C-Reactive Protein and the Risk of Incident Colorectal Cancer FREE

Inflammation has been hypothesized to increase the risk of cancer.¹ Although there is growing evidence from laboratory studies supporting the role of inflammation in the pathogenesis of colorectal cancer, data from epidemiologic studies are sparse. However, inflammation could be particularly important in the pathogenesis of colorectal cancers. Chronic inflammatory bowel diseases, such as ulcerative colitis, have been associated with increased risk of colon cancer.² Moreover, several studies^{3 - 8} have shown a reduced risk of colon cancer with use of aspirin or other anti-inflammatory agents.

C-reactive protein (CRP) is an acute-phase protein produced primarily in the liver in response to stimulation by interleukin 6 (IL-6). In recent years, elevated levels of CRP have been shown to reliably predict cardiovascular events in several populations.⁹ Both CRP and IL-6 have been shown to be associated with total and noncardiovascular mortality as well.^{10 - 11} These findings are consistent with those from studies^{12 - 14} that demonstrate an increased risk of mortality from cancer using other markers of inflammation and raise the possibility that inflammation could play a role in the development of cancer. Therefore, elevated circulating markers of inflammation, in particular CRP, could help identify persons at risk for developing colorectal cancer. To examine this hypothesis, we determined the risk of colon and rectal cancer associated with elevated CRP levels in a prospective, nested case-control study.

Colorectal cancer cases and controls were identified among members of the CLUE II cohort, a prospective study established in May 1989 and named for its campaign slogan, "Give Us A Clue to Cancer and Heart Disease." The cohort consisted of 32 897 individuals, among whom 22 887 were Washington County, Maryland, residents older than 18 years who formed the analytic cohort for this study. Participants provided a blood sample and completed a brief questionnaire at baseline, after providing written informed consent. Loss to follow-up was less than 5% among cohort members who were 45 years or older at baseline, the age range of most cases at baseline and the pool of cohort members eligible to be selected as age-matched controls. Enrollment took place from May to October 1989. Individuals are still under active follow-up. For the purpose of case-control selection from this cohort, the final date of case diagnosis was December 2000 (maximum of 11 years of follow-up). To ensure equal follow-up time between cases and controls, the matched control was selected to have a date of blood draw (enrollment date) within 2 weeks of the case, and the matched control had to still be alive and free of a cancer diagnosis at the date of the case's colorectal cancer diagnosis. The institutional review board at the Johns Hopkins Bloomberg School of Public Health approved the study.

Cases were identified through linkage with the Washington County Cancer Registry and, since 1992, with the Maryland State Cancer Registry. A total of 172 men and women who did not have a prior cancer diagnosis (except possibly for nonmelanoma skin cancer or cervix in situ) were diagnosed as having colon or rectal cancer following the date of blood draw through December 2000. Of these, 131 individuals had cancer of the colon (International Classification of Diseases, Ninth Revision [ICD-9] code 153) and 41 had cancer of the rectum (ICD-9 code 154). Ninety-eight percent of the cases were confirmed by review of pathology report. Information on stage of diagnosis was available for 130 (99 colon, 31 rectal) of the 172 cases.

For each case, up to 2 controls were selected from among CLUE II participants who did not have a diagnosis of cancer through December 2000 and who were not known to be deceased at the time the case was diagnosed (n = 342). Controls were matched to each case on age (plus or minus 1 year), sex, race, date of blood draw (plus or minus 2 weeks), and time since last meal (0-1, 2-3, 4-5, 6-7, and ≥8 hours).

The Maryland Cancer Registry is certified by the North American Association of Central Cancer Registries as being more than 95% complete (http://www.naaccr.org/index.asp?Col_SectionKey=12&Col_ContentID=55). Compared with the Maryland Cancer Registry, the Washington County Cancer Registry captured 98% of the colorectal cancer cases diagnosed in Washington County residents in 1998.

Smoking history and self-reported weight and height were recorded for each participant at baseline. Medication use was assessed at baseline. Hormone use in women was defined as self-report of oral contraceptives or any other estrogen or progesterone use, including hormone therapy, within the last 48 hours. Use of aspirin or nonsteroidal anti-inflammatory drugs (NSAIDs) within the last 48 hours was determined by self-report. Follow-up questionnaires were mailed in 1996 and 2000, which ascertained whether participants had a history of inflammatory bowel disease before 1989 or had a family history (parent or sibling) of colon or rectal cancer. Diabetes was defined as taking any diabetic medication at baseline or having a hemoglobin A_1C of 6.1% or higher.

C-reactive protein concentrations were measured in duplicate among cases and controls from plasma stored at −70°C since baseline examination using a high-sensitivity assay (Dade Behring, Newark, Del).¹⁵ Twenty-seven quality control samples (equal to 5% of total sample) aliquotted from pooled plasma were arranged in triplets among the cases and controls. Each case and its controls (the triplets) were run adjacently. The laboratory was blind to case-control and quality control sample status. The mean intrapair coefficient of variation among quality control samples was 3.3%. Hemoglobin A_1C levels were determined by turbidimetric immunoinhibition in red blood cells (Hitachi 911 analyzer, Boehringer Mannheim, Indianapolis, Ind).

Matched sets that contained the case and at least 1 control with a CRP measurement were used in the main analysis. Of 173 original cases and matched controls (n = 346), 1 case and 2 controls had less than the detectable limit of CRP (0.02 mg/L) and were excluded from the analysis, as were the corresponding controls for the excluded case. Analyses that imputed the values to one half the detection limit did not appreciably change our findings; therefore, these cases and controls were excluded. Because CRP values are right-skewed, the Wilcoxon signed rank test was used to determine differences in the distribution of baseline CRP between cases and controls. Results were replicated using log-transformed CRP and paired t tests. Matched odds ratios (ORs) for colorectal cancer, calculated as an estimate of the relative risk, and corresponding 95% confidence intervals (CIs) were estimated using conditional logistic regression. Quartile cut points for CRP were based on the distribution of concentration among controls. Tests for trend were conducted by entering a single ordinal variable using the median of each category. Additional analyses were conducted according to cancer site (ie, colon vs rectal). Smoking status was not associated with outcomes in this sample, and log CRP levels were similar among never and former smokers after adjusting for body mass index (BMI, calculated as weight in kilograms divided by the square of height in meters; P = .15). However, CRP levels remained significantly higher among smokers compared with former and never smokers in this and other studies¹⁶ ; therefore, we report results separately for nonsmokers. In subsidiary analyses using all controls, the continuous association with colon cancer associated with a 1-SD increase in log CRP (1.02 mg/L) was determined using multivariate logistic regression adjusting for potential confounders (age, sex, smoking status, BMI, and use of hormones, aspirin, and NSAIDs), after excluding cases that occurred within 2 and 5 years of blood draw and stage III or IV cancers at the time of diagnosis (to reduce the effect of subclinical disease), and stratified by sex, overweight status, and use of aspirin or other NSAIDs. Overweight was defined as having a BMI of 25 or higher. Associations remained unchanged after computing new CRP quartile cut points based on nonsmokers only. All analyses were performed using STATA version 7.0 statistical software (STATA Corp, College Station, Tex). Reported P values are for 2-sided statistical tests, and P<.05 was considered statistically significant.

Table 1 shows baseline characteristics of cases and controls. Because of matching, cases and controls were of similar age, race, and sex. Equal proportions of cases and controls were current smokers at baseline. Small differences in BMI (P = .15), hormone use in women (P = .17), and use of aspirin (P = .17) and NSAIDs (P = .13) were not statistically significant.

Among controls, BMI was significantly greater for persons in the highest quartile of CRP compared with persons in the lowest quartile (mean [SD], 27.8 [4.8] vs 24.3 [3.2]; P for trend <.001; Table 2). No statistically significant differences were observed for other covariates.

Table 3 shows baseline geometric mean and median concentrations of CRP in cases and controls. Baseline geometric mean and median concentrations of CRP were higher among persons who subsequently developed colorectal cancer than among those who remained free of disease (P = .01). This association was present for colon cancer but not rectal cancer. Among persons who subsequently developed colon cancer, median CRP concentrations were 2.69 vs 1.97 mg/L for matched controls (P<.001). In contrast, CRP concentrations were not significantly different between cases of rectal cancer and controls (1.79 vs 1.81 mg/L, P = .32). Among nonsmokers, median CRP levels were higher among colorectal cancer cases compared with matched controls (2.44 vs 1.85 mg/L, respectively; P = .002) and among colon cancer cases compared with matched controls (2.67 vs 1.84 mg/L, respectively; P<.001). Only 21 cases and 41 controls were smokers at baseline. No association was found between CRP levels and risk of cancer among smokers.

The odds of developing colorectal cancer increased with higher concentrations of CRP, such that persons in the highest quartile of CRP had a 2-fold increased risk of colorectal cancer compared with persons in the lowest quartile (OR, 2.00; 95% CI, 1.16-3.46; P for trend = .008) (Table 4). This increase in risk was primarily observed among those who developed colon cancer (highest vs lowest quartile, OR, 2.55; 95% CI, 1.34-4.88; P for trend = .002).

The risk of colorectal cancer associated with higher concentrations of CRP was greater among nonsmokers (highest vs lowest quartile, OR, 2.52; 95% CI, 1.36-4.70; P for trend = .004). Nonsmokers in the highest quartile of CRP had more than a 3-fold increased risk of developing colon cancer compared with those in the lowest quartile (OR, 3.51; 95% CI, 1.64-7.51; P for trend<.001). No association was found between CRP and risk of rectal cancer in nonsmokers (P for trend = .63, data not shown).

Additional analyses were conducted to determine the association between CRP and subsequent risk of colon cancer among selected subgroups and to determine the potential influence of undiagnosed colorectal cancer at baseline. In general, associations were similar across subgroups (Table 5). To reduce associations due to occult disease at baseline, we conducted analyses limited to early-stage disease and after excluding cases that occurred within 2 and 5 years of follow-up. Among cases in which the initial stage of tumor was stage I or II (n = 51), the adjusted OR of developing colon cancer associated with a 1-SD increase in log CRP was 1.38 (95% CI, 0.99-1.91). Among nonsmokers, the corresponding OR was 1.51 (95% CI, 1.07-2.14). After excluding cases of colon cancer that occurred within 2 years of follow-up, the adjusted OR of colon cancer associated with a 1-SD increase in log CRP was 1.35 (95% CI, 1.05-1.74) among all participants and 1.57 (95% CI, 1.20-2.07) among nonsmokers. C-reactive protein remained significantly associated with a higher risk of colon cancer in nonsmokers after excluding cases that occurred within 5 years of follow-up (OR, 1.52; 95% CI, 1.08-2.13). In addition, there was no correlation among colon cancer cases between time to diagnosis and CRP level (Spearman r = 0.04), indicating that cases that occurred early did not have appreciably higher levels of CRP at baseline than cases that occurred later.

Consistent with the hypothesis that inflammation increases the risk of colorectal cancer, use of either aspirin or NSAIDs within the last 48 hours was associated with a reduced risk of colorectal cancer (OR, 0.63; 95% CI, 0.41-0.97). Nonsignificant decreases in risk were present for both colon cancer (OR, 0.67; 95% CI, 0.41-1.09) and rectal cancer (OR, 0.53; 95% CI, 0.22-1.31). Of the 337 respondents to the 1996 and 2000 follow-up questionnaires in the current study, 5 (2 cases, 3 controls) had a history of inflammatory bowel disease before 1989, and 23 (10 cases, 13 controls) had a positive family history of colon or rectal cancer in a parent or sibling. All results were unchanged when we repeated analyses after excluding persons with a baseline history of inflammatory bowel disease or a positive family history of colorectal cancer.

Finally, to determine whether the association between CRP and colon cancer risk was independent of diabetes, we limited the analysis to persons without diabetes. After adjusting for all covariates, a 1-SD increase in log CRP was still associated with a significantly increased risk of colon cancer (OR, 1.35; 95% CI, 1.03-1.75). Among nonsmokers, the association was stronger (OR, 1.47; 95% CI, 1.09-1.96). These results suggest that the association between colon cancer and inflammation is independent of diabetes.

These prospective findings demonstrate that prediagnostic concentrations of CRP are strongly associated with the subsequent development of colon cancer. This association was slightly stronger in nonsmokers and remained significant after excluding higher-stage tumors at diagnosis and cases that occurred within 5 years of follow-up, thus reducing the likelihood that our results could reflect the presence of subclinical disease at the time of blood collection. Results were similar in men and women and remained significant when limiting analyses to persons without diabetes.

These findings are consistent with several lines of evidence that suggest that the risk of colon cancer in particular is increased with inflammation. Individuals with chronic inflammatory bowel disease, particularly ulcerative colitis, are at substantially higher risk of developing colorectal cancer than nonaffected persons and the risk increases with duration of disease.¹⁷ In laboratory studies, inflammation has been shown to promote the conversion of colonic adenoma cells to adenocarcinoma cells.¹⁸ Moreover, IL-6, a pleiotropic cytokine that is a potent stimulator of CRP, has been shown to stimulate the growth of primary and metastatic colon cancer cell lines.¹⁹ Observational studies^{12 - 14} have shown that modest elevations in inflammatory markers, including total white blood cell count and fibrinogen, are associated with an increased risk of cancer mortality in persons free of cancer at baseline. Finally, clinical trials have demonstrated that anti-inflammatory agents can reduce the risk of developing colon adenomatous polyps.^{3 - 5 ,8} The lack of an association with rectal cancer could reflect insufficient power to detect associations within this small subset of participants who developed rectal cancer or an alternative biological pathway. Additional studies are needed to clarify this issue.

Despite evidence that inflammation could play a role in the pathogenesis of colon cancer among persons without a history of inflammatory bowel disease, data from epidemiologic studies are sparse. We know of only 1 other prospective study²⁰ that has examined the risk of incident colon cancer associated with higher concentrations of CRP. Although the association was not statistically significant, that study was aimed at risk of any cancer, had few colon cancer cases (n = 44), and had limited follow-up time (58 months). A study conducted in CLUE I (started in 1974 and also conducted among residents of Washington County, Maryland) found that lower concentrations of serum albumin were associated with an increased risk of incident colon cancer.²¹ Because serum albumin is reduced in the presence of inflammation, an inverse association between serum albumin levels and risk of colon cancer supports our hypothesis that inflammation increases the risk of colon cancer.

Several factors should be considered in the interpretation of our findings. A major strength of the current study is that it is a prospective study, and thus, we can more confidently infer a temporal association between inflammation and the occurrence of colon cancer. In addition, most cases (98%) were histologically confirmed, thus reducing the likelihood of misclassification. One potential limitation is that CRP was measured at one point in time. However, unaccounted intraindividual variation would tend to attenuate the association and thus cannot account for our results. Finally, CRP is a nonspecific marker of inflammation, and additional studies of specific cytokines or factors that regulate acute-phase response are necessary to elucidate the mechanisms by which inflammation increases the risk of colon cancer.

Although confirmation of these results is clearly warranted, this finding, if true, could have implications for prevention strategies. Several studies have demonstrated that CRP levels can be reduced with smoking cessation²² and weight loss.^{23 - 24} In addition, several trials have demonstrated a reduced risk of polyps and colon cancer from use of aspirin and NSAIDs.^{3 - 8} Consistent with these findings, in the present study we found a reduced risk of colorectal cancer with use of these agents, but the results of tests for interaction between use of anti-inflammatory agents and CRP were not statistically significant (data not shown). However, data on use of anti-inflammatory agents were limited to use within the last 48 hours. Additional studies are needed to clarify whether use of anti-inflammatory agents modifies the association between CRP and colon cancer risk. Finally, the cut points for CRP associated with moderate (1-3 mg/L) and high (>3 mg/L) risk of cardiovascular disease are in the same range associated with increased colon cancer risk in the present study.²⁵

In summary, this study demonstrates that elevated concentrations of CRP are strongly associated with the development of colon cancer in individuals believed to be free of this disease at baseline. This finding is consistent with literature that supports the role of chronic inflammation in the pathogenesis of colon cancer. Additional studies are needed to confirm these findings and to determine the implications on screening and prevention of colon cancer.