Biomarkers of Oxidative Stress and Risk of Developing Colorectal Cancer: A Cohort-nested Case-Control Study in the European Prospective Investigation Into Cancer and Nutrition

Abstract

Oxidative stress has been shown to play an important role in carcinogenesis, but prospective evidence for an association between biomarkers of oxidative stress and colorectal cancer (CRC) risk is limited. The authors investigated the association between prediagnostic serum levels of oxidative stress indicators (i.e., reactive oxygen metabolites (ROM) and ferric reducing ability of plasma (FRAP)) and CRC risk. This was examined in a nested case-control study (1,064 CRC cases, 1,064 matched controls) in the European Prospective Investigation Into Cancer and Nutrition cohort (1992–2003). Incidence rate ratios and 95% confidence intervals were calculated using conditional logistic regression analyses. ROM were associated with overall CRC risk (highest tertile vs. lowest: adjusted incidence rate ratio (IRR_adj) = 1.91, 95% confidence interval (CI): 1.47, 2.48), proximal (IRR_adj = 1.89, 95% CI: 1.06, 3.36) and distal (IRR_adj = 2.31, 95% CI: 1.37, 3.89) colon cancer, and rectal cancer (IRR_adj = 1.69, 95% CI: 1.05, 2.72). When results were stratified by tertile of follow-up time, the association remained significant only in participants with less than 2.63 years of follow-up (IRR_adj = 2.28, 95% CI: 1.78, 2.94; P-heterogeneity < 0.01). FRAP was not associated with CRC risk. In conclusion, prediagnostic serum ROM levels were associated with increased risk of CRC. However, this association was seen only in subjects with relatively short follow-up, suggesting that the association results from production of reactive oxygen species by preclinical tumors.

biological markers, case-control studies, colorectal neoplasms, oxidative stress

Worldwide, colorectal cancer (CRC) is the third most common cancer, accounting for approximately 1.2 million new cases and 608,000 deaths per year (1). Several endogenous factors, as well as lifestyle and dietary factors, related to CRC risk have been identified (2). A diet rich in fruit and vegetables and containing antioxidants in abundance has been shown to be associated with a lower risk of cancer, including CRC (3, 4). Moreover, oxidative stress has been associated with the development of cancer (5).

Oxidative stress is an imbalance between production of reactive oxygen species (ROS) and the ability to detoxify these intermediates and to repair damage caused by ROS. It is able to cause DNA damage and has been suggested to be associated with the development of cancer (5, 6). This balance is mainly determined by endogenous enzymatic mechanisms but is also affected by exogenous factors such as lifestyle, medications, and diet (7). Tobacco smoke is considered a strong pro-oxidant, as it contains high concentrations of ROS. Additionally, certain foods and iron are documented pro-oxidants, whereas vitamins C and E are among the antioxidant food constituents (6).

Whether oxidative stress has an actual causal role in carcinogenesis or whether it is an epiphenomenon in the pathophysiology of cancer is still being debated (6). For example, oxidative DNA lesions have not been identified in tumor suppressor genes and oncogenes, and in a study by Poulsen (8), high tissue levels of oxidative lesions in animals whose DNA was “knocked out” for specific DNA repair enzymes did not result in cancer development.

Only a few epidemiologic studies have been conducted to examine the role of oxidative stress in carcinogenesis. Results from previously published case-control studies have shown increased blood levels of oxidative stress markers in patients with familial adenomatous polyposis (9) and CRC (10). In 2004, Suzuki et al. (11) prospectively investigated oxidized low density lipoprotein in relation to the risk of developing CRC. They reported a positive association between prediagnostic serum levels of oxidized low density lipoprotein and CRC risk. However, only 161 CRC cases were included in that study.

We conducted a large, prospective epidemiologic study investigating prediagnostic serum levels of oxidative stress markers in relation to subsequent development of CRC. We conducted a case-control study nested within the European Prospective Investigation Into Cancer and Nutrition cohort. Reactive oxygen metabolites (ROM) and ferric reducing ability of plasma (FRAP) were used as indicators of oxidative stress.

MATERIALS AND METHODS

Study population

The European Prospective Investigation Into Cancer and Nutrition was designed to prospectively investigate the relation between diet and various lifestyle factors and the risk of cancer (12). The study was initiated in 1991 and included 521,448 participants, approximately 70% women and mostly aged 35–70 years. Participants were recruited at 23 regional or national centers in 10 European countries (Denmark, France, Germany, Greece, Italy, the Netherlands, Norway, Spain, Sweden, and the United Kingdom). In the majority of study centers, individuals were selected from the general population of a specific town or province. Selection procedures by center were previously reported by Riboli et al. (12). The study was approved by the local ethics committees at the participating centers and by the Internal Review Board of the International Agency for Research on Cancer (Lyon, France). All participants gave written informed consent.

Data collection

Food consumption was assessed using country-specific food frequency questionnaires that had been tested against biomarkers in validation studies before the start of the main study (13). Procedures used for collecting dietary and nondietary information and protocols for blood collection and storage have been reported previously (12). In short, biologic samples including serum were collected at baseline from 385,747 participants and stored for later analysis. The samples from all countries were processed and aliquoted at the local study centers and stored in heat-sealed straws at −196°C under liquid nitrogen at International Agency for Research on Cancer biorepositories, except the samples from Denmark and Sweden, where tubes were stored at −150°C under nitrogen vapor or at −80°C in freezers, respectively.

Follow-up

Incident cancer cases were identified through record linkage with regional cancer registries at most study centers. In Germany, France, Greece, and Naples (Italy), follow-up was based on a combination of methods, including health insurance records, cancer and pathology registries, and active contact with study subjects and their next of kin. The end of follow-up for the current study was defined as the latest date of complete follow-up for both cancer incidence and vital status; it ranged from December 1999 to June 2003 for centers using registry data and from June 2000 to December 2002 for centers using active follow-up procedures.

Nested case-control design

In this nested case-control study, right or proximal colon tumors included tumors of the cecum, appendix, ascending colon, hepatic flexure, transverse colon, and splenic flexure (International Classification of Diseases, Tenth Revision (ICD-10), codes C18.0–C18.5). Left or distal colon tumors included the descending colon (ICD-10 code C18.6) and the sigmoid colon (ICD-10 code C18.7). Overlapping or unspecified-origin tumors (ICD-10 codes C18.8 and C18.9) were grouped with all colon cancers. Rectal cancer was defined as tumors occurring at the rectosigmoid junction (ICD-10 code C19) or rectum (ICD-10 code C20). Anal canal cancers were excluded. CRC was defined as the combination of colon and rectal cancer.

For each case, 1 control was selected by incidence density sampling from eligible cohort members who were alive and free of cancer at the time of the case’s diagnosis and was matched (1:1) by age (±2 years at recruitment), gender, study center, time of day of blood collection (±4 hours), and fasting status at the time of blood collection (nonfasting: <3 hours; in-between: 3–6 hours; fasting: >6 hours). Women were further matched by menopausal status, phase of menstrual cycle at the time of blood collection, and use of oral contraceptives or hormone replacement therapy at the time of blood collection, because of other studies using the same matched case-control sets.

After exclusion of cases with in situ tumors or tumors of nonmalignant morphology (n = 20) or secondary tumors (n = 2), participants with missing data on both exposures of interest (ROM and FRAP; n = 8), and participants with missing data on covariates (n = 166), 1,064 complete pairs of first incident CRC cases (671 colon cancer, 393 rectal cancer) and matched controls were eligible for analysis. Cases were not selected from Norway (blood samples had only recently been collected; few CRC cases were diagnosed after blood donation) or the center in Malmö, Sweden (limited amount of sample blood per subject).

Laboratory analysis

Serum was used for analysis of oxidative stress biomarkers. As a marker of reactive oxygen, the ROM assay was used, which is a spectrophotometric test that determines the concentration of hydroperoxides. A Diacron kit (Grosseto, Italy) was used (d-ROMs test, MC006), and the method has previously been described by Trotti et al. (14). The FRAP antioxidant assay was used as a measure of antioxidant capacity. This assay is dependent on a color change due to the reduction of a ferric complex (Fe³⁺) to a ferrous (Fe²⁺) complex by a reductant at low pH. The relative activities of samples were assessed by comparing their activities with that of Trolox (Sigma-Aldrich, Zwijndrecht, the Netherlands), a water-soluble analog of vitamin E. The assay was performed according to the method of Benzie and Strain (15). ROM and FRAP are relatively easy to measure in large quantities. For technical reasons, approximately 66% of all case-control sets were not measured in the same analytical batch. However, batch-to-batch differences are considered to have been minor; the coefficient of variation (interassay), as determined with 2 kit control samples, was only minimal (5.3% and 4.7% at ROM levels of 174 Carratelli units and 487 Carratelli units and 4.0% and 3.8% at FRAP levels of 859 μmol/L and 1,569 μmol/L, respectively), and no significant between-day drift, time shifts, or other trends were observed. All analyses were performed at the Laboratory for Health Protection Research of the National Institute for Public Health and the Environment (Bilthoven, the Netherlands), where technicians were blinded to the case-control status of the samples.

Statistical analysis

Baseline characteristics were compared for colon and rectal cancer cases and controls separately and are presented as frequencies or median values, whenever appropriate. In order to examine possible associations between ROM and FRAP and various lifestyle and dietary factors that may confound the associations between these biomarkers and CRC, we separately evaluated correlations between ROM and FRAP and these factors in controls using Spearman correlation coefficients, with adjustment for study center, age at blood collection, gender, and body mass index. The factors that were assessed included duration of smoking, height, weight, waist circumference, and the following dietary variables: energy derived from fat, energy not derived from fat, and intakes of fruit, vegetables, fish, red and processed meat, alcohol, fiber, polyunsaturated fatty acids, vitamin C, vitamin E, and beta-carotene. For the identification of biomarkers that may be involved in mechanisms potentially underlying the associations between ROM and FRAP and cancer sites, we assessed correlations between ROM and FRAP and serum levels of iron, C-reactive protein (16), high density lipoprotein cholesterol, apolipoprotein A-I (17), and glycosylated hemoglobin (18) in the same way.

Incidence rate ratios (IRRs) and 95% confidence intervals for CRC and cancer subsites were estimated using conditional logistic regression models. Data were analyzed in tertiles as well as continuously for each increment of 1 standard deviation. Tertile cutoff points and standard deviations were based on control distributions and were used for overall analyses and analyses stratified by anatomic subsite. The trend across tertiles was tested using tertile medians of the site-specific control distribution as a continuous variable in regression models.

In addition to the analysis conditioned on matching factors only, a multivariate model was used. This model included month of blood collection (12 categories), smoking habits (never smoker/former smoker who had quit ≥20 years previously/former smoker who had quit 10–19 years previously/former smoker who had quit <10 years previously/current smoker who smoked <15 cigarettes per day/current smoker who smoked 15–24 cigarettes per day/current smoker who smoked ≥25 cigarettes per day), smoking duration (years, continuous), highest educational level (none/primary school/technical or professional school/secondary school/additional education, including university degree/not specified), physical activity (inactive/moderately inactive/moderately active/active) (19), height (cm, continuous), weight (kg, continuous), waist circumference (cm, continuous), consumption of red and processed meats (g/day, continuous), fish (g/day, continuous), and fruits and vegetables (g/day, continuous), and alcohol intake (g/day, continuous).

In order to examine whether preclinical cancer may have affected the results, the associations between ROM and colorectal cancer, colon cancer, and rectal cancer were examined continuously by tertile of follow-up time (cutoff points: <2.63 years and <4.81 years). Because the results were strongly suggestive of reverse causation, all subsequent analyses were performed after excluding participants in the first tertile of follow-up time. Because of a lower number of cases and controls in the additional analyses, there was insufficient statistical power to present results for proximal and distal colon cancer separately. In addition, to evaluate whether there was a specific cutoff point for this potential reverse causation, we excluded the initial years of follow-up one by one in order to investigate when the effect of ROM on CRC risk disappeared.

For ROM, effect modification by several factors was assessed using 2 different approaches. First, since cases and controls were matched on these variables, interactions with age (in tertiles), gender, fasting status, and region were assessed by means of stratified analysis in which ROM was modeled continuously. Region was based on country (north: Sweden and Denmark; middle: the Netherlands, Germany, France, and the United Kingdom; south: Italy, Spain, and Greece). Possible heterogeneity of effects between categories of these variables was tested using the heterogeneity statistic derived from the inverse variance method (20). In addition, in order to further evaluate the effects of nonmatching variables on the outcome and on each other, we examined interactions (on the additive scale) between ROM and smoking status, physical activity, body mass index, height, weight, waist circumference, alcohol intake, and serum iron level in a joint-effects model. To do this, we constructed combined categories of tertiles of ROM and categories of these separate variables. Except for smoking status and physical activity, which had predefined categories, the other variables were divided into tertiles. Then these combined categories were included in the regression models and were compared with a joint reference category. As a reference category, a combination of the lowest category of the individual variables and the lowest ROM tertile was used.

All statistical analyses were performed using SAS software (version 9.1; SAS Institute Inc., Cary, North Carolina). For all analyses, 2-sided P values less than 0.05 were considered statistically significant.

RESULTS

Of 1,064 CRC cases, 671 colon cancer cases and 393 rectal cancer cases were identified (Table 1). The proportion of males was 45.6% among colon cancer case-control sets and 53.2% among rectal cancer case-control sets. Median follow-up time was 3.7 years for colon cancer and 3.9 years for rectal cancer.

Table 1.

Characteristics of 671 Colon Cancer Cases and 393 Rectal Cancer Cases and Their Matched Controls, European Prospective Investigation Into Cancer and Nutrition, 1992–2003

	Colon						Rectum
	Cases			Matched Controls			Cases			Matched Controls
	No.	%	Median (IQRa)	No.	%	Median (IQR)	No.	%	Median (IQR)	No.	%	Median (IQR)
Gender
Men	306	45.6		306	45.6		209	53.2		209	53.2
Women	365	54.4		365	54.4		184	46.8		184	46.8
Age, years
At recruitment			59.2 (54.1–63.3)			59.2 (54.3–63.4)			58.0 (53.0–62.5)			58.1 (53.0–62.5)
At blood collection			59.3 (54.4–63.5)			59.3 (54.4–63.5)			58.0 (53.4–62.5)			58.2 (53.3–62.5)
At end of follow-up			62.8 (57.9–67.2)						62.2 (57.3–66.2)
Follow-up time, years			3.7 (2.1–5.3)						3.9 (2.1–5.6)
Fasting status at blood collection
Not fasting (<3 hours)	315	46.9		319	47.5		216	55.0		217	55.2
In-between (3–6 hours)	167	24.9		166	24.7		89	22.7		87	22.1
Fasting (>6 hours)	176	26.2		175	26.1		81	20.6		82	20.9
Missing data	13	1.9		11	1.6		7	1.8		7	1.8
Highest educational level
None	38	5.7		34	5.1		20	5.1		19	4.8
Primary school	239	35.6		263	39.2		127	32.3		140	35.6
Technical/professional school	152	22.7		158	23.6		104	26.5		105	26.7
Secondary school	114	17.0		87	13.0		55	14.0		46	11.7
Postsecondary education (including university degree)	112	16.7		114	17.0		79	20.1		78	19.9
Not specified	16	2.4		15	2.2		8	2.0		5	1.3
Smoking habits
Never smoker	292	43.5		314	46.8		154	39.2		161	41.0
Ex-smoker
Quit ≥20 years previously	86	12.8		91	13.6		54	13.7		55	14.0
Quit 10–19 years previously	70	10.4		72	10.7		40	10.2		33	8.4
Quit <10 years previously	63	9.4		51	7.6		33	8.4		33	8.4
Current smoker
<15 cigarettes/day	55	8.2		61	9.1		47	12.0		40	10.2
15–24 cigarettes/day	61	9.1		47	7.0		31	7.9		42	10.7
≥25 cigarettes/day	13	1.9		15	2.2		11	2.8		12	3.1
Missing datab	31	4.6		20	3.0		23	5.9		17	4.3
Duration of smoking, years			10.0 (0.0–32.0)			6.0 (0.0–30.5)			17.0 (0.0–34.0)			12.0 (0.0–35.0)
Physical activity
Inactive	184	27.4		174	25.9		90	22.9		79	20.1
Moderately inactive	221	32.9		198	29.5		113	28.8		114	29.0
Moderately active	142	21.2		138	20.6		94	23.9		91	23.2
Active	124	18.5		161	24.0		96	24.4		109	27.7
Anthropometric factors
Height, cm			167.0 (160.5–174.9)			166.1 (159.0–174.0)			168.0 (161.0–174.0)			168.0 (160.5–175.0)
Weight, kg			74.0 (64.8–84.4)			72.2 (63.6–81.5)			75.0 (65.1–85.2)			74.0 (65.0–83.4)
Body mass indexc			26.5 (23.8–29.2)			26.0 (23.6–28.5)			26.4 (23.9–29.0)			26.0 (23.9–28.3)
Waist circumference, cm			90.0 (80.5–100.0)			88.0 (79.0–97.0)			91.0 (82.0–99.6)			89.5 (80.5–98.0)
Serum biomarker levels
Iron, μmol/L			17.7 (13.9–22.1)			18.9 (15.3–23.0)			19.0 (15.0–22.6)			19.0 (15.3–23.6)
C-reactive protein, mg/L			3.1 (1.2–5.5)			2.3 (1.1–5.0)			2.5 (1.0–4.4)			2.3 (1.0–4.1)
High density lipoprotein cholesterol, mmol/L			1.37 (1.11–1.66)			1.44 (1.19–1.77)			1.43 (1.17–1.73)			1.42 (1.16–1.73)
Apolipoprotein A-I, g/L			1.69 (1.51–1.92)			1.75 (1.54–1.96)			1.75 (1.51–1.97)			1.74 (1.53–1.97)
Glycosylated hemoglobin, %			5.8 (5.5–6.1)			5.7 (5.5–6.0)			5.7 (5.5–6.0)			5.7 (5.5–6.0)
Dietary intake
Fruit, g/day			196.4 (101.4–325.3)			203.1 (114.9–336.4)			182.9 (102.4–310.3)			187.6 (104.9–299.0)
Vegetables, g/day			155.7 (102.4–234.8)			159.1 (110.0–241.1)			159.4 (105.4–233.4)			161.5 (102.9–238.1)
Fish and shellfish, g/day			26.5 (14.5–46.6)			28.5 (13.6–48.7)			28.6 (16.1–51.0)			31.9 (16.2–52.4)
Red meat, g/day			47.2 (25.6–75.0)			48.3 (26.0–75.8)			57.0 (33.9–84.7)			53.2 (31.3–80.6)
Processed meats, g/day			25.6 (13.0–42.6)			23.8 (12.5–43.5)			27.6 (14.6–47.8)			26.7 (13.2–48.1)
Alcohol, g/day			8.7 (1.1–23.8)			8.0 (1.4–21.6)			11.6 (2.5–31.3)			10.5 (2.3–24.9)
Energy from fat, kcal/day			714 (548–886)			707 (554–881)			721 (557–955)			723 (570–909)
Energy not from fat, kcal/day			1,356 (1,101–1,638)			1,318 (1,096–1,604)			1,414 (1,127–1,694)			1,381 (1,106–1,667)
Polyunsaturated fatty acids, g/day			12.4 (9.3–16.8)			12.7 (9.8–16.9)			12.9 (9.5–18.5)			13.0 (10.0–17.1)
Fiber, g/day			22.0 (17.2–27.5)			22.5 (18.3–27.0)			21.8 (17.6–27.1)			22.5 (17.7–27.3)
Vitamin C, mg/day			107.8 (76.2–151.3)			111.9 (78.2–157.9)			104.9 (74.6–147.7)			109.8 (81.4–149.4)
Vitamin E, mg/day			9.1 (7.0–12.7)			9.6 (7.5–12.8)			9.2 (6.8–13.2)			9.1 (7.2–12.1)
Beta-carotene, μg/day			2,436 (1,602–3,574)			2,645 (1,727–3,759)			2,663 (1,788–3,853)			2,508 (1,602–3,744)

Abbreviation: IQR, interquartile range.

a

25th–75th percentile range.

b

Missing data on smoking habits included missing information on smoking status, time since quitting, or number of cigarettes smoked per day.

c

Weight (kg)/height (m)².

The correlations between ROM and FRAP and various lifestyle and dietary factors were weak. The strongest correlations were found between ROM and smoking duration (r = 0.20) and FRAP and waist circumference (r = 0.17). All other correlations ranged from −0.11 (ROM vs. height) to 0.11 (FRAP vs. alcohol intake). For the correlation between ROM and C-reactive protein, a correlation coefficient of 0.39 was found. Correlations with other biomarkers were between −0.12 (FRAP vs. high density lipoprotein cholesterol) and 0.15 (ROM vs. glycosylated hemoglobin). ROM and FRAP were weakly intercorrelated, as demonstrated by a coefficient of 0.08.

As is shown in Table 2, the highest tertile of ROM (compared with the lowest tertile) was significantly positively associated with CRC (IRR = 1.91, 95% confidence (CI): 1.47, 2.48), colon cancer (IRR = 2.15, 95% CI: 1.53, 3.01), proximal colon cancer (IRR = 1.89, 95% CI: 1.06, 3.36), distal colon cancer (IRR = 2.31, 95% CI: 1.37, 3.89), and rectal cancer (IRR = 1.69, 95% CI: 1.05, 2.72). Furthermore, a significant linear trend was observed for all sites (all P’s for trend < 0.05). A clear graded positive dose-response was observed for all sites except the rectum. No associations were found between FRAP and cancer at any of the sites (Table 3).

Table 2.

Incidence Rate Ratios for Colorectal Cancer and Its Subsites, Continuously and by Tertile of Reactive Oxygen Metabolites, European Prospective Investigation Into Cancer and Nutrition, 1992–2003

	No. of Cases	No. of Controls	Serum ROM Level		Crudea		Multivariate-Adjustedb
			Mean (SD)	Median	IRR	95% CI	IRR	95% CI
Colorectum
Tertile of serum ROM levelc
1 (90–348 U/mL)	246	336	306 (36)	317	1.00		1.00
2 (349–409 U/mL)	362	360	381 (17)	382	1.43	1.14, 1.79	1.40	1.10, 1.77
3 (410–735 U/mL)	445	357	464 (46)	452	1.92	1.51, 2.45	1.91	1.47, 2.48
Ptrend							<0.01
Continuous (per 1-SD increased)	1,053	1,053			1.34	1.21, 1.48	1.33	1.19, 1.49
Colon
Tertile of serum ROM level
1	142	206	306 (37)	316	1.00		1.00
2	219	234	381 (17)	383	1.43	1.07, 1.90	1.34	0.98, 1.84
3	304	225	466 (49)	453	2.28	1.67, 3.11	2.15	1.53, 3.01
Ptrend							<0.01
Continuous (per 1-SD increase)	665	665			1.38	1.22, 1.57	1.35	1.18, 1.54
Proximal Colon
Tertile of serum ROM level
1	51	73	306 (39)	317	1.00		1.00
2	89	103	381 (17)	383	1.23	0.79, 1.92	0.95	0.56, 1.62
3	136	100	473 (52)	460	2.25	1.38, 3.66	1.89	1.06, 3.36
Ptrend							0.02
Continuous (per 1-SD increase)	276	276			1.46	1.20, 1.78	1.40	1.12, 1.76
Distal Colon
Tertile of serum ROM level
1	70	103	306 (38)	316	1.00		1.00
2	112	111	381 (18)	382	1.58	1.04, 2.39	1.90	1.15, 3.13
3	143	111	461 (46)	449	2.12	1.37, 3.27	2.31	1.37, 3.89
Ptrend							<0.01
Continuous (per 1-SD increase)	325	325			1.32	1.10, 1.58	1.33	1.07, 1.64
Rectum
Tertile of serum ROM level
1	104	130	307 (34)	318	1.00		1.00
2	143	126	380 (18)	380	1.46	1.01, 2.09	1.69	1.10, 2.61
3	141	132	458 (39)	450	1.42	0.96, 2.11	1.69	1.05, 2.72
Ptrend							0.04
Continuous (per 1-SD increase)	388	388			1.25	1.05, 1.49	1.35	1.09, 1.67

Abbreviations: CI, confidence interval; IRR, incidence rate ratio; ROM, reactive oxygen metabolites; SD, standard deviation.

a

Conditioned on matching factors only (age, gender, study center, time of day, and fasting status at blood collection; for women only, also menopausal status, phase of menstrual cycle, and use of oral contraceptives or hormone replacement therapy at the time of blood collection).

b

Conditioned on matching factors and adjusted for smoking status/dose/duration, physical activity, educational level, month of blood collection, weight, height, waist circumference, and intakes of red meat, processed meats, alcohol, fruit, vegetables, and fish.

c

Tertile cutoff points were set to be identical for all cancer sites.

d

A 1-unit increase in SD was the same for all cancer sites: 74.4 U/mL.

Table 3.

Incidence Rate Ratios for Colorectal Cancer and Its Subsites, Continuously and by Tertile of Ferric Reducing Ability of Plasma, European Prospective Investigation Into Cancer and Nutrition, 1992–2003

	No. of Cases	No. of Controls	Serum ROM Level		Crudea		Multivariate-Adjustedb
			Mean (SD)	Median	IRR	95% CI	IRR	95% CI
Colorectum
Tertile of serum FRAP levelc
1 (444–936 μmol/L)	360	356	800 (96)	819	1.00		1.00
2 (937–1,142 μmol/L)	308	357	1,038 (59)	1,038	0.87	0.70, 1.09	0.82	0.64, 1.04
3 (1,143–2,732 μmol/L)	393	348	1,361 (205)	1,304	1.16	0.91, 1.48	0.97	0.74, 1.27
Ptrend							1.00
Continuous (per 1-SD increased)	1,061	1,061			1.08	0.98, 1.19	0.99	0.89, 1.11
Colon
Tertile of serum FRAP level
1	235	238	795 (99)	812	1.00		1.00
2	204	225	1,038 (59)	1,037	0.94	0.71, 1.24	0.88	0.65, 1.20
3	229	205	1,363 (205)	1,304	1.18	0.87, 1.61	0.98	0.69, 1.39
Ptrend							0.94
Continuous (per 1-SD increase)	668	668			1.10	0.98, 1.25	1.02	0.88, 1.17
Proximal Colon
Tertile of serum FRAP level
1	96	98	797 (99)	819	1.00		1.00
2	83	102	1,040 (62)	1,038	0.90	0.58, 1.39	0.78	0.46, 1.32
3	99	78	1,359 (187)	1,308	1.38	0.85, 2.26	1.17	0.63, 2.17
Ptrend							0.47
Continuous (per 1-SD increase)	278	278			1.18	0.96, 1.44	1.07	0.83, 1.39
Distal Colon
Tertile of serum FRAP level
1	117	119	796 (98)	811	1.00		1.00
2	102	103	1,037 (56)	1,037	1.01	0.69, 1.50	0.93	0.58, 1.48
3	106	103	1,370 (215)	1,307	1.07	0.69, 1.66	0.73	0.42, 1.27
Ptrend							0.28
Continuous (per 1-SD increase)	325	325			1.06	0.89, 1.26	0.93	0.75, 1.16
Rectum
Tertile of serum FRAP level
1	125	118	809 (91)	826	1.00		1.00
2	104	132	1,037 (59)	1,039	0.75	0.51, 1.11	0.66	0.43, 1.03
3	164	143	1,359 (206)	1,304	1.11	0.74, 1.66	0.98	0.61, 1.57
Ptrend							0.66
Continuous (per 1-SD increase)	393	393			1.04	0.88, 1.22	0.95	0.78, 1.15

Abbreviations: CI, confidence interval; FRAP, ferric reducing ability of plasma; IRR, incidence rate ratio; SD, standard deviation.

a

Conditioned on matching factors only (age, gender, study center, time of day, and fasting status at blood collection; for women only, also menopausal status, phase of menstrual cycle, and use of oral contraceptives or hormone replacement therapy at the time of blood collection).

b

Conditioned on matching factors and adjusted for smoking status/dose/duration, physical activity, educational level, month of blood collection, weight, height, waist circumference, and intakes of red meat, processed meats, alcohol, fruit, vegetables, and fish.

c

Tertile cutoff points were set to be identical for all cancer sites.

d

A 1-unit increase in SD was the same for all cancer sites: 267.5 μmol/L.

The associations between ROM and CRC, colon cancer, and rectal cancer were further examined in analyses stratified by tertile of follow-up time (Table 4). For all sites, IRRs in the lowest tertile of follow-up time were significantly higher than those in the middle and highest tertiles (for the lowest, middle, and highest tertiles, IRR = 2.28 (95% CI: 1.78, 2.94), IRR = 1.14 (95% CI: 0.95, 1.38), and IRR = 1.11 (95% CI: 0.89, 1.38), respectively; P-heterogeneity < 0.01). Additional adjustment for serum iron level, C-reactive protein, high density lipoprotein cholesterol, apolipoprotein A-I, and glycosylated hemoglobin did not markedly change the risk estimates (data not shown). When we excluded the initial years of follow-up one by one, the IRR for CRC was no longer statistically significant after 3 years (excluding no years: IRR = 1.33, 95% CI: 1.19, 1.49; excluding the first year: IRR = 1.24, 95% CI: 1.10, 1.40; excluding the first 2 years: IRR = 1.15, 95% CI: 1.01, 1.30; excluding the first 3 years: IRR = 1.12, 95% CI: 0.97, 1.29).

Table 4.

Multivariate-adjusted Continuous Incidence Rate Ratiosa for Colorectal, Colon, and Rectal Cancer According to a 1-SD Increaseb in Reactive Oxygen Metabolites, by Tertile of Follow-up Time, European Prospective Investigation Into Cancer and Nutrition, 1992–2003

Tertile of Follow-up Timec	No. of Casesd	No. of Controlsd	Serum ROM Level				Colorectum (1,053 Cases, 1,053 Controls)		Colon (665 Cases, 665 Controls)		Rectum (388 Cases, 388 Controls)
			Cases		Controls
			Mean	Range	Mean	Range	IRR	95% CI	IRR	95% CI	IRR	95% CI
1	348	348	412	133–639	374	90–633	2.28	1.78, 2.94	2.35	1.69, 3.27	3.07	1.50, 6.03
2	368	368	390	191–581	384	91–735	1.14	0.95, 1.38	1.14	0.89, 1.47	0.93	0.58, 1.50
3	337	337	386	134–643	380	177–639	1.11	0.89, 1.38	1.08	0.82, 1.43	1.60	0.93, 2.74
P for heterogeneitye							<0.01		<0.01		0.02

Abbreviations: CI, confidence interval; IRR, incidence rate ratio; ROM, reactive oxygen metabolites.

a

Conditioned on matching factors and adjusted for smoking status/dose/duration, physical activity, educational level, month of blood collection, weight, height, waist circumference, intake of red meat, processed meats, alcohol, fruit, vegetables and fish.

b

A 1-unit increase in SD was the same for all cancer sites: 74.4 U/mL.

c

Cutoff points for follow-up time were the same for all cancer sites: <2.63 years and <4.81 years.

d

In the total data set.

e

P for heterogeneity across tertiles, calculated using the heterogeneity statistic derived from the inverse variance method.

After excluding participants in the lowest tertile of follow-up time, subsequent continuous analyses (within 442 colon case-control sets and 263 rectum case-control sets) were stratified by tertile of age and by gender, fasting status, and region to examine effect modification by these factors. Regarding age, among participants in the highest age tertile, every 1-standard-deviation increase in ROM level was significantly associated with higher risks of CRC (IRR = 1.37, 95% CI: 1.07, 1.75) and colon cancer (IRR = 1.53, 95% CI: 1.10, 2.14). This was not observed for rectal cancer (IRR = 1.86, 95% CI: 0.79, 4.39), but this could have been caused by low numbers in this group. However, a linear trend across age tertiles was not observed, and only for colon cancer were the risk estimates between tertiles of age statistically significantly different (P-heterogeneity < 0.01). IRRs for all cancer sites were similar in men and women, in different categories of fasting status, and across regions.

After excluding participants in the lowest tertile of follow-up time, interactions between ROM and smoking status, physical activity, body mass index, height, weight, waist circumference, alcohol intake, and serum iron level were examined in a joint-effects model. Among never smokers, a graded increase in risk of CRC by tertile of ROM was seen in comparison with the lowest tertile (tertile 1, joint reference category; tertile 2: adjusted IRR (IRR_adj) = 1.53, 95% CI: 0.98, 2.38; tertile 3: IRR_adj = 1.63, 95% CI: 1.04, 2.57). Among former and current smokers, no pattern in CRC risk by tertile of ROM was observed (former smokers—tertile 1: IRR_adj = 0.90, 95% CI: 0.36, 2.28; tertile 2: IRR_adj = 1.17, 95% CI: 0.48, 2.89; tertile 3: IRR_adj = 0.84, 95% CI: 0.34, 2.03; current smokers—tertile 1: IRR_adj = 1.57, 95% CI: 0.54, 4.55; tertile 2: IRR_adj = 0.86, 95% CI: 0.28, 2.68; tertile 3: IRR_adj = 0.80, 95% CI: 0.26, 2.43). IRRs for physical activity, height, and weight did not show a clear pattern.

DISCUSSION

To our knowledge, this is the largest prospective cohort-nested case-control study to have investigated the association between CRC and indicators of oxidative stress. In this European population of CRC cases and matched controls, a positive association was observed between prediagnostic levels of ROM and the risk of CRC and its subsites. However, this association was apparent only in participants with a relatively short follow-up time. From these patients, blood was collected less than 2.63 years before CRC diagnosis, suggesting that preclinical cancer may have affected the results. For prediagnostic levels of FRAP, no association was found with cancer risk at any of the examined sites.

In a previously published case-control study, Gackowski et al. (10) also reported increased levels of an oxidative stress marker among colon cancer cases. However, because of the retrospective nature of that study, no separate analyses of follow-up time before diagnosis could be conducted as was done in our study. In a prospective study, Suzuki et al. (11) found an association between serum oxidized low density lipoprotein levels and CRC risk. However, only 161 CRC cases were included in that study, and no analyses stratified for time before diagnosis could be conducted.

Among participants in the first tertile of follow-up, it is very likely that tumor development and growth has already started at the time of blood collection. It has been shown that human tumor cells are able to produce large quantities of ROS (21), and it is highly probable that the association between ROM and risk of developing CRC in these subjects is a result of ROS production by preclinical tumors, rather than a causal factor in carcinogenesis (6). Nonetheless, our finding could not conclusively differentiate between increased levels of ROM as a late cause of CRC and increased levels of ROM as an early consequence of CRC.

After exclusion of all cases and controls identified in the first tertile of follow-up, only in participants over 62 years of age was an increased ROM level associated with increased CRC risk. We do not have a clear explanation for this finding, and our finding could be due to chance. It may be that the preclinical phase is longer in older patients; however, based on our data, we cannot rule out a causal effect of ROM.

Our finding that the effect of ROM is seen only in never smokers, not in former and current smokers, is counterintuitive. Perhaps, for reasons unknown to us, former and current smokers more efficiently eliminate ROS than never smokers. However, this finding may also be explained by chance.

In the complex concept of oxidative stress, caused by an imbalance between ROS production and detoxifying or cell-repairing ability (5), it is plausible that measurements of reactive oxygen and antioxidant status made separately do not provide sufficient information about this balance. Although they are expected to be counterparts, ROM and FRAP were not correlated in the present study (r = 0.08), and the expected inverse association between FRAP levels and CRC was not observed. It may be that this is explained by specific characteristics of the FRAP assay. This assay measures the ability of an antioxidant to reduce a ferric complex (Fe³⁺) to a ferrous complex (Fe²⁺). Since the FRAP assay is largely dependent on the concentration of certain serum components, such as albumin and uric acid, this method does not necessarily have as much sensitivity as the ROM assay to detect changes in free radical production (22).

Strengths of our study include its prospective design, with the advantage that all blood samples and information on potential confounders were collected before the occurrence of CRC. Furthermore, the relatively large sample size and the collaboration of European countries from north to south contributed a wide variety of lifestyle and dietary habits and CRC incidence rates.

Nonetheless, several limitations should be considered as well. First, for the examination of ROM and FRAP in relation to CRC risk, only a single measurement from a baseline blood sample was used. Reliability of serum ROM and FRAP levels over several years has been reported previously (23) and was poor for ROM (Spearman rank correlations: r = 0.29 for men and r = 0.39 for women) and moderate for FRAP (r = 0.78 and r =0.63). By using a single measurement of a biomarker, we are likely to have underestimated the associations between ROM and FRAP and CRC. Second, follow-up time in this study was relatively short. The observed association between ROM and CRC was most likely affected by preclinical cancer, and whether high ROM levels increase CRC risk over the long term remains unclear. Third, information on the presence of other inflammatory diseases (e.g., inflammatory bowel disease, rheumatoid arthritis) at baseline was not available in our cohort. Since these conditions are also associated with a higher production of ROS, as well as with a higher colon cancer risk in the case of inflammatory bowel disease (24), this may have caused residual confounding. However, additional adjustment for blood levels of C-reactive protein as a marker of inflammation did not markedly affect our results. Additionally, medications like statins and nonsteroidal antiinflammatory agents can affect oxidative stress (25). Unfortunately, we did not have sufficient information on medication use to address this in our study.

Considering the lack of clear evidence for a role of ROS in the etiology of CRC and the suggestion of ROM being an early consequence of CRC, further studies are needed to examine whether measuring ROM levels may be a useful clinical tool in improving the accuracy of diagnosis and advancing the date of diagnosis among suspected cases.

In conclusion, the results of this study indicate that prediagnostic serum levels of ROM are associated with an increased risk of developing CRC. The fact that this association was observed only in participants with less than 2.63 years of follow-up strongly suggests that the association between ROM and CRC risk is a result of ROS production by preclinical tumors, rather than a causal factor in carcinogenesis (reverse causation). Studies with longer follow-up, repeated measurements, and combined measures of reactive oxygen exposure and antioxidant status are needed to explore this complex, possibly long-term association. Additionally, further studies, particularly studies of the possible diagnostic use of ROM in clinical practice, are indicated.

Abbreviations

Abbreviations
 
• CI

  confidence interval

 
• CRC

  colorectal cancer

 
• FRAP

  ferric reducing ability of plasma

 
• ICD-10

  International Classification of Diseases, Tenth Revision

 
• IRR

  incidence rate ratio

 
• ROM

  reactive oxygen metabolites

 
• ROS

  reactive oxygen species

Author affiliations: Department of Gastroenterology and Hepatology, University Medical Center, Utrecht, the Netherlands (Anke M. Leufkens, Peter D. Siersema, H. Bas Bueno-de-Mesquita); National Institute for Public Health and the Environment, Bilthoven, the Netherlands (Anke M. Leufkens, Fränzel J. B. van Duijnhoven, Sjoukje H. S. Woudt, Eugene H. J. M. Jansen, H. Bas Bueno-de-Mesquita); Julius Center for Health Sciences and Primary Care, University Medical Center, Utrecht, the Netherlands (Fränzel J. B. van Duijnhoven, Petra H. Peeters); International Agency for Research on Cancer, Lyon, France (Mazda Jenab); Department of Epidemiology, German Institute of Human Nutrition, Potsdam-Rehbruecke, Nuthetal, Germany (Tobias Pischon, Heiner Boeing, Krasimira Aleksandrova); Institute of Cancer Epidemiology, Danish Cancer Society, Copenhagen, Denmark (Anne Tjønneland, Anja Olsen); Department of Clinical Epidemiology, Aarhus University Hospital, Aarhus, Denmark (Kim Overvad); Institut National de la Santé et de la Recherche Médicale, Centre for Research in Epidemiology and Population Health, U1018, Institut Gustave-Roussy, Villejuif, France (Marie Christine Boutron-Ruault, Françoise Clavel-Chapelon, Sophie Morois); Unité Mixte de Récherche Scientifique 1018, Paris South University, Villejuif, France (Marie Christine Boutron-Ruault, Françoise Clavel-Chapelon, Sophie Morois); Molecular and Nutritional Epidemiology Unit, Cancer Research and Prevention Institute, Florence, Italy (Domenico Palli); Nutritional Epidemiology Unit, Fondazione IRCCS Istituto Nazionale dei Tumori, Milan, Italy (Valeria Pala); Cancer Registry and Histopathology Unit, “Civile M. P. Arezzo” Hospital, Ragusa, Italy (Rosario Tumino); MRC-HPA Centre for Environment and Health, School of Public Health, Imperial College London, London, United Kingdom (Paolo Vineis); HuGeF Foundation, Turin, Italy (Paolo Vineis); Department of Clinical and Experimental Medicine, Faculty of Medicine, Federico II University, Naples, Italy (Salvatore Panico); Division of Cancer Epidemiology, German Cancer Research Center, Heidelberg, Germany (Rudolf Kaaks, Annekatrin Lukanova); WHO Collaborating Center for Food and Nutrition Policies, Department of Hygiene, Epidemiology and Medical Statistics, University of Athens Medical School, Goudi, Athens, Greece (Antonia Trichopoulou); Hellenic Health Foundation, Athens, Greece (Antonia Trichopoulou, Vardis Dilis); Department of Epidemiology, Harvard School of Public Health, Boston, Massachusetts (Dimitrios Trichopoulos); Bureau of Epidemiologic Research, Academy of Athens, Athens, Greece (Dimitrios Trichopoulos); Department of Community Medicine, Faculty of Health Sciences, University of Tromsø, Tromsø, Norway (Guri Skeie); Unit of Nutrition, Environment and Cancer, Cancer Epidemiology Research Programme, Catalan Institute of Oncology, Barcelona, Spain (Carlos A. González); Public Health and Participation Directorate, Health and Health Care Services Council, Asturias, Spain (Marcial Argüelles); Andalusian School of Public Health, Granada, Spain (María-José Sánchez); Consortium for Biomedical Research in Epidemiology and Public Health, Madrid, Spain (María-José Sánchez, José María Huerta, Eva Ardanaz); Epidemiology and Health Information Unit, Public Health Division of Gipuzkoa, Basque Regional Health Department, San Sebastian, Spain (Miren Dorronsoro); Department of Epidemiology, Murcia Regional Health Authority, Murcia, Spain (José María Huerta); Navarre Public Health Institute, Pamplona, Spain (Eva Ardanaz); Nutrition Research Unit, Department of Public Health and Clinical Medicine, Umeå University, Umeå, Sweden (Göran Hallman); Pathology Unit, Department of Medical Biosciences, Umeå University, Umeå, Sweden (Richard Palmqvist); Department of Public Health and Primary Care, School of Clinical Medicine, University of Cambridge, Cambridge, United Kingdom (Kay-Tee Khaw); MRC Epidemiology Unit, Institute of Metabolic Science, University of Cambridge, Cambridge, United Kingdom (Nick Wareham); Cancer Epidemiology Unit, Nuffield Department of Clinical Medicine, University of Oxford, Oxford, United Kingdom (Naomi E. Allen, Francesca L. Crowe); Nutritional Epidemiology Group, Nutrition and Metabolism Section, International Agency for Research on Cancer, Lyon, France (Veronika Fedirko); and Department of Epidemiology and Biostatistics, School of Public Health, Imperial College London, London, United Kingdom (Teresa Norat, Elio Riboli).

This work was supported by the Directorate-General for Health and Consumers (formerly known as the Directorate-General for Health and Consumer Protection) of the European Commission during 1993–2004; the Directorate-General for Research of the European Commission during 2005–2008; the Ligue contre le Cancer, Institut Gustave-Roussy, the Mutuelle Générale de l’Education Nationale, and the Institut National de la Santé et de la Recherche Médicale (France); German Cancer Aid, the German Cancer Research Center, and the Federal Ministry of Education and Research (Germany); the Danish Cancer Society (Denmark); the Health Research Fund of the Spanish Ministry of Health and participating regional governments and institutions (Spain); the Carlos III Institute, Cancer Research UK, and the Medical Research Council (United Kingdom); the Stavros Niarchos Foundation, the Hellenic Health Foundation, and the Hellenic Ministry of Health and Social Solidarity (Greece); the Italian Association for Research on Cancer and the National Research Council (Italy); the Dutch Ministry of Public Health, Welfare and Sports, the Netherlands Cancer Registry, LK Research Funds, Dutch Prevention Funds, Dutch ZON (Zorg Onderzoek Nederland), the World Cancer Research Fund, and Statistics Netherlands (the Netherlands); the Swedish Cancer Society, the Swedish Scientific Council, and the Regional Government of Västerbotten (Sweden); and Nordforsk—Centre of Excellence Programme HELGA (Norway).

Conflict of interest: none declared.