowed Colorectal neoplasm and advanced le- the OR for the association of the 2 dis- sion were more prevalent (34.0% and Bivariate Logistic eases to depend on explanatory vari- 18.4%, respectively) in the CAD- Regression Analysis ables like age, sex, smoking status, positive group than in the CAD- The 2 control groups were merged for metabolic syndrome, and family his- negative (18.8% and 8.7%) and gen- the bivariate logistic regression analy- tory of colorectal cancer. To identify the eral population (20.8% and 5.8%) sis because no significant differences be- risk factors and/or predictors that pre- groups (PϽ.001 for both by Pearson tween them were found for any demo- dispose to the development of both ␹2 test and adjusted P =.002 and graphic or clinical characteristics (age, CAD and colorectal neoplasm, the vari- P=.001, respectively). In addition, using sex, family history of colorectal can- ables of age, sex, smoking history, meta- the Holm method, there was no differ- cer, smoking status, presence of the bolic syndrome, and family history of ence in the prevalence of endoscopic metabolic syndrome, and high fasting colorectal cancer were included as the polyps (40.8% vs 28.4% vs 33.3%; ad- glucose level) and the 2 groups ap- explanatory variables for the log of the justed P=.055 by Pearson ␹2 test), hy- peared to be reasonably homogeneous OR while adjusted for the variables of perplastic polyps (8.3% vs 4.8% vs (Table 1). age and sex in the marginal probabili- 7.2%; adjusted P=.33 by Fisher exact Assuming that the OR of the asso- ties P(Yi=1) and P(Wi=1). A backward- test), and adenocarcinoma (4.4% vs ciation of the 2 diseases was constant elimination procedure was adopted to 0.5% vs 1.4%; adjusted P=.06 by Fisher (ie, not affected by any independent remove the most nonsignificant vari- exact test). Differences in the preva- variables) when adjusting only for the able in the regression of the log of the

Downloaded From: https://jamanetwork.com/ on 09/25/2021 COLORECTAL NEOPLASM IN NEWLY DIAGNOSED CORONARY ARTERY DISEASE

OR at each step. The final model indi- cruited prospectively for colonos- neoplasm and CAD (OR, 1.88) and be- cated that the association between CAD copy. We observed that there was a tween advanced colonic lesion and CAD and colorectal neoplasm did not de- strong association between colorectal (OR, 2.51). The prevalence of colorec- pend on any of the aforementioned explanatory variables (PϾ.05) and that Table 4. Results of the Bivariate Logistic Regression Analysis on CAD and Colorectal the OR (estimated to be 1.88; Table 4) Neoplasm Assuming a Constant Odds Ratio ␥ was constant across all participants. Factor (␣, ␤,or␥) However, we found that both the (95% Confidence Interval) P Value metabolic syndrome and history of Coefficients for the marginal probability of CAD (␣) Ͻ smoking were strong independent pre- Intercept −4.53 (−5.77 to −3.21) .001 Ͻ dictive factors for the positive associa- Age/100 4.97 (2.94 to 6.79) .001 Ͻ tion between advanced lesions and CAD Male 0.98 (0.63 to 1.35) .001 Coefficients for the marginal probability (for metabolic syndrome, OR, 5.99; 95% of colorectal neoplasm (␤) CI, 1.43-28.0; 2-tailed P=.02; for smok- Intercept −4.97 (−6.32 to −3.54) Ͻ.001 ing, OR, 4.74; 95% CI, 1.38-19.0; Age/100 5.00 (2.80 to 7.02) Ͻ.001 2-tailed P=.02; TABLE 6). This indi- Male 0.88 (0.40 to 1.30) Ͻ.001 cates that participants who were smok- Coefficients for the log odds ratio (␥) ers and/or who had the metabolic syn- Intercept 0.61 (0.22 to 0.99) .002 drome were much more likely to Odds ratio 1.88 (1.25 to 2.70) .002 Abbreviation: CAD, coronary artery disease. develop both CAD and advanced lesions. The OR of the metabolic syndrome in predicting the positive asso- Table 5. Results of Bivariate Logistic Regression Analysis on CAD and Advanced Lesions ␥ ciation of the 2 diseases was higher than Assuming a Constant Odds Ratio Factor (␣, ␤,or␥) the OR of the individual components (95% Confidence Interval) P Value of the metabolic syndrome (TABLE 7) Coefficients for the marginal probability of CAD (␣) after adjustment for age and sex. The Intercept −3.98 (−5.19 to −2.89) Ͻ.001 wide CI for this OR was because the OR Age/100 4.14 (2.51 to 5.95) Ͻ.001 was larger in magnitude and fewer par- Male 0.96 (0.61 to 1.36) Ͻ.001 ticipants had the metabolic syndrome Coefficients for the marginal probability of advanced lesions (␤) than had its individual components. Intercept −5.06 (−7.03 to −3.43) Ͻ.001 Age/100 3.76 (1.28 to 6.62) .005 Complications Male 0.78 (0.18 to 1.40) .01 One patient from the CAD-positive Coefficients for the log odds ratio (␥) group developed colonic perforation Intercept 0.88 (0.35 to 1.47) .001 during polypectomy for a 2-cm sessile Odds ratio 2.51 (1.43 to 4.35) .001 villous adenoma. She recovered after Abbreviation: CAD, coronary artery disease. operation. Another patient in the CAD- positive group developed congestive Table 6. Final Results of the Bivariate Logistic Regression Analysis on CAD and Advanced heart failure after bowel preparation for Lesions Allowing the Odds Ratio ␥ to Depend on Some Explanatory Variables colonoscopy and was treated conser- Factor (␣, ␤,or␥) vatively. (95% Confidence Interval) P Value Coefficients for the marginal probability of CAD (␣) Ͻ COMMENT Intercept −4.01 (−5.11 to −2.46) .001 Age/100 4.10 (1.73 to 5.70) Ͻ.001 Previous studies have reported the as- Male 0.95 (0.57 to 1.34) Ͻ.001 sociation between colorectal neo- Coefficients for the marginal probability plasm and CAD,21-23 and some have re- of colorectal neoplasm (␤) futed the association.24 The current Intercept −5.96 (−7.93 to −3.42) Ͻ.001 study confirms the association be- Age/100 5.09 (1.10 to 8.02) .002 tween the 2 diseases and our previous Male 0.71 (0.09 to 1.36) .03 3 Coefficients for the log odds ratio (␥) retrospective study observation. The Intercept −0.29 (−1.35 to 0.62) .58 design of the current study is robust Smoking status 1.56 (0.32 to 2.94)a .02 21-24 compared to others in that CAD was Metabolic syndrome 1.79 (0.35 to 3.33)b .02 defined by coronary angiogram, which Abbreviation: CAD, coronary artery disease. is the criterion standard for diagnos- a Estimated odds ratio for smoking status: exp(1.56)= 4.74. b Estimated odds ratio for the metabolic syndrome: exp(1.79)=5.99. ing CAD, and that patients were re-

Downloaded From: https://jamanetwork.com/ on 09/25/2021 COLORECTAL NEOPLASM IN NEWLY DIAGNOSED CORONARY ARTERY DISEASE

nary angiogram. Similarly, few young Table 7. Age- and Sex-Adjusted Odds Ratios for Independent Predictive Factors for Association of Advanced Colonic Lesions and CAD women have CAD (the study group). No. (%) of Sample Odds Ratio (95% Therefore, we have included the age- With Characteristic Confidence Interval) P Valuea and sex-matched general population as Metabolic syndrome 136 (21.9) 5.99 (1.43-27.94)b .02 another control group. In addition, the Individual components of the factors of age and sex were further ad- metabolic syndrome Diabetes 171 (27.5) 2.39 (1.33-3.98) .002 justed in the marginal logistic regres- Hypertension 170 (27.4) 2.41 (1.33-4.03) .002 sion part of the analysis. We chose coro- Triglycerides 163 (26.2) 2.17 (1.21-3.74) .007 nary angiogram to be the diagnostic High-density lipoprotein 167 (26.9) 2.12 (1.15-3.47) .008 criterion for the presence of CAD be- cholesterol cause diagnosis based on symptoms Waist circumference 198 (32.0) 2.29 (1.29-3.72) .002 alone is not reliable and coronary an- Abbreviation: CAD, coronary artery disease. giogram has the highest sensitivity a P values are for 2-tailed alternative hypotheses. b The odds of the metabolic syndrome in predicting the association of the 2 diseases was higher than the odds of its among the tests (treadmill, thallium individual components after adjustment for age and sex in the marginal probabilities of the events. scan, or computed tomography angiogram). However, the current design tal neoplasm (34%) was much higher pivotal in the pathogenesis of athero- might have potential bias in that we in the CAD-positive group than in the sclerosis and, hence, CAD.32,33 Colorec- were only assessing the prevalence of CAD-negative group (18.8%) or in the tal cancer is also thought to progress colorectal neoplasm in CAD patients age- and sex-matched general popula- through the pathway of inflamma- who presented for coronary angio- tion group (20.8%), as well as that re- tion.34 This is evidenced by the devel- gram. The study might not be able to ported by Sung et al15 (21%). More- opment of colorectal cancer in patients estimate the true magnitude of asso- over, the prevalence of advanced lesions with ulcerative colitis. Inflammation, ciation between CAD and colorectal (18.4%) and adenocarcinoma (4.4%) which may result from underlying risk neoplasm because there might be a per- was also observed to be much higher factors, may be the culprit for the simul- centage of patients in the general popu- in the CAD-positive group. taneous development of the 2 condi- lation with CAD who have not had a The high prevalence of advanced le- tions. In addition, insulin resistance is coronary angiogram. However, the sions in CAD-positive patients in the recognized as an important metabolic study highlights the important point study is remarkable. This is much defect linking the components of the that, at least in those with CAD pre- higher than the 12.5% reported in the metabolic syndrome. The direct prolif- senting for coronary angiogram, a high general population in Hong Kong by erative/antiapoptotic effects of insulin prevalence of colorectal neoplasm was Sung et al15 and the 10.5% in the US and related insulinlike growth factor 1 observed. The predictive value of the general population.25 Importantly, we on both colorectal neoplasm and CAD metabolic syndrome and smoking on observed that the metabolic syndrome has gained support recently.35,36 Peroxi- predisposing the positive association of and smoking, after adjusting for age and some proliferator–activated receptor ␥ colorectal neoplasm and CAD is lim- sex, were important predictive factors may play an important role in meta- ited by the nature of the cross- for the association of advanced co- bolic syndrome to regulate metabolic sectional study. A prospective study lonic lesions and CAD. It has been well and vascular pathways. These may be evaluating the role of the metabolic syn- documented that old age and male sex important potential mechanisms for the drome and smoking on the 2 condi- are important risk factors for CAD and simultaneous evolution of the 2 dis- tions is desirable. colorectal neoplasm. However, these 2 eases. Interestingly, statins have been In this study population undergo- factors are not modifiable. But the meta- shown to have beneficial effects in both ing coronary angiography, the preva- bolic syndrome and smoking are envi- colorectal cancer and CAD, probably lence of colorectal neoplasm was greater ronmental factors that can be re- through an anti-inflammatory mecha- in patients with CAD. The association versed. Smoking has been demonstrated nism.37 Aspirin has long been proven to between the presence of advanced co- as a major risk factor in the develop- be beneficial in both conditions, albeit lonic lesions and CAD was stronger in ment of the 2 diseases.26,27 There are also through different mechanisms. persons with the metabolic syndrome reports on insulin resistance syn- It would be ideal to perform an age- and a history of smoking. drome and colorectal cancer,28,29 as well and sex-matched study in both the Author Contributions: Dr Chan had full access to all as metabolic syndrome, predisposing to CAD-positive and CAD-negative groups of the data in the study and takes responsibility for colorectal adenomas.30,31 to study the association of the 2 dis- the integrity of the data and the accuracy of the data analysis. Both colorectal neoplasm and CAD eases. However, in reality, this is nearly Study concept and design: A. O. O. Chan, Jim, Morris, probably develop through the mecha- impossible. Asymptomatic men with Lai, S. K. Lam. Acquisition of data: A. O. O. Chan, Jim, Siu, Tong, nism of chronic inflammation. Inflam- healthy coronary arteries (the control Ng, S. Y. Wong, C. K. Chan, Cheung, P. Chan, mation is now recognized as being group) would rarely undergo a coro- G. Wong, Yuen, Lau, Lee, B. C. Y. Wong.

Downloaded From: https://jamanetwork.com/ on 09/25/2021 COLORECTAL NEOPLASM IN NEWLY DIAGNOSED CORONARY ARTERY DISEASE

Analysis and interpretation of data: A. O. O. Chan, RB. Clustering of metabolic factors and coronary ar- with coronary heart disease. Am J Cardiol. 2001; K. F. Lam, Morris, Hui. tery disease. Arch Intern Med. 1999;159(10):1104- 87(4):459-462, A6. Drafting of the manuscript: A. O. O. Chan, Tong, 1109. 24. Neugut AI, Rosenberg DJ, Ahsan H, et al. Asso- C. K. Chan, Lai, P. Chan, Lau, Szeto, S. K. Lam. 10. Hospital Authority of Hong Kong Special Admin- ciation between coronary heart disease and cancers Critical revision of the manuscript for important in- istrative Region. Hong Kong Cancer Registry. http: of the breast, prostate, and colon. Cancer Epidemiol tellectual content: A. O. O. Chan, Jim, K. F. Lam, //www3.ha.org.hk/cancereg/. Accessed August 29, Biomarkers Prev. 1998;7(10):869-873. Morris, Siu, Ng, S. Y. Wong, Hui, Cheung, Yuen, Lee. 2007. 25. Lieberman DA, Prindiville S, Weiss DG, Willett W; Statistical analysis: A. O. O. Chan, K. F. Lam, Morris. 11. American Cancer Society. Cancer Facts and Fig- VA Cooperative Study Group 380. Risk factors for ad- Obtained funding: A. O. O. Chan, S. K. Lam. ures 2000. Atlanta, GA: American Cancer Society; vanced colonic neoplasia and hyperplastic polyps in asymp- Administrative, technical, or material support: 2000. tomatic individuals. JAMA. 2003;290(22):2959-2967. A. O. O. Chan. 12. Parkin DM. Global cancer statistics in the year 26. Lakier JB. Smoking and cardiovascular disease. Am Study supervision: A. O. O. Chan. 2000. Lancet Oncol. 2001;2(9):533-543. J Med. 1992;93(1A):8S-12S. Financial Disclosures: None reported. 13. National Cholesterol Education Program. 27. Giovannucci E. An updated review of the epide- Funding/Support: The study was partially sponsored Executive summary of the third report of the miological evidence that cigarette smoking increases by the Gastrointestinal Cancer Research Fund, Uni- National Cholesterol Education Program (NCEP) risk of colorectal cancer. Cancer Epidemiol Biomark- versity of Hong Kong. Expert Panel on Detection, Evaluation, and Treat- ers Prev. 2001;10(7):725-731. Role of the Sponsor: The Gastrointestinal Cancer Re- ment of High Blood Cholesterol in Adults (Adult 28. Bowers K, Albanes D, Limburg P, et al. A pro- search Fund had no role in the design and conduct of Treatment Panel III). JAMA. 2001;285(19):2486- spective study of anthropometric and clinical mea- the study, in the collection, analysis, and interpreta- 2497. surements associated with insulin resistance syn- tion of the data, or in the preparation, review, or ap- 14. Simmons DT, Harewood GC, Baron TH, et al. Im- drome and colorectal cancer in male smokers. Am J proval of the manuscript. pact of endoscopist withdrawal speed on polyp yield: Epidemiol. 2006;164(7):652-664. implications for optimal colonoscopy withdrawal 29. Ahmed RL, Schmitz KH, Anderson KE, Rosa- time. Aliment Pharmacol Ther. 2006;24(6):965- mond WD, Folsom AR. The metabolic syndrome and REFERENCES 971. risk of incident colorectal cancer. Cancer. 2006;107 15. Sung JJ, Chan FK, Leung WK, et al. Screening for (1):28-36. 1. Parkin DM, Bray F, Ferlay J, Pisani P. Global can- colorectal cancer in Chinese: comparison of fecal oc- 30. Morita T, Tabata S, Mineshita M, Mizoue T, Moore cer statistics, 2002. CA Cancer J Clin. 2005;55(2): cult blood test, flexible sigmoidoscopy, and colonoscopy. MA, Kono S. The metabolic syndrome is associated 74-108. Gastroenterology. 2003;124(3):608-614. with increased risk of colorectal adenoma develop- 2. Miniño AM , Heron MP, Murphy SL , Kochanek 16. Chan JS, Kuk AY, Bell J. A likelihood approach to ment: the Self-Defense Forces health study. Asian Pac KD. Deaths: final data for 2004. Natl Vital Stat Rep. analyzing longitudinal bivariate binary data. Biom J. J Cancer Prev. 2005;6(4):485-489. 2007;55(19):1-120. 1997;4:409-421. 31. Wang YY, Lin SY, Lai WA, Liu PH, Sheu WH. As- 3. Chan AO, Lam KF, Tong TS, et al. Coexistence be- 17. Ananth CV, Preisser JS. Bivariate logistic regres- sociation between adenomas of rectosigmoid colon tween colorectal cancer/adenoma and coronary ar- sion: modelling the association of small for gesta- and metabolic syndrome features in a Chinese tery disease: results from 1382 patients. Aliment Phar- tional age births in twin gestations. Stat Med. 1999; population. J Gastroenterol Hepatol. 2005;20(9): macol Ther. 2006;24(3):535-539. 18(15):2011-2023. 1410-1415. 4. Winawer SJ. Natural history of colorectal cancer. 18. Holm S. A simple sequentially rejective multiple 32. Tiong AY, Brieger D. Inflammation and coronary Am J Med. 1999;106(1A):3S-6S. test procedure. Scand J Stat. 1979;6:65-70. artery disease. Am Heart J. 2005;150(1):11-18. 5. Framingham Heart Study Web site. http://www 19. Kwon HS, Prak YM, Lee HJ, et al. Prevalence and 33. Sepulveda JL, Mehta JL. C-reactive protein and .framinghamheartstudy.org. Accessed August 30, clinical characteristics of the metabolic syndrome in cardiovascular disease: a critical appraisal. Curr Opin 2007. middle-aged Korean adults. Korean J Intern Med. 2005; Cardiol. 2005;20(5):407-416. 6. Multiple Risk Factor Intervention Trial Group. Mul- 20(4):310-316. 34. Clevers H. At the crossroads of inflammation and tiple Risk Factor Intervention Trial Proceedings of the 20. Lin RT, Lee WJ, Jeng CY, Sheu WH, Chen YT. cancer. Cell. 2004;118(6):671-674. Nutrition Behavioral Research Conference. Bethesda, Metabolic syndrome and its contribution to CAD in 35. Renehan AG, Zwahlen M, Minder C, O’Dwyer MD: National Institutes of Health; 1975. DHEW pub- non-DM participants. J Formos Med Assoc. 2004; ST, Shalet SM, Egger M. Insulin-like growth factor lication (NIH) 76-978. 103(4):317-320. (IGF)-I, IGF binding protein-3, and cancer risk: sys- 7. Ford ES, Giles WH, Dietz WH. Prevalence of the 21. Correa P, Strong JP, Johnson WD, et al. Athero- tematic review and meta-regression analysis. Lancet. metabolic syndrome among US adults: findings from sclerosis and polyps of the colon: quantification of pre- 2004;363(9418):1346-1353. the third National Health and Nutrition Examination cursors of coronary heart disease and colon cancer. 36. McEwen JE, Zimniak P, Mehta JL, Reis RJ. Mo- Survey. JAMA. 2002;287(3):356-359. J Chronic Dis. 1982;35(5):313-320. lecular pathology of aging and its implications for se- 8. Lakka H-M, Laaksonen DE, Lakka TA, et al. The 22. Neugut AI, Jacobson JS, Sherif G, et al. Coronary nescent coronary atherosclerosis. Curr Opin Cardiol. metabolic syndrome and total and cardiovascular dis- artery disease and colorectal neoplasia. Dis Colon 2005;20(5):399-406. ease mortality in middle-aged men. JAMA. 2002; Rectum. 1995;38(8):873-877. 37. Demierre MF, Higgins PD, Gruber SB, et al. Stat- 288(21):2709-2716. 23. Reicher-Reiss H, Jonas M, Goldbourt U, Boyko ins and cancer prevention. Nat Rev Cancer. 2005; 9. Wilson PW, Kannel WB, Silbershatz H, D’Agostino V, Modan B. Selectively increased risk of cancer in men 5(12):930-942.

Downloaded From: https://jamanetwork.com/ on 09/25/2021