|Home | About | Journals | Submit | Contact Us | Français|
Mutations in PIK3CA (the gene encoding the p110α catalytic subunit of phosphatidylinositide-3-kinase, PI3K) play an important role in colorectal carcinogenesis. Experimental evidence suggests that PIK3CA exon 9 and exon 20 mutations trigger different biological effects, and that concomitant mutations in both exons 9 and 20 synergistically enhance tumorigenic effects. Thus, we hypothesized that PIK3CA exon 9 and exon 20 mutations might have differential effects on clinical outcome in colorectal cancer, and that concomitant PIK3CA exon 9 and 20 mutations might confer aggressive tumor behavior.
We sequenced PIK3CA by pyrosequencing in 1170 rectal and colon cancers in two prospective cohort studies, and found 189 (16%) PIK3CA-mutated tumors. Mortality hazard ratio (HR) according to PIK3CA status was computed using Cox proportional hazards model, adjusting for clinical and molecular features including microsatellite instability, CpG island methylator phenotype, LINE-1 methylation, and BRAF and KRAS mutations.
Compared to PIK3CA wild-type cases, patients with concomitant PIK3CA mutations in exons 9 and 20 experienced significantly worse cancer-specific survival [log-rank P=0.031; multivariate HR=3.51; 95% confidence interval (CI), 1.28–9.62] and overall survival (log-rank P=0.0008 ; multivariate HR=2.68; 95% CI, 1.24–5.77). PIK3CA mutation in either exon 9 or 20 alone was not significantly associated with patient survival. No significant interaction of PIK3CA mutation with BRAF or KRAS mutation was observed in survival analysis.
Co-existence of PIK3CA (the PI3K p110α subunit) exon 9 and 20 mutations, but not PIK3CA mutation in either exon 9 or 20 alone, is associated with poor prognosis of colorectal cancer patients.
Phosphatidylinositide-3-kinases (PI3K) are lipid kinases that promote various biological processes including cellular proliferation and survival (1). Mutations in the PIK3CA gene, which encodes the p110α catalytic subunit of PI3K, have been identified in many human solid tumors, including colon, breast, brain, ovarian, liver, and lung cancers (1). In colorectal cancers, PIK3CA mutations, which are found in 10–20% of tumors, have been reported to be associated with specific clinicopathological features and molecular events, such as proximal tumor location, microsatellite instability (MSI), and KRAS mutation (2–10).
The prognostic significance of PIK3CA mutation in colorectal cancer remains unclear (Table 1) (2, 4, 7, 8, 11–16). The majority of activating PIK3CA mutations map to three sites: exon 9, codons 542 and 545 in the helical domain, and exon 20, codon 1047 in the kinase domain. Mutation at any one of these sites has been shown to result in a gain of enzymatic function and to promote oncogenic transformation in vitro and in vivo (17–19). Interestingly, the mechanisms through which helical and kinase domain mutations augment enzyme function differ (20). Furthermore, the coexistence of mutations in both exons 9 and 20 of the same p110α molecule (PIK3CA) leads to a synergistic gain of function, with a potent transforming capacity in vitro (20). Thus, we hypothesized that PIK3CA exon 9 and exon 20 mutations might have differential effects on tumor behavior, and that the coexistence of mutations in both exons 9 and 20 might result in more aggressive tumor behavior compared to cancers with wild-type PIK3CA, or a single mutation in either exon 9 or exon 20.
The interaction of EGF with EGFR triggers two main signaling pathways, RAS-RAF-MAPK and PI3K-AKT. Activation of these pathways by mutations in KRAS, BRAF and/or
PIK3CA is an established mechanism that drives colorectal carcinogenesis (21). In thyroid cancers, the coexistence of BRAF and PIK3CA mutations is associated with aggressive tumor behavior (22, 23). Based on these findings, our third hypothesis was that PIK3CA and BRAF mutations might interact synergistically to confer a more aggressive colorectal cancer phenotype.
In order to test these hypotheses, we utilized our molecular pathological epidemiology (24–26) database based on two ongoing U.S. nationwide prospective cohort studies. We assessed various additional molecular features, including KRAS mutation, CpG island methylator phenotype (CIMP), microsatellite instability (MSI), TP53 negativity, and LINE-1 hypomethylation, and could therefore control for confounding by these potential predictors of outcome.
We used the database of two prospective cohort studies, the Nurses’ Health Study (NHS, N = 121,700 women observed since 1976) and the Health Professionals Follow-Up Study (HPFS, N = 51,500 men observed since 1986). Every two years, participants were sent follow-up questionnaires to update information on potential risk factors, and to identify newly diagnosed cancers and other diseases. Paraffin embedded tissue blocks were collected from hospitals where participants with colorectal cancer underwent resection of their primary tumors. We collected diagnostic biopsy specimens for rectal cancers patients who received per-operative therapy in order to avoid treatment-related artifact or bias. The tissue retrieval rate was approximately 70% when specimens were requested within five years of diagnosis. All colorectal cancer cases were confirmed through review of histology by a pathologist (S.O.) blinded to other data. Tumor grade was categorized as high (≤50% glandular area) or low (>50% glandular area). Based on the availability of DNA (at least some amount of DNA was available in 1267 cases), PIK3CA sequencing data, and survival data, a total of 1170 colorectal cancer cases diagnosed up to 2006 were included in this study. Patients were observed until death, or January 2011, whichever came first. Ascertainment of deaths included reporting by family members or, where study correspondence had been returned, by postal authorities. The National Death Index was used to ensure completeness of ascertainment. The cause of death was assigned by study physicians. Written informed consent was obtained from all study subjects. Tissue collection and analyses were approved by the Human Subjects Committees at Harvard School of Public Health and Brigham and Women’s Hospital.
Genomic DNA was extracted from paraffin-embedded tissue. Methods for PCR and pyrosequencing targeted at PIK3CA exons 9 and 20 were adapted from those previously-described (10) with the following modifications: we replaced the sequencing primer PIK3CA 9-RS2 with 5′-TTCTCCTT/GCTT/CAGTGATTT-3′, and employed a new nucleotide dispensation order (ATACACATGTCAGTCAGACTAGCTAGCTAGCTAG), which was particularly sensitive for c.1624G>A; sequencing primer PIK3CA 9-RS3 was replaced with 5′-TAGAAAATCTTTCTCCTGCT-3′, and a new dispensation order (ATAGCACTGACTGACTGACTACTGACTGACTG) used to detect the most common mutations, c.1633G>A and c.1624G>A.
PCR and pyrosequencing assays targeted at BRAF (codon 600) (27) and KRAS (codons 12 and 13) mutations (28) were performed as previously described. Microsatellite instability was assessed using a panel of 10 microsatellite markers (D2S123, D5S346, D17S250, BAT25, BAT26, BAT40, D18S55, D18S56, D18S67 and D18S487) (29). MSI-high was defined as the presence of instability in ≥30% of the markers, and MSI-low/microsatellite stability (MSS) as 0–29% unstable markers (29).
Sodium bisulfite treatment of DNA and real-time PCR assays (MethyLight) were performed as previously described (29, 30). We quantified promoter methylation at eight CIMP-specific loci: CACNA1G, CDKN2A (p16), CRABP1, IGF2, MLH1, NEUROG1, RUNX3, and SOCS1 (31–33). CIMP-high was defined as ≥6 (of 8) methylated promoters, and CIMP-low/0 as 0–5 (of 8) methylated promoters (32). To accurately quantify small methylation changes on a background of relatively high methylation, a LINE-1 PCR-pyrosequencing assay was employed (34, 35).
Tissue microarray blocks were constructed and immunohistochemistry for TP53 (p53) was performed (36). Positive and negative controls were included in each run of immunohistochemistry. All immunostaining slides were scored by a pathologist (S.O.) blinded to other data. A random sample of 118 tumors was re-examined by a second observer (K.N.) unaware of other data. The concordance between the two observers was 0.87 (κ = 0.75; P < 0.0001), indicating substantial agreement.
All statistical analyses were performed using SAS software (version 9.1; SAS Institute Inc, Cary, North Carolina). For categorical data, the chi-square test or Fisher’s exact test was performed. All P values were two sided. When multiple hypothesis testing was performed, the P value for significance was adjusted to P = 0.0038 (= 0.05/13) by Bonferroni correction. To compare mean age and mean LINE-1 methylation levels, a t-test or ANOVA, assuming equal variances, was performed. To assess whether associations between PIK3CA mutation and the variables in Table 2 were independent of other variables, a multivariate logistic regression analysis was conducted for cross sectional analyses. Odds ratios (OR) were adjusted for age at diagnosis (continuous), sex, tumor location (proximal vs. distal), CIMP status (high vs. low/0), MSI status (high vs. low/MSS), LINE-1 methylation (continuous), BRAF mutation and KRAS mutation. A backward stepwise elimination with a threshold of P = 0.05 was used to select variables in the final model to avoid overfitting.
The Kaplan-Meier method and the log-rank test were performed for survival analysis. Deaths from causes other than colorectal cancer were censored in colorectal cancer-specific mortality analyses. We performed power calculations. Assuming a total number of patients of 1170, 7 cases with PIK3CA mutations in both exons 9 and 20, 50% mortality, and an alpha (type I error rate) of 0.05, there was a 50% power to detect an HR of 4.6. To control for confounding, we used Cox proportional hazards models to calculate HR of death according to tumor PIK3CA status, adjusting for age at diagnosis (continuous), sex (NHS vs. HPFS), year of diagnosis (continuous), tumor location (proximal vs. distal colon vs. rectum), tumor grade, MSI (high vs. low/MSS), CIMP (high vs. low/0), LINE-1 methylation (continuous), KRAS mutation, and BRAF mutation. To minimize residual confounding and overfitting, disease stage (I, II, III, IV, or unknown) was used as a stratifying variable using the “strata” option in the SAS “proc phreg” command. To avoid overfitting, variables in the final model were selected using backward stepwise elimination with a threshold of P = 0.05. Interaction was assessed using the Wald test on the cross-product of PIK3CA and another variable of interest (excluding cases missing data) in a multivariate Cox model. The proportionality of hazards assumption was satisfied by evaluating time-dependent variables, which were the cross products of the PIK3CA variable and survival time (P = 0.44 for colon cancer-specific mortality; P = 0.95 for overall mortality).
To avoid overfitting, cases with missing data in any of the categorical variables [tumor location (0.6%), tumor grade (0.6%), CIMP (6.6%), MSI (1.4%), BRAF (0.5%), and KRAS (0.4%)], were included in the majority category for that variable. We confirmed that excluding cases with missing information in any of the covariates did not substantially alter the results (data not shown).
PIK3CA pyrosequencing analysis was successful in 95.6% (1212/1267) of colorectal cancers. Pyrosequencing has been shown to be a reproducible, precise and sensitive method of mutation analysis in paraffin-embedded tumor tissues (10, 28). Cases lacking survival data were excluded. Of the remaining 1170 colorectal cancer cases, 536 cases were from HPFS (men) and 634 cases were from NHS (women).
PIK3CA mutation was detected in 189 (16%) out of 1170 cases, among which 109 cases (58%) had mutations in only exon 9, 73 cases (39%) had mutations in only exon 20, and seven cases (3.7%) had mutations in both exons 9 and 20. Supplementary Table 1 shows the frequencies of specific PIK3CA mutations. The relationships between PIK3CA mutation and clinical, pathological, and molecular features in each cohort (NHS and HPFS) are presented separately in Supplementary Table 2 and Supplementary Table 3, and demonstrate general consistency between the two cohorts. Table 2 summarizes clinical, pathological and molecular features of colorectal cancer according to PIK3CA mutation status.
In contrast to mutations in exon 9, mutations in exon 20 were associated with MSI-high (P = 0.0007), CIMP-high (P = 0.028), and BRAF mutation (P = 0.030). Associations between concomitant PIK3CA exon 9 and exon 20 mutation status and family history of colorectal cancer (P = 0.030), MSI-high (P = 0.0050), CIMP-high (P = 0.025), and TP53 negativity (P = 0.0080) were of borderline significance taking into account multiple hypothesis testing (requiring P = 0.0038 as a significance level). Supplementary Table 4 shows the clinical, pathological, and molecular features of colorectal cancers categorized by overall PIK3CA mutation status (any mutation vs. no mutation). PIK3CA overall mutation status was significantly associated with KRAS mutation (P < 0.0001), however there was no significant difference in the frequency of KRAS mutation between exon 9 and exon 20 mutants (P = 0.29).
Detailed information on the seven cases with concomitant PIK3CA exon 9 and exon 20 mutations is shown in Table 3. One case had three PIK3CA mutations (c.1624G>A, c.1631C>A and c.3140A>G). Although the number of patients with cancers harboring concomitant PIK3CA exon 9 and exon 20 mutations was small, this subgroup appeared to have an association with family history of colorectal cancer. Notably, all seven patients died of colorectal cancer or other causes within the follow-up period.
We performed multivariate logistic regression analysis to assess for independent relationships between KRAS mutation, and other factors, and PIK3CA overall mutation. In multivariate model analysis, KRAS mutation remained significantly associated with PIK3CA overall mutation [multivariate odds ratio (OR) = 2.65; 95% confidence interval (CI), 1.89–3.73; P < 0.0001]. In addition, CIMP-status remained in the final model (multivariate OR = 1.65; 95% CI, 1.07–2.54; P = 0.024), with borderline significance given multiple hypothesis testing (Supplementary Table 5).
We assessed the prognostic role of PIK3CA mutation in 1170 colorectal cancers to test the hypotheses that PIK3CA exon 9 and exon 20 mutations might have differential effects on tumor behavior, and that the presence of mutations in both exon 9 and exon 20 might result in more aggressive tumor behavior. During a median follow-up period of 141 months for survivors (interquartile range, 105–192), there were 552 deaths, including 328 colorectal cancer-specific deaths. Notably, patients with PIK3CA mutations in both exons 9 and 20 (henceforth referred to as “exon 9 and 20 double mutants”) experienced significantly shorter cancer-specific survival (log-rank P = 0.031) (Figure 1A) and overall survival than patients with wild-type PIK3CA (log-rank P = 0.0008) (Figure 1B). In Cox regression analysis, compared to PIK3CA wild-type cases, exon 9 and 20 double mutant status was associated with significantly higher colorectal cancer-specific mortality (univariate HR = 2.84; 95% CI, 1.05–7.69; multivariate HR = 3.51; 95% CI, 1.28–9.62) and overall mortality (univariate HR = 3.37; 95% CI, 1.58–7.15; multivariate HR = 2.68; 95% CI, 1.24–5.77) (Table 4).
In contrast, the presence of a single PIK3CA mutation, in either exon 9 or 20, was not significantly associated with patient survival (Figure 1A, 1B; Table 4). The prognostic association of PIK3CA exon 9 mutations in colorectal cancer (regardless of exon 20 status) was examined. No significant difference was found between exon 9 mutants and wild-type cases in colorectal cancer-specific or overall survival. The HRs for exon 9 mutants were not influenced by the inclusion of double mutants, even after adjusting for exon 20 mutation status. Likewise, the prognostic association of PIK3CA exon 20 mutations (regardless of exon 9 status) was similar to wild-type cases in colorectal cancer-specific and overall survival analysis (Supplementary Table 6). Overall PIK3CA mutation status (exon 9 or 20 mutation) was not significantly associated with colorectal cancer-specific or overall survival when compared to wild-type cases (Figure 1C, 1D; Table 4). When each cohort was analyzed separately, overall PIK3CA mutation status was not significantly associated with colorectal cancer-specific or overall survival (Supplementary Table 7). HRs were similar for both cohorts and 95% CI were largely overlapping, showing the consistency of results between the two cohorts.
To test the third hypothesis, that the presence of both BRAF and PIK3CA mutations might result in aggressive tumor behavior, we examined combined BRAF and PIK3CA mutation status and patient prognosis (Supplementary Table 8). Compared to PIK3CA wild-type/BRAF wild-type cases, the presence of mutations in both PIK3CA and BRAF was not significantly associated with colorectal cancer-specific mortality in univariate analysis (HR = 1.24; 95% CI, 0.61–2.52). However, in multivariate analysis, the presence of mutations in both PIK3CA and BRAF was significantly associated with colorectal cancer-specific mortality (multivariate HR = 2.40; 95% CI, 1.12–5.16). We found that MSI and CIMP status were confounders; when we simply adjusted for MSI and CIMP, the adjusted HR (PIK3CA mutated/BRAF mutated vs. PIK3CA wild-type/BRAF wild-type) was 3.08 (95% CI, 1.44–6.61).
We also examined the influence of KRAS and BRAF mutation status on the prognostic association of mutations in PIK3CA. We classified colorectal cancers into three subtypes according to KRAS and BRAF status: BRAF wild-type/KRAS wild-type, BRAF-mutated/KRAS wild-type, and BRAF wild-type/KRAS-mutated. No substantial effect modification by KRAS/BRAF mutation status was observed in survival analyses (Table 5).
In further exploratory analyses, the prognostic effect of PIK3CA mutation in strata of other variables was evaluated. The effect of PIK3CA on cancer-specific mortality did not significantly differ according to disease stage (Pinteraction = 0.93), tumor location (Pinteraction = 0.099), or any of the other variables examined (all Pinteraction > 0.05).
We conducted this study to test the hypotheses that PIK3CA exon 9 and exon 20 mutations might have differential effects on colorectal cancer behavior, and that the presence of concomitant mutations in both exons 9 and 20 might lead to aggressive tumor behavior. We found no significant association between overall or exon-specific PIK3CA mutation status and survival. The concomitant presence of mutations in both exons 9 and 20 was, however, associated with a poorer prognosis for colorectal cancer patients, although a confirmation by other studies would be essential. Our data support the hypothesis that concomitant exon 9 and 20 mutations may have a synergistic effect on tumor behavior, and are consistent with experimental data by Zhao et al (20), that show potent synergistic transforming effects of concomitant PIK3CA exons 9 and 20 mutations. Patients with concomitant mutations in exons 9 and 20 of PIK3CA were more likely to report a family history of colorectal cancer. At present, the cause of this potential association remains obscure. One could speculate that a family history of colorectal cancer might confer a genetic predisposition to the development of the PIK3CA mutations. The number of cases with the concomitant PIK3CA exon 9 and 20 mutations was, however, very small, and we should exercise caution in the interpretation of any apparent clinical, pathological and molecular associations within this subgroup. These findings must be validated by independent datasets.
The assessment of prognostic factors or biomarkers is important in cancer research (37–45). A number of previous studies have examined the prognostic role of PIK3CA mutation in colorectal cancer (Table 1) (2, 4, 7, 8, 11–16, 46). Although some studies have shown that PIK3CA mutation is associated with shorter survival (2, 11), the statistical power of these studies is quite limited (sample size N < 160). In a larger study (N = 586), the presence of a mutation in any of PIK3CA, BRAF or KRAS was associated with poor three-year survival, but the effect of PIK3CA mutation in isolation was not studied (7). Two additional studies demonstrated that, compared to patients with wild-type PIK3CA, patients with exon 20 mutations experienced worse survival while exon 9 mutation was not associated with outcome (14, 16). In one of these two studies, tumor KRAS and BRAF mutations were not examined (16). In our current study, cancers with PIK3CA exon 20 mutation were found to have a higher frequency of BRAF mutation. BRAF mutation has been associated with a poorer prognosis in colorectal cancer (47), and might therefore confound analyses of PIK3CA exon 20 mutation and survival in colorectal cancer. Other groups have reported that PIK3CA mutations are not associated with liver metastasis (48) or with overall survival (with or without adjuvant treatment) (4, 8, 15).
In our current study, neither PIK3CA overall mutation status, nor PIK3CA mutation in exon 9 or 20 alone, was significantly associated with patient survival. This is in contrast to some of the published literature. Notwithstanding the potential for confounding by BRAF status in other studies, it is worth bearing in mind that small studies with null results have a higher probability of being unpublished compared to similarly sized datasets with “significant” findings. Large studies with adequate statistical power are less prone to this type of “publication bias”. We should therefore place more emphasis on the results of large-scale studies when we evaluate publications on the prognostic significance of cancer biomarkers. Furthermore, experimental evidence suggests that PIK3CA mutation alone has a relatively modest effect on tumor cell growth, and that PIK3CA mutations need to cooperate with other PI3K enzyme mutations for effective cellular transformation (17, 49). Because of the power limitations in the previous studies on PIK3CA mutation, we feel that there is currently insufficient evidence to support a role for PIK3CA exon 9 or 20 mutation alone as a prognostic biomarker in colorectal cancer. Our findings warrant validation in additional large cohort studies.
We previously described an association between PIK3CA mutation and shorter cancer-specific survival among 450 stage I-III colon cancer cases (12). All of those 450 cases were included in our current study. Sample size and adequate statistical power are critically important in such exploratory studies (48). Since our previous study was restricted to stage I-III colon cancers, the numbers of adverse events (66 colon cancer-specific deaths and 152 overall deaths) was also much smaller than in our current study (328 colorectal cancer-specific deaths and 552 overall deaths), which included cancers of all stages. As a result, our current findings are more robust than those of our previous study where, as a result of smaller sample size and lower power, there would have been increased risk of finding chance associations. This underscores the critical importance of careful study design, adequate statistical power, and cautious interpretation of data, which are prerequisites for exploratory studies of this nature (50).
Even in our larger dataset of 1170 colorectal cancers, there were only seven patients who harbored PIK3CA mutations in both exons 9 and 20. However, given that over 550,000 individuals are diagnosed with colorectal cancer each year in the U.S. and Europe, we estimate that, in these regions combined, there would be approximately 3,300 colorectal cancer patients every year with PIK3CA mutations in both exons 9 and 20. The incidence of this potentially aggressive type of colorectal cancer may in fact be similar to the combined sum of the incidences of Burkitt lymphoma, hairy cell leukemia, ALK-positive large B-cell lymphoma, and angioimmunoblastic T-cell lymphoma in these western countries. Other cancers with a similar incidence include osteosarcoma, medulloblastoma, gestational choriocarcinoma, and ovarian clear cell carcinoma. Thus, those colorectal cancers with PIK3CA mutations in both exons 9 and 20 may represent as significant a cancer burden as these other cancer types in our society.
Caveats of our current study include the limited data on cancer treatment in the cohorts, which prevented the inclusion of treatment as a variable in our analyses. Nonetheless, it is unlikely that chemotherapy use or regimens differed substantially by PIK3CA mutation status given that a vast majority of cases were diagnosed prior to 2006, and PIK3CA mutation data were unavailable to physicians or patients. In addition, our multivariable Cox regression analysis adjusted for disease stage (I, II, III, or IV), on which treatment decision-making was mostly based. Colorectal cancer-specific survival is therefore a reasonable colorectal cancer-specific outcome.
Our findings relating to survival in patients whose tumors harbored mutations in both exons 9 and 20 of PIK3CA are novel. However, given that the number of such cases in our study was small, and statistical power was consequently limited, these findings warrant validation by independent studies.
Our study gains several strengths through utilization of the database of two U.S. nationwide prospective cohort studies. Clinicopathological information, various exposures, and tumor molecular data have been integrated into our molecular pathological epidemiology (24–26) database. Cohort participants with colorectal cancer sought medial attention and were treated at hospitals throughout the U.S. Hence, our sample is more representative of colorectal cancer in the general U.S. population than a convenience sample collected at one or a few hospitals. Moreover, our extensive tumor database enabled us to assess the prognostic association of PIK3CA mutations independent of other critical molecular events such as BRAF and KRAS mutations, LINE-1 hypomethylation, MSI, and CIMP, which have all been associated with colon cancer outcome (29, 34).
In conclusion, in our study of 1170 colorectal cancers, concomitant PIK3CA mutation of both exons 9 and 20 was associated with a poorer prognosis, although a statistical power was limited due to only 7 cases with the concomitant mutations. In contrast, neither PIK3CA exon 9 nor exon 20 mutation alone appeared to have substantial prognostic influence. The robustness of our findings would be enhanced by replication in other large studies. Our findings might give additional insight into the relevance of the PI3K pathway in colorectal cancer progression, and suggest that detailed genotyping of PIK3CA might serve towards personalized medicine.
PIK3CA mutation is present in various human cancers, and plays a role in cancer cell proliferation and survival. The relationship between PIK3CA mutation in colorectal cancer and patient survival remains controversial. In this study we utilized a database of 1170 colorectal cancers in two prospective cohort studies. Our study benefitted from adequate participant follow-up, and the availability of clinical information and data on additional molecular characteristics that are important in colorectal carcinogenesis. This is, by far, the largest study on the prognostic role of PIK3CA mutations in colorectal cancer to date, and suggests that patients with concomitant PIK3CA mutations in both exons 9 and 20 might be associated with worse survival. The presence of a single PIK3CA mutation in either exon 9 or 20 was not significantly associated with patient survival. Considering the role of the PI3K signaling pathway in cancer, our findings might be relevant towards personalized medicine.
Funding: This work was supported by U.S. National Institute of Health [P01CA87969 (to S.E. Hankinson), P01CA55075 (to W.C. Willett), P50CA127003 (to C.S.F.), and R01CA151993 (to S.O.)]; the Bennett Family Fund for Targeted Therapies Research; and the Entertainment Industry Foundation through National Colorectal Cancer Research Alliance. The content is solely the responsibility of the authors and does not necessarily represent the official views of NCI or NIH. Funding agencies did not have any role in the design of the study; the collection, analysis, or interpretation of the data; the decision to submit the manuscript for publication; or the writing of the manuscript.
We would like to thank the participants and staff of the Nurses’ Health Study and the Health Professionals Follow-Up Study, for their valuable contributions as well as the U.S. State Cancer Registries for their help.
Disclosure of Potential Conflict of Interest: no potential conflicts of interest exist.