Because of the discrepancies in the relation between serum cholesterol and death from colorectal cancer, we decided to approach the problem by studying colorectal adenomas, the benign precursors of colorectal carcinoma. Much evidence has emerged to suggest that most adenocarcinomas of the large bowel arise from adenomatous polyps [15, 16], a process referred to as the adenoma-dysplasia-carcinoma sequence [17]. An investigation of the clinical precursors may elucidate risk factors in the pathogenesis of large-bowel cancer while avoiding the possible influence of the metabolic consequences of an established cancer. Using this approach, we recently showed a positive association between the serum cholesterol level and the frequency of colorectal adenoma in men alone and in both sexes combined [18]. We have now done a cross-sectional study of the association between the prevalence of colorectal adenoma and serum lipoprotein levels.

Methods

Top Methods Results Discussion Author & Article Info References

Patient Evaluation

Between March 1988 and February 1989, we studied 1124 inpatients and outpatients undergoing colonoscopy up to and including the cecum. We obtained personal and medical histories from all patients and measured their height and weight. Indications for colonoscopy included occult or overt blood in the stools (but not acute intestinal bleeding); abdominal pain of unknown origin; diarrhea; constipation; suspected neoplasia (unexplained anemia and weight loss); and follow-up after resection for large-bowel cancer (but not in the immediate postoperative period); follow-up after polypectomy; and follow-up of chronic inflammatory bowel disease. Patients serving as controls in our study had to have been free of adenomas at previous examinations.

Patients with previous or current malignant disease, chronic inflammatory bowel disease, familial polyposis, or partial colectomy were excluded from the study. We also excluded from analysis patients on a strict cholesterol-lowering diet, patients on cholesterol-lowering medication, and patients who had become vegetarians less than 5 years previously. These exclusion criteria were stipulated in the protocol before the study was begun.

Colonoscopy was done in our endoscopy unit with Olympus CF-IB, CF-l0I, and CF-20I fiberoptic colonoscopes (Olympus, Hamburg, Germany). In addition to performing a macroscopic evaluation, we removed or biopsied polyps and tumors endoscopically or surgically. Adenomas were classified histologically as tubular, tubulovillous, or villous, according to the criteria of the World Health Organization [19]. Epithelial dysplasia of removed adenomas was graded as mild, moderate, or severe [19]. Patients with villous adenomas, tubulovillous adenomas, adenomas with severe dysplasia, adenomas with a diameter of more than 2 cm, and multiple adenomas (more than six) were classified as being at high risk because such patients have an increased risk for developing malignancy [20].

Fasting blood samples were collected before preparation for colonoscopy. More than 90% of the blood collections were done between 0800 hours and 0900 hours on the day of admission, and the remainder were done on the day after admission during the same period. The patients with and without adenoma were homogeneous with regard to blood sampling on the first and second days of hospitalization.

Blood sampling for serum cholesterol and serum lipoprotein determinations was done after patients had fasted for at least 12 hours. Very low-density lipoprotein (d < 1.006 g/mL) was isolated from serum by ultracentrifugation at 105 000 g for 22 h at 4 °C in a Beckman Ti 40.3 rotor (Beckman, Munich, Germany), and the lipoprotein fractions were recovered by tube slicing. In the infranatant, HDL was separated from LDL by precipitation using sodium phosphotungstate and magnesium chloride. Levels of cholesterol and triglycerides in whole serum and of the separated lipoproteins were determined in an automated analyzer using enzymatic techniques (EPOS, Eppendorf, Hamburg, Germany [kits from Boehringer Mannheim, Germany]).

All determinations were done in duplicate and did not differ by more than 1%. Throughout the study, external quality control was done with serum material from the National Center of Laboratory Quality Control (INSTAND).

Statistical Analysis

Patients to be evaluated were divided into two groups: those with adenoma and those with no adenomatous polyps (controls). Patients with and without colorectal adenomas were compared with respect to total serum cholesterol level; serum HDL, LDL, and VLDL cholesterol levels; age; sex; and body mass index (ratio of weight to the square of the height [kg/m²]).

The Wilcoxon test and the chi-square test were used to examine continuous variables and frequency data, respectively.

The association of the total serum cholesterol, HDL, LDL, or VLDL level to the frequency (prevalence) of colorectal adenomas was analyzed in men and women separately and in both sexes combined. The significance of the association between each variable and the frequency of colorectal adenomas and of high-risk adenomas was analyzed by logistic regression analysis using the BMDP-LR program (BMDP Statistical Software, Berkeley, California). Adenoma frequencies were analyzed by lipoprotein quintile. All odds ratios were calculated using the coefficients and standard errors of the respective variable in the logistic regression analysis. The 95% CIs were calculated on the basis of the 95% range of each variable. All P values are two-tailed.

Results

Top Methods Results Discussion Author & Article Info References

Of the 1124 patients examined, 302 were excluded from the study: 148 with malignant disease, 44 of whom either had current colorectal cancer or had colorectal cancer in the past; 24 with ulcerative colitis; 36 with Crohn disease; and 6 with wasting diseases. Patients on a strict cholesterol-lowering diet (n = 68), patients taking medication (n = 7), and patients who were vegetarians (n = 13) were also excluded.

Of the 822 patients included in the study, 194 had adenomas. Characteristics of men and women with and without adenomas are shown in Table 1. The age and sex distributions of serum lipoprotein levels in patients who had colonoscopy were similar to those of all hospital patients and to those observed in a population-based analysis.

We analyzed adenoma frequency according to quintile of total cholesterol, LDL cholesterol, HDL cholesterol, and VLDL cholesterol level for men, for women, and for both sexes combined. Thus, results of the univariate analyses could be compared with data from other studies. In addition, such an approach would improve the clarity of the evaluation.

The relation between the total cholesterol level and frequency of colorectal adenomas was significant in men (odds ratio, 1.66; P < 0.05) and slightly positive but not significant in women (odds ratio, 1.47) (Table 2).

We found an inverse relation between the HDL cholesterol level and the frequency of colorectal adenoma (Table 3). This association was significant in women (odds ratio, 0.58; P < 0.025) and showed a trend toward significance in men (odds ratio, 0.79).

The analysis of the relation between the LDL or the VLDL cholesterol serum level and the frequency of colorectal adenomas showed a positive association (Table 4). Odds ratios for the relation between LDL cholesterol and the frequency of colorectal adenomas were as follows: 1.36 in men (P > 0.2) and 2.1 in women (P < 0.01) (Table 4). Corresponding odds ratios for VLDL cholesterol were 1.31 in men (P < 0.01) and 1.27 in women (P = 0.18) (Table 5).

A logistic regression analysis was done to assess the relative strength of the associations of the individual serum lipoprotein levels with the frequency of colorectal adenoma. In descending order, the relative strength of the individual serum lipoprotein levels was as follows: HDL (P = 0.0001); LDL (P = 0.001); VLDL (P = 0.006); and total cholesterol (P = 0.008).

We analyzed the relation between high-risk adenoma and the serum lipoprotein levels; the presence of such an adenoma increases the risk for malignant degeneration. Eighty-nine of 194 patients with adenoma were classified as being at an increased risk for malignant degeneration. Because of the small number of patients with high-risk adenomas, we restricted our analysis to the total patient group. The logistic regression analysis showed no clear association between the frequency of high-risk colorectal adenoma and the total serum cholesterol and LDL cholesterol levels (Table 6). A significant inverse association between the HDL cholesterol levels and the frequency of high-risk colorectal adenoma was observed (odds ratio, 0.51; P < 0.05). A slight positive association was found between VLDL cholesterol and the presence of high-risk colorectal adenoma (Table 6). The relative strength (in descending order) of the association between individual serum lipoprotein levels and the frequency of high-risk colorectal adenoma was as follows: HDL (P = 0.025); LDL (P = 0.15); VLDL (P = 0.17); and total cholesterol (P > 0.2).

Discussion

Top Methods Results Discussion Author & Article Info References

A similarity in the epidemiology of colorectal cancer and coronary heart disease has been noted by several investigators [1, 21-23]. That the two diseases have common etiologic factors has been further suggested by a strong correlation between the frequency of the two diseases [24, 25] in different countries. Death from both diseases is strongly correlated with the per capita consumption of total calories, total fat, saturated fat, and animal proteins [23, 26]. However, several investigators [1-8, 27-29] have reported a negative correlation between the serum cholesterol level and the risk for colorectal cancer but a positive correlation between the risk for coronary heart disease and serum cholesterol [21, 24, 29, 30]. Two hypotheses have been advanced to explain this discrepancy.

Because patients with advanced colorectal cancer are known to have significantly lower serum cholesterol levels [31, 32], the inverse association reported in some studies might be attributable to the inclusion of patients with cancer that was undetected at the baseline examination. Indeed, the inverse relation virtually disappeared in some [5-7], but not all, studies [2-4] when cases of cancer diagnosed early in the study—up to 7 years after the initiation of study [6]—were excluded from the analysis. However, the discrepancy between the results of cohort or case–control studies and epidemiologic data remains.

Another methodologic question that has been raised is whether the inverse relation could be accounted for by higher rates of death from cardiovascular disease among patients with higher cholesterol levels. Cohort studies of death from coronary heart disease are especially prone to this form of bias. When all patients in a cohort were followed until death, those at low risk for coronary heart disease could be expected to have a high risk for dying later from other causes, such as colorectal cancer. However, when standard actuarial techniques were used, the inverse relation in two such studies could not be accounted for by the competing risk from cardiovascular disease [2, 4].

To clarify these inconsistencies, we conducted a new type of study; we investigated patients with colorectal adenomas, which may be considered benign precursors in nearly all patients with colorectal cancer [15-17, 33]. The main advantage of this approach is the significantly lower age of patients in our study. Because death from colorectal cancer peaks in the seventh decade of life [34], the frequency of colorectal cancer may be influenced by numerous other diseases with an earlier peak mortality (fifth or sixth decade of life). A major bias of previous studies might be that they did not exclude other deaths, in particular from coronary heart disease, for the calculation of deaths from colon cancer. In our study, patients had a mean age of 53 years, which is about 15 years younger than the mean age of patients with colon cancer.

Patient age is the leading risk factor for the development of colorectal adenoma. Patient age is, moreover, strongly correlated with serum total cholesterol, LDL cholesterol, and VLDL cholesterol levels and is inversely correlated with the serum HDL cholesterol level [18]. Thus, all frequency data in the analysis of patients by lipoprotein quintile were adjusted for age. The mean values for lipoprotein levels in patients with and without adenomas (see Table 1) were corrected using logistic regression analysis. The age and sex distributions of serum lipoprotein levels in our group of patients who underwent colonoscopy were found to be similar to those in all hospitalized patients and to those observed in a population-based study. Further, the indications for colonoscopy, which are associated with different adenoma frequencies, were similarly distributed throughout the different age groups. Both findings suggest that no major selection bias operated in our study.

The association between the total cholesterol level and the frequency of adenomas was similar to that observed by Mannes and coworkers [18]. In our study, the odds ratios for the highest cholesterol quintile relative to the lowest quintile were 1.36 for all patients, 1.66 for men, and 1.47 for women; the corresponding values were 2.0, 2.2, and 1.7 in the study by Mannes and colleagues [18], who investigated 842 patients and calculated the odds ratios on the basis of highest and lowest quintiles. In our study, the relative risk expressed by the odds ratio was significant only for men; however, in the study by Mannes and colleagues, the relative risk was significant for all patients and for men. The relative risk for colorectal adenomas in the cholesterol quintiles in this study was also similar to that observed by Tornberg and coworkers [34], who studied the relation between ß-lipoproteins and colorectal cancer. A significant association between the cholesterol level and colorectal cancer was not observed by Yaari and coworkers [14], but was confirmed for a small number of patients in 1987 [35]. Thomas and colleagues [32] also found a positive association in a study sample comprising a small number of medical students. In a case–control study, Jain and coworkers [36] found a positive association between alimentary fat and cholesterol intake and colorectal cancer.

The relation of the serum HDL cholesterol level to cancer, including colorectal cancer, has been previously studied. Yaari and colleagues [14] were unable to find any clear-cut association between the HDL cholesterol level and cancer in their first study. Wilson and colleagues [37] also failed to find any association between the serum HDL cholesterol level and cancer. However, in 1987 Holtzman and coworkers [35] found a positive association between the HDL cholesterol level and colon cancer in a study involving a small number of patients. In our study, the odds ratios expressing the relative risk for colorectal adenomas associated with the HDL level were significant for patients overall and for women alone. The logistic regression analysis indicated that HDL cholesterol is the lipoprotein most strongly associated with the frequency of colorectal adenoma.

The relation of the serum LDL cholesterol level to the frequency of colorectal adenomas has not been previously investigated. Evaluating colorectal carcinomas, Tornberg and colleagues [34] found that the association between the ß-lipoprotein level, which provides a rough estimate of the LDL cholesterol level, and the development of colorectal cancer was as strong as that of the total serum cholesterol level to colorectal cancer. No further studies have reported any association between LDL cholesterol and the development of colon cancer.

The association between the serum VLDL cholesterol level and colorectal cancer has not been studied. It has been reported that the HDL cholesterol level is inversely related to the VLDL cholesterol level [38, 39]. Our study confirmed this finding and, as expected, the VLDL level correlated positively with the frequency of colorectal adenomas in patients overall, as well as in each sex separately (see Table 3). Parallelism between LDL and VLDL cholesterol levels has been described in studies of lipoprotein metabolism [40, 41]. Further, a role for triglyceride-rich VLDL cholesterol has been described for coronary heart disease [41, 42], suggesting that such an association may also exist for colorectal cancer.

Shared epidemiologic factors have been used to explain the similar incidence of coronary heart disease and colorectal cancer [22, 23]. Our data support these epidemiologic data, which show a parallelism between animal fat and cholesterol consumption and the incidence of colorectal cancer and coronary heart disease [26, 43-46]. In our study, we found that colorectal adenomas, like coronary heart disease, seem to be associated with an elevated total serum cholesterol level, an elevated serum LDL cholesterol level, and a decreased serum HDL cholesterol level. Further, the logistic regression analysis of all patients with adenomas and of patients with high-risk adenomas showed that the risk profile for lipoprotein levels is similar to that seen in coronary heart disease [14]. However, it is not possible to decide, on the basis of our data, whether the evidence of a possible role for serum lipoproteins in the pathogenesis of colorectal cancer is as strong as for coronary heart disease [14].

The mechanism underlying increased serum lipoprotein levels may be the relation between lipoproteins and the secretion of bile acids. The quantity of fat ingested appears to be the major determinant of the amount of bile acids secreted into the intestine. In the colon, bile acids are transformed to secondary bile acids, which exert a cocarcinogenic effect [47-49]. The proportion of fat in the diet has an influence on the composition of gut bacteria responsible for its degradation and thus for the amount of secondary bile acids present [50]. One of the first notable steps in the neoplastic transformation of colonic mucosa is the upward shift of the proliferative compartment in the mucosal glands. Secondary bile acids were found to damage colonic mucosa with subsequent enhanced proliferation [51]. Enhanced colonic mucosal proliferation precedes the development of colorectal neoplasia and is usually found in "normal" mucosa of patients with colorectal adenoma or colorectal carcinoma [51]. Thus, secondary bile acids provide the important link to the epidemiologic finding of increased lipid levels in colon cancer.

We found a negative association between the HDL cholesterol level and the frequency of colorectal adenomas and a positive association between serum LDL and VLDL cholesterol levels and the frequency of colorectal adenomas. A logistic regression analysis showed that the HDL cholesterol level was the strongest indicator of the risk for colorectal adenomas, followed by LDL cholesterol. We conclude that patients with colorectal adenomas have lower levels of HDL cholesterol and higher levels of LDL and VLDL cholesterol, and that measurements of these lipoproteins may have prognostic significance for the development and the prevalence of colorectal adenomas.