Associations among IRS1, IRS2, IGF1, and IGFBP3 Genetic Polymorphisms and Colorectal Cancer

Introduction

Uncontrolled cell growth is central to the carcinogenic process. Factors that regulate and control cell growth are undoubtedly important to the etiology of cancer. While the biological rationale for the importance of looking at factors influencing cell growth is clear, determination of which factors may be most informative is not. There is a growing body of evidence that suggests that insulin-like growth factors (IGF), IGF binding proteins (IGFBPs), especially IGFBP-3, and insulin play a significant role in the initiation of cell growth and proliferation of colorectal cancers (1-5). Insulin has been regarded as primarily a metabolic signal, while IGF-I has been implicated as an important mitogen and cell differentiation factor (6, 7). The insulin receptor substrate (IRS) protein family contains several members, of which IRS-1 and IRS-2 are expressed in almost all cells and tissues (8-10). While IRS-1 controls body growth and peripheral insulin action, IRS-2 regulates body weight control and glucose homeostasis (9). Within tumors, IRS-1 may be a marker of an active IGF signal transduction pathway (4, 11), although IRS is involved in insulin signaling.

Many factors are involved in the regulation of insulin and IGFs, including diet, lifestyle, hormonal, and genetic factors. Studies have shown that diet, physical activity, body size, and sex steroids are involved in the regulation of these hormones (12-18); some data suggest that serum levels of IGF, IGFBP-3, and IRS also may be affected by polymorphisms in these genes (19-22). Now, it is somewhat uncertain which variants in these genes are most relevant in the regulation of serum levels of IGF-I, IGFBP-3, IRS-1, and IRS-2 and in the etiology of colorectal cancer. Few variants of these genes have been evaluated with cancer, most notably a CA repeat of IGF1 (23-26), with inconsistent results. We could find no studies assessing associations with any variants of these genes and colorectal cancer. However, based on the literature, there are indications that certain genetic polymorphisms may be important.

The G972R polymorphism in IRS1 has been associated with insulin resistance and type 2 diabetes, leading to the hypothesis that the IRS1 R allele would increase risk of colon cancer (27). The IRS2 G1057D polymorphism has been associated with insulin resistance, although the association has been shown to differ based on body size (28). High IGF-I and low IGFBP-3 serum levels have been associated with colon cancer risk in some but not all studies. Variation in serum IGF-I levels has been associated with a polymorphism in the IGF1 gene 1 kb upstream of the transcription start site (29). This is a CA repeat polymorphism, with the most common allele containing 19 CA repeats, denoted “192” for the size of the PCR product (29). Because IGF-I serum levels have been shown to be lower among men with the 192/192 genotype, one could hypothesize that the 192/192 genotype would decrease colorectal cancer risk. In IGFBP3, an A > C polymorphism at nucleotide −202 has been associated with different levels of IGFBP-3 in a dose response fashion (i.e., AA > AC > CC; ref. 30). Therefore, one could predict that the AC or CC genotype could be associated with an increased risk of colorectal cancer, as some studies indicate that high levels of IGFBP-3 reduce colon cancer risk and those with the AC and CC genotypes should have the lowest IGFBP-3 levels.

In this article, we evaluate the associations of genetic polymorphisms in the IGF1, IGFBP3, IRS1, and IRS2 genes with colorectal cancer both independently and in conjunction with each other. We hypothesize that the 192/192 IGF1 genotype reduces risk of colorectal cancer and that the AA genotype of the IGFBP3 gene reduces risk of colorectal cancer; the joint effects of these genotypes may be more important than each genotype independently, given that IGFBP-3 regulates bioavailability of IGF-I. We hypothesize that having a R allele of the IRS1 gene will increase risk of colon cancer and that the DD genotype of the IRS2 gene will decrease risk of colon cancer. We also evaluate if these genes relate to colorectal cancer risk differently among those with and without a family history of colorectal cancer. We look at family history of cancer, because there is limited information on the association of relatively common, low penetrance genes and colon cancer risk among those with and without a family history. We hypothesize that the importance of these genes may vary by family history of colorectal cancer.

Methods

Study Population

Participants in the study were from the Kaiser Permanente Medical Care Program of (northern California) and Utah. All eligible cases within these defined geographic areas were identified and recruited for the study. Two study populations are included in these analyses. The first study includes cases and controls from a population-based case-control study of first primary colon cancer (International Classification of Diseases for Oncology, Second Edition codes 18.0 and 18.2 to 18.9) diagnosed between October 1, 1991 and September 30, 1994. The second study includes cases with a first primary tumor in the rectosigmoid junction or rectum, identified between May 1997 and May 2001 in Utah and in Kaiser Permanente Medical Care Program. Case eligibility was determined by the Surveillance, Epidemiology, and End Results cancer registries in northern California and in Utah. In both studies, cases were identified using rapid reporting systems, and eligibility included being between ages 30 and 79 years at the time of diagnosis, English speaking, mentally competent to complete the interview, no previous history of colorectal cancer (31), and no known (as indicated on the pathology report) familial adenomatous polyposis, ulcerative colitis, or Crohn's disease.

Controls were matched to cases by sex and by 5-year age groups. At the Kaiser Permanente Medical Care Program, controls were randomly selected from membership lists; and in Utah, controls ages ≥65 years were randomly selected from Health Care Financing Administration lists and controls ages <65 years were randomly selected from driver's license lists. The analysis includes 1,346 colon cancer cases and 1,544 controls interviewed between February 1991 and May 1994 and 952 rectal cancer cases and 1,205 controls interviewed between October 1997 and January 2002. For the colon study, 80.8% of cases and 71.6% of controls whom we were able to contact were interviewed; for the rectal study, we interviewed 73.2% of cases and 68.8% of controls contacted. Response rates, or the number interviewed divided by overall persons identified, were 71.8% for colon cancer cases and 68.0% for controls selected for the colon cancer study and 65.2% for cases and 65.3% for controls for the rectal cancer study. The primary reasons for nonparticipation of cases were death prior to interview or too ill to be interviewed (∼11%) and refusal for both cases (20%) and controls (29%).

Questionnaire Data. Family history of colorectal cancer was determined from an interviewer-administered questionnaire. Participants were asked to list all first-degree family members, including parents, siblings, and children. After enumerating family members, participants were asked to report family members that had ever been diagnosed with cancer and specific type of cancer they were diagnosed with. In this analysis, family history of colorectal cancer included cancer of the colon, rectum, or large bowel. Other questionnaire data included age, diet history, physical activity history, medical history, and reproductive history.

Genotyping

DNA was extracted from peripheral blood leukocytes. Of the colon cancer cases and controls who were interviewed, 1,181 cases and 1,194 controls had germ line DNA available for analysis; many samples were no longer available given previous analyses. Of the 952 rectal cancer cases and 1,205 controls interviewed, 827 cases and 1,031 controls had DNA extracted. Of these, genotyping data were available for 792 rectal cancer cases and 985 controls for IGF1, 794 rectal cancer cases and 989 controls for IGFBP3, 796 rectal cancer cases and 988 controls for IRS1, and 766 rectal cancer cases and 983 controls for IRS2. For quality control, known controls representing all polymorphisms and blanks were included in each 96-well tray. All genotypes were scored by two individuals to help insure accuracy.

IRS1. The G972R polymorphism was detected using PCR amplification with primers 5′-CTTCTGTCAGGTGTCCATCC and 5′-TGGCGAGGTGTCCACGTAGC (32). PCR cycling consisted of an initial denaturation at 94°C for 2 minutes, 10 cycles at 94°C for 10 seconds, 60°C for 10 seconds, and 72°C for 10 seconds followed by 30 cycles at 94°C for 10 seconds, 55°C for 10 seconds, and 72°C for 10 seconds. BstNI was used to digest the PCR products following manufacturer's instructions. Alleles were scored as either G for glycine or R for arginine (absence or presence of the restriction site, respectively).

IRS2. The G1057D polymorphism was detected using a previously described TaqMan assay (33) with minor modifications. Primer and probe sequences for the TaqMan assay were as follows: primer IRS-2 F 5′-GGAGCTGTACCGCCTGCC-3′, primer IRS-2 R 5′-ACCAAAAGCCATCTCGGTGT-3′, G-probe FAM-CCGGGCGCCGCCTCAT-Tamra, and A-probe FAM-CCGGGCGCCGCCTCAT-Tamra, and A-probe VIC-CGGACGCCGCCTCATCGTT-Tamra. Each 17 μL PCR reaction contained 20 ng genomic DNA, 900 nmol/L of each primer, 130 nmol/L of each TaqMan probe, and 8.5 μL TaqMan 2× Universal PCR Master Mix (contains AmpErase UNG and AmpliTaq Gold enzymes, deoxynucleotide triphosphates, and reaction buffer). PCR was carried out under the following conditions: 95°C for 10 minutes followed by 40 cycles of 95°C for 15 seconds and 62°C for 1 minute using IQ detection system (BioRad, Hercules, CA). The fluorescence of each sample was colected and analyzed by iCycler IQ real-time detection software (version 3.0).

IGF1. The IGF1 CA repeat was amplified using PCR primers forward 5′-GCTAGCCAGCTGGTGTTATT-3′ and R 5′-ACCACTCTGGGAGAAGGGTA-3′ (29, 34). PCR conditions consisted of a 2-minute denaturation at 94°C followed by 30 cycles of 94°C for 10 seconds, 57°C for 10 seconds, and 72°C for 15 seconds. The IGF1 products were electrophoresed on 6% denaturing polyacrylamide gels at 70 W for 3 hours. The gels were dried and exposed to X-ray film. Alleles were assigned by size (bp) of fragment and classified as “192” or “non-192.” “192” is the PCR product size of the most common allele, which contains 19 CA repeats.

IGFBP3. The −202 A > C polymorphism was amplified using primers F 5′-CCACGAGGTACACACGAATG-3′ and R3 5′-TGAGCAGCCGGGGCCGAG (35). Amplitaq Gold (0.5 units) and 5% DMSO were used to increase efficiency of amplification. PCR conditions were 9-minute initial denaturation at 95°C followed by 40 cycles at 95°C for 10 seconds and 66°C for 20 seconds. The resulting PCR product was digested with Alw21I (4 units) at 37°C overnight. Digested products were separated on a 2% NuSieve gel stained with ethidium bromide and visualized with UV light. Alleles were scored as either A or C (presence or absence of the restriction site, respectively).

Statistical Methods

IRS1 genotypes were GG, GR, and RR, with the R allele being less common. IRS2 genotypes were GG, GD, and DD, with the GG genotype being most common. IGF1 genotypes were 192/192, heterozygous, or non-192 alleles; the absence of the 192 allele was less common than the presence of the 192 allele. IGFBP3 genotypes were CC, CA, and AA. Tumor site was defined as proximal (cecum through transverse colon), distal (splenic flexure, descending, and sigmoid colon), and rectal (rectosigmoid junction and rectum).

SAS Statistical Package (version 8.2) was used to conduct the analyses. Analyses included evaluating the distribution of the genotypes in the population, the independent associations of genetic polymorphisms with colorectal cancer risk, and the joint effect of genotypes on colorectal cancer risk. Logistic regression models were used to assess risk associations. Because controls were frequency matched to cases, matched logistic regression models were not used. In all logistic models, age and sex were adjusted, unless data were stratified by sex, where only age was used as an adjustment variable. Other lifestyle factors, such as diet and physical activity, did not appreciably alter associations. No differences were detected by center, so center was not adjusted in analyses. Because the majority of the population was non-Hispanic White, adjustment for racial group did not influence results; stratification by ethnic group was not meaningful given small sample size for participants who were not White (see Table 1). Sex- and age-specific analyses were done; categorization of age was done for those ages ≥65 and <65 years. Several tests for interaction between genes were performed The relative excess risk from interaction and corresponding 95% confidence interval (95% CI) was calculated to provide insight into differences that might be expected on an additive scale of relative excess risk from interaction (36). The Wald χ² test was used to determine significant differences in slopes. Hardy-Weinberg equilibrium and allele frequency was determined using the SAS Genetics program.

Results

Rectal cancer cases were younger than colon cancer cases (Table 1). For both colon and rectal cancer, there were more men than women interviewed. In both studies, the population was predominately non-Hispanic White. Family history of colorectal cancer was reported in the same frequency for both colon controls and rectal controls, although more colon cancer cases than rectal cancer cases reported a history of colorectal cancer. Few differences in genotype distribution existed by study controls, where ∼11% had either GR or RR genotype for IRS1, 12% to 14% had the DD IRS2 genotype, 40% had the 192/192 IGF1 genotype, and 27% to 29% had the IGFBP3 CC genotype.

Based on data from the control population, IGFBP3, and IRS1, and IRS2 were in Hardy-Weinberg equilibrium, while IGF1 was not (data not shown). Results were similar from controls selected for the colon study and those selected for the rectal cancer study (P = 0.35 and 0.13 for IGFBP3 for colon and rectal, respectively; P = 0.64 and 0.26 for IRS1 for colon and rectal, respectively; P < 0.01 for colon and 0.07 for IGF1; and P = 0.44 for both colon and rectal studies for IRS2). Allele frequencies also were similar between the two samples. The A allele frequency for IGFBP3 was 0.463 (95% CI 0.448-479) for colon and 0.468 (95% CI 0.446-0.492) for rectal; the C allele frequency was 0.537 (95% CI 0.521-0.552) for colon and 0.531 (95% CI 0.508-0.554) for colon and rectal studies, respectively. The IRS1 G allele frequency was 0.939 (95% CI 0.932-0.947) and 0.939 (95% CI 0.928-0.949) for colon and rectal controls, respectively; the R allele frequency was 0.606 (95% CI 0.531-0.680) and 0.612 (95% CI 0.511-0.724) for colon and rectal controls, respectively. The IRS2 G allele frequency was 0.658 (95% CI 0.643-0.672) and 0.645 (95% CI 0.630-622), while the D allele frequency was 0.342 (95% CI 0.329-.366) and 0.355 (95% CI 0.338-370) for colon and rectal controls, respectively. There were 13 alleles detected for IGF1. The 192 allele was the most common, with a frequency of 0.642 (95% CI 0.627-0.657) and 0.634 (95% CI 0.613-0.553) for the colon and rectal cancer control participants. Other common IGF1 alleles were the 194 allele (frequency 0.184 for colon and 0.172 for rectal), the 190 allele (frequency 0.054 for colon and 0.070 for rectal), and the 196 allele (frequency 0.073 for colon and 0.080 for rectal).

Analyses were conducted separately for colon cancer and rectal cancer. For main effects, there was a significant association of the IRS1 R allele and colon cancer (Table 2). Among women, having at least one copy of the IRS1 R allele (i.e., GR or RR genotypes) was associated with a significantly increased risk of colon cancer. The associations for IRS1 were significantly stronger among individuals diagnosed prior to age 65 years [data not shown; odds ratio (OR) 1.7, 95% CI 1.2-2.5; P interaction < 0.05]. The IRS2 G972D heterozygous GD genotype was associated with a significant 20% to 30% decrease in colon cancer risk overall. There were no associations of the main effects of the IGF1 and IGFBP3 variants and colon cancer risk. There were no significant associations between any of the variants evaluated and rectal cancer risk.

In the analyses of associations of the gene-gene interactions and colon cancer risk (Table 3), there was a significant increase in risk for individuals with an IRS1 R allele and with at least one IGF1 non-192 allele (OR for 192 heterozygote 1.6, 95% CI 1.1-2.3; OR for non-192 alleles 2.3, 95% CI 1.2-4.4), although the P for interaction did not reach statistical significance (0.20). The IRS1 GR/RR genotypes also resulted in a 70% increased risk of colon among those who had either AC or AA genotype in IGFBP3. Similar associations were not observed with rectal cancer (data not shown).

Assessment of associations among those with and without a family history of colorectal cancer showed that the GR/RR IRS1 genotypes were associated significantly with an increased risk of colon cancer among those without a family history of colorectal cancer (Table 4). Having non-192 IGF1 allele was associated with an increased risk of colon cancer among those with a family history of colorectal cancer, although given the small sample size, this association was not statistically significant (OR 1.8, 95% CI 0.8-3.9). These genotypes did not appear to be importantly related to rectal cancer in people with or without a family history of colorectal cancer (data not shown).

Discussion

It has been suggested that insulin and IGF may importantly contribute to risk of colorectal cancer. Polymorphisms in genes that involved in the regulation of serum levels of IGF, insulin, and IRS may be associated with colorectal cancer. While many genes are involved in the regulation of insulin-related factors, we assessed four polymorphisms that had been shown to have functional significance in the regulation of hormone levels and may therefore influence colorectal cancer risk. Evaluation of genetic polymorphisms in four genes along this pathway suggests that having at least one R allele of the IRS1 gene (GR/RR genotypes) may increase risk of colon cancer. The GR/RR genotypes of the IRS1 gene in combination with specific genotypes of IGFBP3 (AC or AA) and IGF1 (non-192 allele) also appear to increase risk of colon cancer, while a significant inverse association was observed for those with the IRS2 heterozygous GD genotype.

IRS-1 is the major cytoplasmic substrate of the insulin receptor in most insulin-sensitive tissues, and some studies suggest that IRS-1 plays an important role in regulating insulin secretion in pancreatic β cells (37-39). Of the many polymorphisms of the IRS1 gene described, the glycine-to-arginine substitution at codon 972 (G972R) has been studied in conjunction with obesity, polycystic ovary syndrome, and non-insulin-dependent diabetes (32, 40, 41), making this a plausible variant that may alter cancer risk. The R allele has been associated with impaired insulin-associated signaling (32, 41) and lower fasting plasma concentrations of insulin and C-peptides (32). Insulin resistance has been hypothesized as being associated with colon cancer (17), so a slight increase in risk associated with the R allele may be the result of its association with obesity, poorer insulin sensitivity, diabetes, and altered insulin action and secretion (27, 32, 42, 43). We observed that the R allele slightly increased risk (OR 1.4) of colon cancer overall, with a slightly stronger association for those diagnosed prior to age 65 years (OR 1.8).

IRS2, like IRS1, is thought to be involved in insulin signaling and glucose intolerance (8, 10, 44). There is little information about the G1057D polymorphism of the IRS2 gene, although there is reason to believe that IRS2 may be an important risk factor for colon cancer, given its previously reported association with obesity and diabetes (9, 28, 45). IRS2^−/− mice have been shown to be obese (9). The G1057D IRS2 polymorphism is a reasonable candidate to evaluate in conjunction with colon cancer given its relationship to IRS2 haplotypes associated with obesity (28). Lautier et al. (28) also found that the G1057D IRS2 polymorphism was heterogeneous in its association with obesity and that having the R allele of the IRS1 gene influenced the association seen with the G1057D IRS2 gene and obesity. In our study, we observed that people eith the GD genotype were at reduced risk of colon cancer, with stronger associations among women, although neither GD nor DD genotype was associated with significant altered rectal cancer risk in either men or women.

IGF-I may be important for colorectal cancer risk because of its role in cell growth and differentiation (34, 46, 47). High IGF-I serum levels have been associated with increased risk of colorectal cancer (relative risk 2.51, 95% CI 1.15-5.46; ref. 48). In one study, high levels of IGF-I were associated with a 5-fold increase in risk of colorectal cancer (48, 49). Laboratory studies also have shown the presence of IGF-I receptors in normal intestine (suggesting a role for normal growth and function of intestinal cells) and colorectal cancer cell lines (2, 50, 51). Variation in serum IGF-I levels has been associated with a polymorphism 1 kb upstream of the transcription start site of the IGF1 gene (29). For men, serum IGF-I concentrations were lower with the 192/192 genotype than for other genotypes. One report showed that individuals who were 194/192 heterozygotes had 25% higher IGF-I serum levels than those who were homozygous for the 192 alleles. In our study, the 192/192 genotype did not affect the risk of colon cancer by itself, although the lack of the 192 allele did appear to increase colon cancer risk in combination with the IRS1 R allele. This is consistent with the data showing that the lack of the 192 allele is associated with increased IGF-I levels, which in turn is associated with an increased cancer risk.

IGFBP-3 modulates the activity of IGF-I (5, 52). In human colorectal cancer cell lines, studies have shown that binding protein levels may dictate growth response to IGF-I (2, 53, 54). High levels of IGFBP-3 have been associated with reduced risk of colorectal cancer (relative risk 0.28, 95% CI 0.12-0.66; ref. 30). IGFBP-3 also appears to play a role in regulation of cancer cell growth, independent of the IGF signaling pathway, through inducing apoptosis (52, 55). Sequence alterations have been reported in the IGFBP3 gene in a significant proportion of gastrointestinal tumors (22). The −202 C > A transition has been shown to result in different levels of IGFBP-3 in a dose response fashion (i.e., AA > AC > CC). We observed little variation in risk associated with the IGFBP3 −202 C > A polymorphism for colon cancer independently, although there was an association in conjunction with the IRS1 R allele, where those with the AC or CC genotype were at greater risk if they also had an IRS1 R allele. Although the IGFBP3 polymorphisms did not appear to affect cancer risk independently, the effect of the C allele in conjunction with the IRS1 genotype is in the direction predicted by the effect of this allele on IGFBP-3 levels.

These data suggest that genes involved in the insulin-related pathway may be more important for colon cancer than for rectal cancer, because we saw no effect of these polymorphisms on rectal cancer risk. Significant associations of the G972R variant and colon cancer were observed alone and in association with the IGFBP3 and IGF1 variants. However, for rectal cancer, we observed little variation in risk for any of the variants whether alone or in combination. Support for a weaker role of the insulin-related pathway for rectal cancer than for colon cancer also exists in the epidemiology literature, where obesity generally is not associated with rectal cancer (56, 57) but is associated with colon cancer (31); physical activity is more consistently associated with colon than rectal cancer. Dietary glycemic index has been associated with colon cancer (58, 59), although less is known about the association between rectal cancer and dietary components that may be related to insulin.

A major strength of the study is our inclusion of several variants in several genes along an insulin-related pathway that may be associated with colorectal cancer. The large size of our study allows us to study multiple variants; although given the rare frequency of some alleles, we still are limited in our ability to assess statistically significant associations. Focusing on a disease pathway rather than isolated genetic polymorphisms along that pathway allows for a more complete examination of the pathway and interpretation of results. We believe that looking at multiple genes along a pathway provides more information on the importance of the pathway to the etiology of the disease. A limitation of this study is that we examined only four polymorphisms in four genes along this pathway; one of the genes evaluated was not in Hardy-Weinberg equilibrium. This most likely suggests that many of the alleles are rare given that IGF1 is a length variant and that there is sampling variation for rare genotypes such as the non-192/192 IGF-I. Although these polymorphisms have been reported to show associations with hormone levels and/or clinic parameters, there may be other polymorphisms in these and other genes in this pathway that importantly regulate insulin, IGF, and IRS and may be involved in the carcinogenic process. It will be important in the future to investigate other functional polymorphisms of genes along an insulin-related pathway to further our understanding of insulin in the etiology of colon cancer.

A second limitation is the lack of ethnic diversity in the population study, although ethnic distribution was similar between cases and controls. These findings may be most applicable to non-Hispanic White populations, as it is known that there is ethnic variation in allele frequency. For example, a study evaluating the IGF1 CA repeat showed different genotype frequencies based on ethnic background: 33.9% of 123 African Americans had non-192/192 alleles, 32.0% of 71 Japanese Americans had non-192/192 alleles, 6.4% of 58 non-Latino Whites had non-192/192 alleles, and 14.7% of 154 Latino Americans had non-192/192 IGF1 alleles (25). In our control population, ∼14% had non-192 alleles. One study examining the IGFBP-3 −202 polymorphism in 943 mostly non-Hispanic White women observed the AA genotype in 21.6% (60); a study of 478 men in the Physician's Health Study showed that 25% had the AA IGFBP3 genotype (20). In our study, ∼21% of controls had the AA IGFBP3 genotype. Likewise, our estimate of ∼12% of controls having a R allele of the IRS1 gene is lower than was observed in 157 subjects whose body mass index was >30 kg/m² (∼20%) but higher than observed in 157 subjects with a body mass index of <28 kg/m² (8.3%; ref. 42). Information from this study adds to our knowledge of the frequency of these variants in the population, given the limited number of studies of these variants, many of which are not population based or are limited in number of individuals with available genetic data. A major strength is our ability to reconfirm allele frequencies of these genes from two large control samples that were collected at two different points in time.

In conclusion, polymorphisms in IRS1 and IRS2 have modest independent effects on colon cancer risk. However, when the IRS1, IRS2, IGF1, and IGFBP3 variants are evaluated together, a more substantial and significant effect on colon cancer risk is observed. These data provide some support for an insulin pathway in the etiology of colon but not rectal cancer. As with all such studies, confirmation of our results in other populations is necessary to establish their significance. We are currently assessing these genotypes in conjunction with diet and lifestyle factors to more fully understand the disease pathway and its importance in colorectal neoplasia, as interactions between such factors and the respective genotypes also may influence cancer risk.