p53 Arg72Pro polymorphism and risk of colorectal adenoma and cancer
A single nucleotide polymorphism (SNP) at codon 72 of the p53 gene (Arg72Pro) alters the p53 protein structure and affects its activity. We investigated this SNP in relation to colorectal adenoma and cancer among men and women from case–control studies nested within the Nurses' Health Study, the Health Professionals Follow-up Study and the Physicians' Health Study. Among 856 colorectal adenoma cases and 1,184 controls, we observed a modest association with p53 Arg72Pro genotype (multivariate odds ratio (OR) = 1.25, 95% confidence interval (CI) = 1.04–1.50 for Arg/Pro and Pro/Pro vs. Arg/Arg). This association did not vary by colorectal site or by sex. Among 442 colorectal cancer cases and 904 controls, we observed no significant overall association between p53 Arg72Pro genotype and colorectal cancer (multivariate OR = 1.14, 95% CI = 0.90–1.45). However, when colorectal site and sex was accounted for, the Pro carrier genotypes compared to Arg/Arg were associated with an increased risk of proximal colon cancers in women (multivariate OR = 2.59, 95% CI = 1.49–4.52) though not with distal colon or rectal cancers, while among men the same genotypes were associated with an increased risk of distal colon cancers (multivariate OR = 2.09, 95% CI = 1.28–3.40) but not proximal colon or rectal cancers. Our results suggest that Arg72Pro may play a role in the early stages of colorectal neoplasia and possibly in progression to invasive disease, depending on site and sex. © 2006 Wiley-Liss, Inc.
The p53 tumor suppressor gene encodes for a transcription factor that initiates numerous cellular signals, most notably cell cycle arrest and apoptosis.1, 2 Mutations in p53 occur in at least 50% of all cancers and over 90% of these eliminate the ability of the p53 protein to bind to its DNA targets,2 underscoring the importance of p53 in tumorigenesis. A single nucleotide polymorphism (SNP) in the p53 gene resulting in the substitution of Arginine (Arg) by Proline (Pro) at codon 72 was identified and shown to alter the primary structure of the p53 protein.3 Though both structural forms are normal in their sequence-specific DNA binding activities, some functional differences have been identified.4 The 2 forms of p53 differ in their susceptibility to degradation by the human papillomavirus E6 protein.5 In addition, Thomas et al. observed that the Pro form activated transcription to a higher level than the Arg form, while the kinetics of apoptosis was faster with the Arg form.4 Results from more recent studies have confirmed that the Arg form is more efficient at inducing apoptosis,6, 7 and it has been shown that the Pro form is better at inducing cell cycle arrest.7
Colorectal carcinogenesis is described by a multistep genetic model, where mutations within several genes, including p53, accumulate in the progression from a normal epithelium to adenoma to invasive cancer.8, 9 Given the experimental evidence for functional differences between the 2 polymorphic forms of p53, genotype at codon 72 may alter susceptibility to colorectal tumor development. The association between the p53 Arg72Pro polymorphism and colorectal cancer risk has been examined in 6 previous studies10, 11, 12, 13, 14, 15 with inconsistent results; 2 studies12, 15 suggested an elevation in risk associated with the Pro carrier genotypes, while in the other studies, the Pro carrier genotypes were associated with a null or statistically nonsignificant inverse association. To our knowledge, the risk of colorectal adenoma has not been examined in association with this polymorphism. In this study, we examined the p53 Arg72Pro polymorphism in relation to colorectal cancer among men and women from case–control studies nested in the Nurses' Health Study (NHS) and the Physicians' Health Study (PHS). In addition, we examined this polymorphism in association with colorectal adenoma among women in the NHS and men in the Health Professionals Follow-up Study (HPFS).
Material and methods
Each of the nested case–control studies included here has been described previously.16, 17, 18 Among women, colorectal adenomas and cancers were examined in the NHS, an ongoing prospective study of 121,700 female registered nurses in the United States. Women aged 30–55 years in 1976 provided questionnaire information on demographic, lifestyle and reproductive factors. Dietary information was first obtained in 1980. Updated information on risk factors, screening history (including endoscopy) and disease diagnoses has been collected biennially through mailed questionnaires (diet is collected quadrennially). Self-reported disease diagnoses are confirmed by medical record review. Between 1989 and 1990, blood specimens were collected from 32,826 women in the NHS. Women with colorectal adenomas or cancer and their controls were selected from among this subcohort of women with a blood sample. Women were eligible for the adenoma case–control study if they had had a sigmoidoscopy or colonoscopy after providing a blood sample and did not have a history of cancer (except nonmelanoma skin cancer) or colorectal adenoma before the date of endoscopy. Between 1989 and 1998, 557 women with colorectal adenoma were identified, and 1 control was selected for each case, matched on year of birth, date of blood draw, time period of endoscopy, indication for endoscopy and time period of first or most recent endoscopy excluding the one in the current time period. One case was subsequently found to have hyperplastic polyps instead of adenoma and thus excluded, leaving 556 cases and 557 controls. Women included in the cancer case–control study were eligible if they were free of inflammatory bowel disease and did not have a history of cancer (except nonmelanoma skin cancer). Between 1989 and 2000, 197 women with incident colorectal cancer were identified. For each case, 2 or 3 women without colorectal cancer and matched to cases on year of birth, date of blood collection and fasting status were selected as controls (n = 499).
Colorectal adenomas among men were examined in the HPFS, an ongoing prospective study of 51,529 male dentists, optometrists, osteopaths, podiatrists, pharmacists and veterinarians in the United States. In 1986, men aged 40–75 years provided questionnaire information on demographic, lifestyle, medical history and dietary factors. Updated information has been collected biennially through mailed questionnaires, except diet, which was collected quadrennially. Self-reported disease diagnoses were confirmed by medical record review. Between 1993 and 1994, blood specimens were collected from 18,018 men in the HPFS. We identified men with colorectal adenomas and their matched controls from the subcohort of men who had completed a valid dietary questionnaire in 1986, had undergone sigmoidoscopy or colonoscopy after the date of return of the 1986 dietary questionnaire, and did not have a history of cancer (except nonmelanoma skin cancer) or colorectal adenoma before the date of endoscopy. Between 1986 and 1994, 367 cases were identified and 736 men who matched to cases on year of birth, year of endoscopy and whether they had had a previous endoscopy were selected at random as controls. Subsequently, 9controls were identified as cases, leaving 376 cases and 725 controls.
Colorectal cancer in men was examined in the PHS, a randomized, double-blind, placebo-controlled 2 × 2 factorial trial of low-dose aspirin and β-carotene, now followed as an observational cohort. In 1982, 22,071 male physicians aged 40–84 years were successfully enrolled and randomized. A blood specimen was also provided at baseline in 1982 by 14,916 (68%) of the randomized men. Information on study compliance, medication use, risk factors for various diseases and the occurrence of any illness has been assessed annually through mailed questionnaires. Self-reported disease diagnoses are confirmed by medical record review. Between 1982 and 2000, 272 men were diagnosed with colorectal cancer from among the 14,916 that provided a blood specimen. For each case, 1 or 2 men who were free from diagnosed cancer at the time of case ascertainment and who had also provided a blood specimen were selected as controls (n = 456), and were matched to cases on age and smoking status at baseline (current, former, and never-smokers).
DNA extraction and genotyping
In the PHS, genomic DNA was extracted from whole blood, while in the NHS and HPFS, DNA was extracted from 50 μl of buffy coat diluted with 150 μl of phosphate-buffered saline using the QIAmp (Qiagen, Chatsworth, CA) 96-spin blood protocol according to the manufacturer's instructions. The 5′ nuclease (TaqMan) assay19 was used to genotype the p53 Arg72Pro SNP (rs1042522). Specific primer and probe sequences are available on request. Allelic discrimination to determine genotype was conducted by measuring end-point fluorescence with the ABI PRISM Sequence Detection System software. A random 10% of quality control samples were inserted, and concordance for these samples was 100%. Laboratory personnel were blinded to the case–control status of the specimens, as well as to the quality control samples. The overall genotyping success rate was 93.8%, and the genotype distributions among cases and controls in each of the nested case–control studies were in Hardy–Weinberg equilibrium.
To account for the matching between cases and controls, conditional logistic regression was used to estimate odds ratios (OR) and 95% confidence intervals (CI) for the association between p53 Arg72Pro genotype and colorectal adenoma and cancer. To increase statistical power, we also used unconditional logistic regression, adjusted for the matching factors, and compared these results to those from the conditional logistic regression models. Multivariate models for men in the HPFS and women in the NHS included family history of colorectal cancer, pack-years of smoking, body mass index, physical activity level, multivitamin use, aspirin use, alcohol consumption, dietary energy-adjusted folate intake, red meat consumption, total energy intake and postmenopausal hormone use (among women only). Among men in the PHS, multivariate models included body mass index, physical activity level, multivitamin use, aspirin use and alcohol consumption. Modification of the ORs by colorectal cancer risk factors was evaluated by including product terms for genotype and the effect modifier of interest. The p-value for multiplicative interaction was based on the likelihood ratio test comparing models with and without the product terms. All statistical tests were 2-sided.
Arg72Pro genotype was modeled as both a 3-level categorical variable with Arg/Arg as the reference and as a dichotomous variable comparing Pro carrier genotypes (Pro/Pro and Arg/Pro) to Arg/Arg. ORs and 95% CIs were estimated for men and women from the different cohort studies separately. The presence of heterogeneity between the ORs for men and women was tested for using the Q statistic, and if appropriate, the ORs were pooled using the inverse variance method.20 We evaluated the association between genotype and colorectal adenoma and cancer. In addition, we examined associations separately for each colorectal site (i.e. proximal colon, distal colon and rectum).
Overall, there were a total of 856 adenoma cases and their 1,184 controls and 442 cancer cases and their 904 controls with an informative genotype result. The mean age of adenoma cases and their matched controls was 62.3 years (standard deviation (S.D.) 6.6) for women and 60.9 years (S.D. 8.3) for men. Cancer cases and their matched controls were older with a mean age of 65.4 years (S.D. 6.6) for women and 67.5 years (S.D. 8.5) for men. Over 93% of each of the study populations were of Caucasian ancestry (Table I). The distribution of Arg/Arg, Arg/Pro and Pro/Pro genotypes among controls was similar across each study population and was in Hardy–Weinberg equilibrium (Table I). The genotype distributions corresponded to an Arg allele frequency of 74–76% across the populations, which is consistent with what has been observed previously in populations of predominantly Caucasian origin. Both adenomas and cancers were more often diagnosed in the colon compared to the rectum (Table II). The majority of colon adenomas were located in the distal colon for both women (53.5%) and men (60.1%; Table II). For cancer, women tended to have more proximal colon cancers than distal colon cancers, while the distribution of cancers by colon site in men was more equal (Table II).
Table 1. Characteristics of the Study Population
| Women||96.5||497||490||288 (58.8)2||172 (35.1)||30 (6.1)||0.53|
| Men||97.7||359||694||381 (54.9)||275 (39.6)||38 (5.5)||0.20|
| Women||95.9||187||473||266 (56.2)||178 (37.6)||29 (6.1)||0.91|
| Men||>933||255||431||232 (53.8)||173 (40.1)||26 (6.0)||0.40|
Table 2. Distribution of Colorectal Adenoma and Cancer by Site
|Missing/ Multiple3||694|| ||384|| ||19|| ||22|| |
The results from the conditional and unconditional logistic regression analyses were not appreciably different, thus, only the results from the unconditional analyses are presented. The associations between each of the risk factors included in the multivariate models and colorectal cancer and adenoma were generally similar across all 4 nested case–control studies (results not shown). For the association between the Arg72Pro SNP and colorectal adenoma risk (Table III), matching-factor adjusted and multivariate analyses were similar. The magnitudes of the ORs for Arg/Pro and for Pro/Pro separately compared to the Arg/Arg genotype were similar and indicated an increased risk of adenoma associated with the Pro carrier genotypes among women. For the dichotomous representation of genotype, the multivariate OR (95% CI) for women was 1.37 (1.05–1.78) for Pro carrier genotypes vs. Arg/Arg. Among men, the magnitude of the association was lower and not statistically significant [multivariate OR (95% CI): 1.14 (0.87–1.48) for Pro carrier genotypes vs. Arg/Arg]. There was no statistical evidence of heterogeneity between the results for men and women (p-value for heterogeneity = 0.32 in the analysis of Pro carrier genotypes vs. Arg/Arg) and the pooled multivariate OR (95% CI) was 1.25 (1.04–1.50) for Pro carrier genotypes vs. Arg/Arg. There was no appreciable difference in association for large (≥1 cm) versus small (<1 cm) adenomas (results not shown). In a case–case analysis, the multivariate OR (95% CI) for the risk of developing large versus small adenomas was 1.05 (0.70–1.58) for women and 0.89 (0.53–1.49) for men. In addition, the ORs for men and women separately were not modified by colorectal cancer risk factors (results not shown), including body mass index, aspirin use, dietary folate intake, multivitamin use, family history of colorectal cancer, pack-years of smoking, and postmenopausal hormone use among women (all p-values for interaction >0.13).
Table 3. Odds Ratios for the Association between p53 Codon 72 Genotype and Colorectal Adenoma
|Women2||497||490|| || |
| Arg/Arg||263 (52.9)3||288 (58.8)||1.00 [referent]4||1.00 [referent]|
| Arg/Pro||199 (40.0)||172 (35.1)||1.29 [0.99–1.68]||1.37 [1.04–1.80]|
| Pro/Pro||35 (7.0)||30 (6.1)||1.30 [0.77–2.20]||1.37 [0.80–2.36]|
| Arg/Pro + Pro/Pro||234 (47.1)||202 (41.2)||1.29 [1.00–1.67]||1.37 [1.05–1.78]|
|Men||359||694|| || |
| Arg/Arg||189 (52.7)||381 (54.9)||1.00 [referent]||1.00 [referent]|
| Arg/Pro||147 (41.0)||275 (39.6)||1.08 [0.83–1.41]||1.12 [0.85–1.48]|
| Pro/Pro||23 (6.4)||38 (5.5)||1.21 [0.70–2.10]||1.23 [0.70–2.16]|
| Arg/Pro + Pro/Pro||170 (47.4)||313 (45.1)||1.10 [0.85–1.42]||1.14 [0.87–1.48]|
|Pooled2,5||856||1,184|| || |
| Arg/Arg||452 (52.8)||669 (56.5)||1.00 [referent]||1.00 [referent]|
| Arg/Pro||346 (40.4)||447 (37.8)||1.18 [0.98–1.42]||1.24 [1.02–1.50]|
| Pro/Pro||58 (6.8)||68 (5.7)||1.26 [0.86–1.84]||1.30 [0.88–1.92]|
| Arg/Pro + Pro/Pro||404 (47.2)||515 (43.5)||1.19 [0.99–1.43]||1.25 [1.04–1.50]|
A similar pattern was observed for the association between p53 Arg72Pro genotype and colorectal cancer (Table IV), where a slight elevation in risk associated with the Pro carrier genotypes, though not statistically significant, was suggested for women [multivariate OR (95% CI): 1.23 (0.86–1.77) for Pro carrier genotypes vs. Arg/Arg], while the association was essentially null for men [multivariate OR (95% CI): 1.08 (0.79–1.49) for Pro carrier genotypes vs. Arg/Arg]. The results for men and women were not significantly different from each other (p-value for heterogeneity = 0.59 in the analysis of Pro carrier genotypes vs. Arg/Arg) and the pooled multivariate OR (95% CI) was 1.14 (0.90–1.45) for Pro carrier genotypes vs. Arg/Arg. For both men and women, the ORs were not modified by colorectal cancer risk factors including body mass index, aspirin use, multivitamin use and smoking behavior (all p-values for interaction >0.11). We also examined interactions with dietary folate intake and family history of colorectal cancer among women only because the data was not available for men in the PHS, and we observed no evidence of effect modification by these factors (p-values for interaction >0.42). We observed a marginally significant interaction with postmenopausal hormone use (p-value for interaction = 0.06), suggesting that the Pro carrier genotypes were more strongly associated with an increased risk of colorectal cancer among women who were currently using postmenopausal hormones. When compared with the ORs for never or past users of postmenopausal hormones with the Arg/Arg genotype, the ORs (95% CIs) were 0.99 (0.63–1.55) for never or past users of postmenopausal hormones who were Pro carriers, 0.41 (0.24–0.71) for current users of postmenopausal hormones with the Arg/Arg genotype, and 0.85 (0.48–1.51) for current users of postmenopausal hormones who were Pro carriers.
When we analyzed adenomas occurring in the proximal colon, distal colon or rectum separately, the magnitudes of the site-specific associations were not appreciably different from each other for both men and women (Table V). Analyses of cancer (Table VI) suggested that the Pro carrier genotypes were more strongly associated with proximal colon cancers [multivariate OR (95% CI): 2.59 (1.49–4.52)] compared to the distal colon or rectum among women. However among men, the Pro carrier genotypes were found to be more strongly associated with the risk of distal colon cancers [multivariate OR (95% CI): 2.09 (1.28–3.40)] compared to the proximal colon or rectum. In a case–case analysis, the multivariate OR (95% CI) for the risk of developing proximal versus distal colon cancer associated with the Pro carrier genotypes was 6.52 (2.38–17.86) for women (p-value for difference <0.001) and 0.34 (0.17–0.68) for men (p-value for difference = 0.003).
Table 4. Odds Ratios for the Association between p53 Codon 72 Genotype and Colorectal Cancer
|Women2||187||473|| || |
| Arg/Arg||94 (50.3)||266 (56.2)3||1.00 [referent]4||1.00 [referent]|
| Arg/Pro||82 (43.9)||178 (37.6)||1.29 [0.91–1.84]||1.24 [0.85–1.80]|
| Pro/Pro||11 (5.9)||29 (6.1)||1.09 [0.52–2.27]||1.19 [0.54–2.63]|
| Arg/Pro + Pro/Pro||93 (49.8)||207 (43.7)||1.26 [0.90–1.78]||1.23 [0.86–1.77]|
|Men||255||431|| || |
| Arg/Arg||134 (52.5)||232 (53.8)||1.00 [referent]||1.00 [referent]|
| Arg/Pro||104 (40.8)||173 (40.1)||1.06 [0.76–1.46]||1.06 [0.76–1.47]|
| Pro/Pro||17 (6.7)||26 (6.0)||1.09 [0.79–1.51]||1.13 [0.81–1.57]|
| Arg/Pro + Pro/Pro||121 (47.5)||199 (46.2)||1.07 [0.78–1.47]||1.08 [0.79–1.49]|
|Pooled2,5||442||904|| || |
| Arg/Arg||228 (51.6)||498 (55.1)||1.00 [referent]||1.00 [referent]|
| Arg/Pro||186 (42.1)||351 (38.8)||1.16 [0.91–1.47]||1.13 [0.88–1.45]|
| Pro/Pro||28 (6.3)||55 (6.1)||1.09 [0.81–1.47]||1.13 [0.83–1.54]|
| Arg/Pro + Pro/Pro||214 (48.4)||406 (44.9)||1.15 [0.92–1.45]||1.14 [0.90–1.45]|
Table 5. Multivariate Odds Ratios for the Association between p53 Codon 72 Genotype and Colorectal Adenoma1 in Women and Men, by Site
| Arg/Arg||288 (58.8)3||74 (53.6)||1.00 [referent]4||124 (54.2)||1.00 [referent]||32 (52.5)||1.00 [referent]|
| Arg/Pro||172 (35.1)||48 (34.8)|| ||93 (40.6)|| ||25 (41.0)||1.52 [0.85–2.74]|
| Pro/Pro||30 (6.1)||16 (11.6)||1.32 [0.88–1.98]||12 (5.2)||1.25 [0.89–1.76]||4 (6.6)||1.52 [0.85–2.74]|
| Arg/Arg||381 (54.9)||31 (54.4)||1.00 [referent]||102 (52.9)||1.00 [referent]||39 (54.9)||1.00 [referent]|
| Arg/Pro||275 (39.6)||21 (36.8)|| ||76 (39.4)|| ||30 (42.3)|| |
| Pro/Pro||38 (5.5)||5 (8.8)||1.01 [0.57–1.80]||15 (7.8)||1.11 [0.80–1.55]||2 (2.8)||1.04 [0.62–1.73]|
Table 6. Multivariate Odds Ratios for the Association between p53 Codon 72 Genotype and Colorectal Cancer in Women and Men, by Site
| Arg/Arg||266 (56.2)3||27 (36.0)||1.00 [referent]4||37 (64.9)||1.00 [referent]||20 (55.6)||1.00 [referent]|
| Arg/Pro||178 (37.6)||42 (56.0)||2.59 [1.49–4.52]||19 (33.3)||0.66 [0.36–1.23]||14 (38.9)|| |
| Pro/Pro||29 (6.1)||6 (8.0)||1 (1.8)||2 (5.6)||0.76 [0.34–1.71]|
| Arg/Arg||232 (53.8)||54 (62.8)||1.00 [referent]||34 (38.2)||1.00 [referent]||31 (53.5)||1.00 [referent]|
| Arg/Pro||173 (40.1)||30 (34.9)|| ||46 (51.7)|| ||21 (36.2)|| |
| Pro/Pro||26 (6.0)||2 (2.3)||0.70 [0.42–1.15]||9 (10.1)||2.09 [1.28–3.40]||6 (10.3)||1.06 [0.59–1.88]|
In this analysis of men and women in the NHS, HPFS and PHS, the p53 Arg72Pro polymorphism was not associated with colorectal cancer; however, there was a possible modest increased risk of colorectal adenoma associated with the Pro carrier genotypes. Analyses of gene–environment interactions by previously identified colorectal cancer risk factors did not reveal any subgroups of men or women among which this polymorphism was more strongly associated with colorectal adenoma or cancer, although there was some suggestion of an interaction with postmenopausal hormone use and the risk of cancer. In addition, the observed elevation in risk for adenoma did not appreciably vary by colorectal site. However, when site was taken into account for colorectal cancer, the Pro carrier genotypes were more strongly associated with proximal colon cancers in women while in men the association was stronger for distal colon cancers. No association with Arg72Pro genotype was observed for rectal cancers among both men and women.
The association between this polymorphism and cancer risk has been most extensively examined for cancers of the cervix and lung, with some studies showing an increased risk of cervical cancer with the Arg/Arg genotype,22, 23 while for lung cancer, an increased risk has been associated with the Pro carrier genotypes.24 However, the findings for both cervical and lung cancers have been very inconsistent. The association between the p53 Arg72Pro polymorphism and the risk of colorectal cancer has been previously examined in only four studies.10, 12, 14, 15 In an additional 2 studies of lung cancer and the Arg72Pro polymorphism, colorectal cancer cases were included as a second control group.11, 13 The findings from these 6 studies have not strongly supported an association between the p53 Arg72Pro polymorphism and colorectal cancer. In 2 studies, the reported genotype distributions corresponded to a nonsignificant12 and marginally significant15 1.3-fold increase in colorectal cancer risk associated with the Pro carrier genotypes, though in one of these studies the association disappeared when adjusted for another p53 polymorphism involving a 16-bp duplication in intron 3.15 In another study, the Pro carrier genotypes were associated with a nonsignificant 40% reduction in colorectal cancer risk,14 while in 3 other studies, the observed associations were null.10, 11, 13 Analyses by colorectal cancer site was conducted in only 1 study, and the genotype distributions were not found to vary greatly by site among 115 men and women (combined) with colorectal cancer.10 In a study of men and women with colorectal cancer, the Pro carrier genotypes were found to be associated with increasing stage of colorectal cancer.25 The association between the p53 Arg72Pro polymorphism and colorectal adenomas has not been previously reported.
Our results are consistent with previous findings and do not support an appreciable association between the p53 Arg72Pro polymorphism and colorectal cancer overall. However, we were able to further examine whether the association for the p53 Arg72Pro polymorphism differed by colorectal site and sex and observed increased risks of proximal colon cancers in women and distal colon cancers in men associated with the Pro carrier genotypes. The Pro carrier genotypes were also modestly associated with the risk of colorectal adenoma overall. It is possible that these observed associations are, in fact, due to another polymorphism in linkage disequilibrium with the Arg72Pro SNP, such as the intron 3 16-bp duplication polymorphism.15 Alternatively, the Pro form of p53 compared to the Arg form has been experimentally shown to activate transcription4 and induce cell cycle arrest7 to a higher level than the Arg form, and the Pro form has also been shown to be less efficient at inducing apoptosis.4, 6, 7 Thus, the increased risk of colorectal adenoma associated with the Pro carrier genotypes may suggest that apoptosis is a more important mechanism of tumor suppression during the earlier stages of colorectal carcinogenesis. For invasive disease, the tumor suppression mechanism of p53 appears to vary by site and sex with less efficient apoptosis being a more important influence on progression toward proximal colon cancers in women and distal colon cancers in men.
Although chance may account for the differences in the site- and sex-specific associations, given the relatively small numbers among each subgroup, several lines of evidence from embryology, physiology and epidemiology support the notion that cancers of the proximal and distal colon are somewhat different diseases.26, 27, 28 Recent evidence shows that 2 distinct pathways of genetic instability contribute to carcinogenesis in the colon, with chromosomal instability occurring more often in cancers of the distal colon, while microsatellite instability predominates in proximal colon cancers.27, 28 Thus, the importance of p53-induced cell cycle arrest versus apoptosis may vary according to the pertinent etiological pathway. The reason why the site-specific associations were different for men and women is unclear given that the morphology of the colorectum and colorectal cancers do not differ by sex.29 However, differences between men and women have been observed in the descriptive epidemiology of colon cancer.30 In particular, proximal colon cancers have been shown to be more frequent than distal colon cancers in women, while the converse is true for men.30 Interestingly, postmenopausal hormone use among women has been inversely associated with the risk of colorectal cancer.31 In fact, there was some suggestion in our study, though not quite statistically significant, that the Pro carrier genotypes compared to Arg/Arg were associated with a stronger risk of colorectal cancer among women who were currently using postmenopausal hormones compared to never or past users, suggesting that hormones may modify the association between the p53 polymorphism and colorectal cancer. It is possible that the differences between men and women in the observed site-specific associations may reflect a hormonal influence on the different pathways involved in colorectal carcinogenesis.29 Unfortunately, because of the small numbers of cases, we were not able to reliably examine the interaction between Arg72Pro genotype, colon site and postmenopausal hormone use together.
This study had a large sample size compared to previous studies, with 442 participants with colorectal cancer and 857 with colorectal adenoma, allowing us to conduct analyses separately by sex and colorectal site and to examine potential effect modification by colorectal cancer risk factors. Because each case–control study was nested within large prospective cohorts, case and control participants were selected from the same underlying population, thus minimizing the potential for selection bias. In addition, we used a high-quality genotyping method,19 which yielded 100% concordance on the quality control samples, and genotype distributions were in Hardy–Weinberg equilibrium, suggesting that the potential for genotype misclassification was minimal. Although the p53 Arg72Pro genotype distribution has been shown to vary according to ancestral origin,32, 33 over 96% of the study population overall was of Caucasian ancestry. Thus, bias due to population stratification was not likely of concern.34 In a recent analysis of the controls in the NHS, PHS and HPFS, the genotype distributions of 100 SNPs were found to be very similar across studies (personal communication, P. Kraft), further suggesting genetic similarity among individuals and a low likelihood of population stratification.
In summary, our results may suggest that a reduced efficiency of p53-induced apoptosis due to the presence of the Pro allele may play a role in the early stages of colorectal neoplasia, and as well, may be a factor in progression to invasive disease when colorectal site and sex are taken into account. Further studies are warranted to confirm these findings.
We are grateful to the participants of the Nurses' Health Study, the Physicians' Health Study and the Health Professionals Follow-up Study for their participation, and Mr. Hardeep Ranu and Ms. Patrice Soule for technical assistance.