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

Colorectal cancer is the second most prevalent cancer worldwide.¹ There were about 1 million new cases and 500 000 deaths due to colorectal cancer in 2002.¹ It has been estimated that 1 in 20 healthy individuals will eventually develop colorectal cancer. Coronary artery disease (CAD) is the single leading cause of death in the United States and other industrialized countries.² We previously published a retrospective study that reported a strong association between colorectal cancer/adenoma and CAD, possibly due to the sharing of common environmental risk factors.³

Colorectal cancer and CAD share similar environmental risks factors, such as diabetes mellitus; smoking; hyperlipidemia; sedentary lifestyle; high-fat, low-fiber diet; obesity; and hypertension.^4- 6 The metabolic syndrome is being increasingly recognized as a significant health hazard worldwide.⁷ It comprises a constellation of metabolic risk factors, including most of the underlying risk factors for both colorectal cancer and CAD: diabetes or impaired glucose tolerance, hypertriglyceridemia, low high-density lipoprotein cholesterol level, central obesity, and hypertension. Persons with the metabolic syndrome have been reported to have increased risk of developing CAD.^8- 9 We postulated that the metabolic syndrome might also be an important risk factor for the development of both colorectal cancer and CAD.

Hong Kong is an industrialized region with incidences of and mortality due to colorectal cancer and CAD similar to that in western countries.^10- 12 Although we observed an association between colorectal cancer and CAD in our previous study,³ we were not able to identify the risk factors involved because of its retrospective nature.³ We thus designed and conducted the current cross-sectional study, the primary aim of which was to investigate the prevalence of colorectal cancer and adenoma (colorectal neoplasms) in patients with newly diagnosed CAD. A secondary aim was to identify the underlying risk factors, after adjusting for age and sex, that predisposed to the 2 conditions. The results have important implications in prevention of both colorectal neoplasm and CAD, as well as for the screening strategy of colorectal cancer.

The study period encompassed November 2004 to June 2006. Participating hospitals include the cardiac divisions of Queen Mary Hospital, Grantham Hospital, and Ruttonjee Hospital. These are regional hospitals serving patients with heart disease in the Hong Kong West cluster.

The study was approved by the institutional review boards of the University of Hong Kong/Queen Mary Hospital, Grantham Hospital, and Ruttonjee Hospital, all in Hong Kong. All participants provided written informed consent.

Consecutive patients with suspected CAD (ie, those with angina or abnormal exercise stress test results) who presented for the first time for coronary angiography were invited to participate in the study. Because we aimed at studying the prevalence of colorectal neoplasm in patients with newly diagnosed CAD, and to avoid the potential protective effect of aspirin and statins on colorectal neoplasm, patients with a history of CAD for more than 1 year or who had been taking aspirin or a statin for more than 1 year were excluded. In addition, those presenting for coronary angiography for reasons other than suspected CAD (eg, congestive heart failure, cardiomyopathy) were also excluded. Additional exclusion criteria were history of colonic disease, such as colorectal cancer, polyp, and inflammatory bowel disease, and history of colorectal surgery or colonoscopy within the previous 10 years.

Consecutive patients with coronary angiogram were invited for colonoscopy regardless of hemoglobin status to avoid preselection in the CAD-positive group of patients with gastrointestinal tract bleeding due to aspirin or clopidogrel therapy. In accordance with American College of Cardiology/American Heart Association guidelines, patients were defined as CAD-positive if at least 50% diameter stenosis in any 1 of the major coronary arteries was found on coronary angiography; otherwise, patients were defined as CAD-negative.

Because age and sex matching was not available in the CAD-negative group, a second control group from the general population was recruited. They were healthy participants who were asymptomatic other than having functional epigastric pain but with a normal result on upper endoscopy performed in our unit. They were age- and sex-matched to the CAD-positive group within 5 years on a nearly case-to-case basis. This group did not undergo coronary angiogram. The same exclusion criteria were also applied in this second control group.

All individuals in the 3 groups were invited to participate. Information on age, sex, history of smoking, diabetes mellitus, hypertension, family history of colorectal cancer, and use and duration of aspirin and statins was recorded. Waist circumference was measured. Blood was drawn and fasting glucose level and complete lipid profile were measured. The metabolic syndrome was defined as at least 3 of the following criteria set forth by the National Cholesterol Education Program's modified Adult Treatment Panel III (Asian Pacific Region criteria)¹³: (1) abdominal obesity: waist circumference of at least 36 in (91 cm) for men and 32 in (81 cm) for women; (2) low high-density lipoprotein cholesterol level: 40 mg/dL (1.03 mmol/L) or lower for men and 50 mg/dL (1.3 mmol/L) or lower for women; (3) hypertriglyceridemia: 150 mg/dL (1.7 mmol/L) or higher; (4) hypertension: blood pressure 130/85 mm Hg or higher; and (5) impaired glucose tolerance: fasting glucose 110 mg/dL (6.1 mmol/L) or higher.

Colonoscopy was scheduled within 8 weeks after assessing for eligibility or after revascularization for critical patients. All patients received the same bowel preparation. Colonoscopies were repeated the next day for those with poor bowel preparation. Patients taking clopidogrel had their treatment changed to subcutaneous heparin 2 days before colonoscopy. Endoscopists were blinded to the CAD status of the patients. The withdrawal time of the colonoscopy procedure was more than 6 min to minimize the chance of lesions being missed.¹⁴ Incomplete examination was excluded from analysis.

Advanced colonic lesion was defined as presence of cancer or adenomas with villous component, with high-grade dysplasia, or 1 cm or larger. The size of a polyp was assessed by comparing it with an open colonoscope biopsy forcep (Bard, Murray Hill, NJ). The site of colorectal neoplasm was defined as right-sided if proximal to splenic flexure. The size, site, histologic findings, and number of polyps were recorded.

The polyp prevalence in Hong Kong was assumed to be 24% (including hyperplastic polyps).¹⁵ Assuming an unadjusted odds ratio of 2.0,³ the sample size required to achieve a 2-tailed α level of significance of .05 with a power of 1−β = .9 was 207 in the CAD-positive, CAD-negative, and general population groups.

The Pearson χ² test or the Fisher exact test (for small prevalences) was carried out to perform simple comparisons of prevalence based on various demographic and clinical characteristics among participants in the CAD-positive, CAD-negative, and general population groups. F tests based on a 1-way analysis of variance model was used to compare the ages of the patients in the 3 groups. Descriptive statistics and simple analyses were carried out using JMP software, release 6.0.2 (SAS Institute Inc, Cary, North Carolina).

A bivariate logistic regression analysis^16- 17 was carried out to examine the association between colorectal neoplasm and CAD, adjusted accordingly for potential risk factors χ₁, . . . , χ_k. In the current application, we let Y = 1 denote presence of CAD and W = 1 denote presence of colorectal neoplasm, while Y = 0 and W = 0 denote the absence of CAD and colorectal neoplasm, respectively. The marginal probabilities of Y and W were modeled by the usual logistic regression analysis with models

logit[P(Y_i = 1)] = α₁ + α₂x_1i +  . . .  + α_k+1x_k,i

and

logit[P(W_i = 1)] = β₁ + β₂x_1i +  . . .  +β_k+1x_k,i for i = 1, . . . ,  n.

The odds ratio [OR] is defined as

This OR measures the association of CAD and colorectal neoplasm. An OR of γ_i = 1 corresponds to no association between CAD and colorectal neoplasm, and γ_i>1 means that a patient with CAD is also more likely to have colorectal neoplasm, with larger values of γ_i indicating a stronger association. While it is possible to model this OR γ_i as a constant γ, forcing it to be the same for all participants in the population, we chose to also let it depend on covariates, thus allowing patients with certain characteristics to have a stronger association between CAD and colorectal neoplasm than others. This relationship is mitigated through the equation

log(γ_i) = θ₁ + θ₂x_1i +  . . .  + θ_k+1x_k,i

With this specification, the joint probabilities of Y and W are given by

where

δ_i = 1 + (γ_i − 1){γ_i[P(Y_i = 1) − P(W_i = 1)]² − [P(Y_i = 1)+P(W_i = 1)]² + 2[P(Y_i = 1) + P(W_i = 1)]}

and

P(Y_i = 1,W_i = 0) = P(Y_i = 1) − P(Y_i = 1,W_i = 1),

P(Y_i = 0,W_i = 1) = P(W_i = 1) − P(Y_i = 1,W_i = 1),

P(Y_i = 0,W_i = 0) = 1 − P(Y_i = 1) − P(W_i = 1) + P(Y_i = 1,W_i = 1).

Maximum likelihood estimates of the regression coefficients, namely α, β, and θ, were computed using a tailor-made computer program in FORTRAN language. The 95% confidence intervals (CIs) were computed based on the asymptotic χ² distributed properties of the log-likelihood ratio. The CAD-negative and general population groups were pooled together as the control group in the bivariate logistic regression analysis because they were not statistically significantly different in terms of sex and age distribution, family history of colorectal cancer, smoking history, metabolic syndrome, or diabetes (Table 1). Our aim was to (1) assess if any association existed between CAD and colorectal neoplasm and (2) identify the risk factors that predispose to both diseases simultaneously.

There are a number of potential covariates, but not all of them are useful predictors in the marginal probabilities of Y or W or the ORs of the 2 events. The effects of age and sex were included in the modeling of the marginal probabilities P(Y = 1) and P(W = 1) since they are well known to be the risk factors that predispose to the development of both CAD and colonic lesions. We did not consider other factors in the marginal models because our primary objective was to investigate the effect of factors on the association between CAD and colorectal neoplasm or advanced lesions, and the inclusion of too many simultaneous factors in the model could result in colinearity or other difficulties fitting the model.

In the first phase of our analysis, we simply assumed that the OR γ was a constant that, thus, was not affected by any covariates. We tested the null hypothesis of no association between CAD and colorectal neoplasm using a Wald test. In the second phase of the bivariate logistic regression analysis, we allowed the OR γ to depend on some explanatory variable that may play an important role in mitigating the association of the 2 diseases. The factors of age, sex, metabolic syndrome, smoking history, and family history of colon cancer were included in the regression model for the log OR, log(γ_i); while the effects of age and sex were adjusted in the modeling of P(Y_i = 1) and P(W_i = 1). A backward-elimination procedure was adopted to remove the most insignificant variable in the regression model for log(γ_i) at each step until the P values for the variables remained in the working model were all less than .10.

The P values reported are based on 2-tailed alternatives in the cases in which comparisons were made between 2 groups and for the regression coefficients of the log OR in the bivariate logistic regression analyses. The method of Holm¹⁸ was used to adjust the P values in multiple testing, where appropriate, particularly on the demographic variables and prevalence of colonic lesions of the 3 groups. For example, when reporting KP values for K distinct tests, the Holm method is to compare the r^th smallest P value (for r = 1, . . . , K) among the KP values with .05/(K − r + 1), and the test result is considered statistically significant after adjustment for the multiple tests if the r^th smallest P value is less than .05/(K − r + 1). However, if the r^th smallest P value is the first that exceeds .05/(K − r + 1), then the test results associated with the (K − r + 1) largest P values are considered statistically nonsignificant according to the Holm method. To make the presentation simpler, we let the adjusted P value be (K − r + 1) times the original P value and simply compare the adjusted P value with .05 to determine whether a particular test result is statistically significant after adjustment.

There were 706 participants assessed for eligibility in the 3 groups. The response rates are summarized in the Figure. The final numbers of participants for analysis were 206, 208, and 207 in the CAD-positive, CAD-negative, and general population groups, respectively.

The baseline demographic and clinical characteristics in each group are summarized in Table 2. The CAD-positive and general population group were age- and sex-matched. According to the Holm adjustment, we concluded that the mean age and distributions of family history of colorectal cancer were insignificantly different among the 3 groups, but there were differences in the distributions of sex, smoking history, the metabolic syndrome, and the individual factors of the metabolic syndrome among the 3 groups (Table 2). Eighty-five percent of the patients in the CAD-positive group underwent revascularization and 10% underwent coronary artery bypass graft surgery; the remaining 5% were treated noninvasively. None of the participants in the CAD-negative and general population groups underwent revascularization or coronary artery bypass graft surgery. The prevalence of the metabolic syndrome in the CAD-positive patients in our study is similar to that reported in the Asia/Pacific region.^19- 20

The characteristics and prevalence of colonic lesions in the 3 groups are summarized in Table 3 together with the original and Holm-adjusted P values. Colorectal neoplasm and advanced lesion were more prevalent (34.0% and 18.4%, respectively) in the CAD-positive group than in the CAD-negative (18.8% and 8.7%) and general population (20.8% and 5.8%) groups (P < .001 for both by Pearson χ² test and adjusted P = .002 and P = .001, respectively). In addition, using the Holm method, there was no difference in the prevalence of endoscopic polyps (40.8% vs 28.4% vs 33.3%; adjusted P = .055 by Pearson χ² test), hyperplastic polyps (8.3% vs 4.8% vs 7.2%; adjusted P = .33 by Fisher exact test), and adenocarcinoma (4.4% vs 0.5% vs 1.4%; adjusted P = .06 by Fisher exact test). Differences in the prevalence of colorectal neoplasm (34.0% vs 20.8%; adjusted P = .01 by Pearson χ² test) and advanced lesions (18.4% vs 5.8%; adjusted P<.001 by Pearson χ² test) were still observed when comparing only the CAD-positive and general population groups, which were age- and sex-matched.

Of the 9 adenocarcinomas (4.4%) detected in the CAD-positive group, 8 were asymptomatic. Five (55%) of the 9 cancers detected were at an early stage (Dukes A or B) and were small (<3 cm). In contrast, only 1 cancer (0.5%) and 3 cancers (1.4%) were detected in the CAD-negative and general population groups, respectively. All were asymptomatic and 75% (3/4) were of small size (1-2 cm) and early stage (Dukes A).

The 2 control groups were merged for the bivariate logistic regression analysis because no significant differences between them were found for any demographic or clinical characteristics (age, sex, family history of colorectal cancer, smoking status, presence of the metabolic syndrome, and high fasting glucose level) and the 2 groups appeared to be reasonably homogeneous (Table 1).

Assuming that the OR of the association of the 2 diseases was constant (ie, not affected by any independent variables) when adjusting only for the marginal probabilities for age and sex, we observed evidence of a strong association between colorectal neoplasm and CAD (OR, 1.88; 95% CI, 1.25-2.70; 2-tailed P = .002; Table 4), and between advanced lesions and CAD (OR, 2.51; 95% CI, 1.43-4.35; 2-tailed P = .001; Table 5). If all explanatory variables (age, sex, smoking status, family history of colon cancer, and the metabolic syndrome) were included in the estimate of the marginal probabilities P(Y_i = 1) and P(W_i = 1) and assuming the OR did not depend on any explanatory variables, the constant OR for CAD and colorectal neoplasm is estimated to be 1.54 (95% CI, 1.04-2.30; 2-tailed P = .048).

In the second part of the bivariate logistic regression analysis, we allowed the OR for the association of the 2 diseases to depend on explanatory variables like age, sex, smoking status, metabolic syndrome, and family history of colorectal cancer. To identify the risk factors and/or predictors that predispose to the development of both CAD and colorectal neoplasm, the variables of age, sex, smoking history, metabolic syndrome, and family history of colorectal cancer were included as the explanatory variables for the log of the OR while adjusted for the variables of age and sex in the marginal probabilities P(Y_i = 1) and P(W_i = 1). A backward-elimination procedure was adopted to remove the most nonsignificant variable in the regression of the log of the OR at each step. The final model indicated that the association between CAD and colorectal neoplasm did not depend on any of the aforementioned explanatory variables (P > .05) and that the OR (estimated to be 1.88; Table 4) was constant across all participants.

However, we found that both the metabolic syndrome and history of smoking were strong independent predictive factors for the positive association between advanced lesions and CAD (for metabolic syndrome, OR, 5.99; 95% CI, 1.43-28.0; 2-tailed P = .02; for smoking, OR, 4.74; 95% CI, 1.38-19.0; 2-tailed P = .02; Table 6). This indicates that participants who were smokers and/or who had the metabolic syndrome were much more likely to develop both CAD and advanced lesions. The OR of the metabolic syndrome in predicting the positive association of the 2 diseases was higher than the OR of the individual components of the metabolic syndrome (Table 7) after adjustment for age and sex. The wide CI for this OR was because the OR was larger in magnitude and fewer participants had the metabolic syndrome than had its individual components.

One patient from the CAD-positive group developed colonic perforation during polypectomy for a 2-cm sessile villous adenoma. She recovered after operation. Another patient in the CAD-positive group developed congestive heart failure after bowel preparation for colonoscopy and was treated conservatively.

Previous studies have reported the association between colorectal neoplasm and CAD,^21- 23 and some have refuted the association.²⁴ The current study confirms the association between the 2 diseases and our previous retrospective study observation.³ The design of the current study is robust compared to others^21- 24 in that CAD was defined by coronary angiogram, which is the criterion standard for diagnosing CAD, and that patients were recruited prospectively for colonoscopy. We observed that there was a strong association between colorectal neoplasm and CAD (OR, 1.88) and between advanced colonic lesion and CAD (OR, 2.51). The prevalence of colorectal neoplasm (34%) was much higher in the CAD-positive group than in the CAD-negative group (18.8%) or in the age- and sex-matched general population group (20.8%), as well as that reported by Sung et al¹⁵ (21%). Moreover, the prevalence of advanced lesions (18.4%) and adenocarcinoma (4.4%) was also observed to be much higher in the CAD-positive group.

The high prevalence of advanced lesions in CAD-positive patients in the study is remarkable. This is much higher than the 12.5% reported in the general population in Hong Kong by Sung et al¹⁵ and the 10.5% in the US general population.²⁵ Importantly, we observed that the metabolic syndrome and smoking, after adjusting for age and sex, were important predictive factors for the association of advanced colonic lesions and CAD. It has been well documented that old age and male sex are important risk factors for CAD and colorectal neoplasm. However, these 2 factors are not modifiable. But the metabolic syndrome and smoking are environmental factors that can be reversed. Smoking has been demonstrated as a major risk factor in the development of the 2 diseases.^26- 27 There are also reports on insulin resistance syndrome and colorectal cancer,^28- 29 as well as metabolic syndrome, predisposing to colorectal adenomas.^30- 31

Both colorectal neoplasm and CAD probably develop through the mechanism of chronic inflammation. Inflammation is now recognized as being pivotal in the pathogenesis of atherosclerosis and, hence, CAD.^32- 33 Colorectal cancer is also thought to progress through the pathway of inflammation.³⁴ This is evidenced by the development of colorectal cancer in patients with ulcerative colitis. Inflammation, which may result from underlying risk factors, may be the culprit for the simultaneous development of the 2 conditions. In addition, insulin resistance is recognized as an important metabolic defect linking the components of the metabolic syndrome. The direct proliferative/antiapoptotic effects of insulin and related insulinlike growth factor 1 on both colorectal neoplasm and CAD has gained support recently.^35- 36 Peroxisome proliferator–activated receptor γ may play an important role in metabolic syndrome to regulate metabolic and vascular pathways. These may be important potential mechanisms for the simultaneous evolution of the 2 diseases. Interestingly, statins have been shown to have beneficial effects in both colorectal cancer and CAD, probably through an anti-inflammatory mechanism.³⁷ Aspirin has long been proven to be beneficial in both conditions, albeit through different mechanisms.

It would be ideal to perform an age- and sex-matched study in both the CAD-positive and CAD-negative groups to study the association of the 2 diseases. However, in reality, this is nearly impossible. Asymptomatic men with healthy coronary arteries (the control group) would rarely undergo a coronary angiogram. Similarly, few young women have CAD (the study group). Therefore, we have included the age- and sex-matched general population as another control group. In addition, the factors of age and sex were further adjusted in the marginal logistic regression part of the analysis. We chose coronary angiogram to be the diagnostic criterion for the presence of CAD because diagnosis based on symptoms alone is not reliable and coronary angiogram has the highest sensitivity among the tests (treadmill, thallium scan, or computed tomography angiogram). However, the current design might have potential bias in that we were only assessing the prevalence of colorectal neoplasm in CAD patients who presented for coronary angiogram. The study might not be able to estimate the true magnitude of association between CAD and colorectal neoplasm because there might be a percentage of patients in the general population with CAD who have not had a coronary angiogram. However, the study highlights the important point that, at least in those with CAD presenting for coronary angiogram, a high prevalence of colorectal neoplasm was observed. The predictive value of the metabolic syndrome and smoking on predisposing the positive association of colorectal neoplasm and CAD is limited by the nature of the cross-sectional study. A prospective study evaluating the role of the metabolic syndrome and smoking on the 2 conditions is desirable.

In this study population undergoing coronary angiography, the prevalence of colorectal neoplasm was greater in patients with CAD. The association between the presence of advanced colonic lesions and CAD was stronger in persons with the metabolic syndrome and a history of smoking.