Effects of Supplemental Calcium and Vitamin D on Expression of Toll-Like Receptors and Phospho-IKKα/β in the Normal Rectal Mucosa of Colorectal Adenoma Patients

Chronic inflammation has been linked to the development and progression of cancer (1). Inflammation in the gut can occur due to innate immune responses to changes in the microbiota and exposure to pathogenic bacteria (2). In the colorectum, Toll-like receptors (TLR) play a key role in the immune response by recognizing pathogen-associated molecular patterns (PAMP). Toll-like receptor 4 (TLR4) recognizes lipopolysaccharides (LPS), an integral part of the outer membrane of gram-negative bacteria cell walls (3), and TLR5 recognizes flagellin, the major component of the motility apparatus of flagellated bacteria (4). Activation of TLRs induces the nuclear factor κβ (NF-κB)–mediated production and secretion of proinflammatory cytokines, such as TNFα, IL1, and IL6 (4). Activation of the NF-κB signaling pathway was found to contribute to colon cancer development and progression via transcriptional upregulation of cell proliferation and angiogenesis, inhibition of apoptosis, and overexpression of cyclooxygenase-2, which in turn also promotes inflammation and cell proliferation (5, 6).

There is strong experimental, and generally supportive human observational, evidence to support vitamin D's antineoplastic effects (7); however, the exact mechanisms of its antineoplastic actions in the colon are not well understood. Experimental evidence suggests that vitamin D and calcium having direct effects on cell proliferation, differentiation, and apoptosis; vitamin D's modulation of inflammation, growth factor signaling, androgen and estrogen receptor pathways, immune function, and angiogenesis; and calcium's ability to bind bile acids and fatty acids in the stool, resulting in compounds that are less likely to damage colonocyte DNA (7–9). Recent data indicate that calcium and vitamin D may be involved in regulating the TLR4 signaling pathway and maintaining the integrity of the intestinal mucosal barrier, the dysfunction of which results in exposure of the host to luminal bacteria and endotoxins, leading to endotoxemia and chronic colon inflammation (10, 11). Moreover, vitamin D can regulate the expression and release of human antimicrobial peptides (e.g., cathelicidin and defensin β2) that protect against bacterial and viral infections (12, 13). It was also found that the active form of vitamin D, 1,25(OH)₂D₃, may prevent TNFα-induced degradation of IκBα (14–16), leading to retention of NF-κB in the cytoplasm where it cannot induce downstream transcriptional activity (14, 15). By forming a complex with calmodulin, calcium can also attenuate the TLR-mediated response to PAMPs by reducing the accumulation of TNFα and NF-κB (17). Intracellular calcium may also reduce IKK activity, thereby decreasing NF-κB transcriptional activity downstream (18).

All these data support the hypothesis that vitamin D and calcium may beneficially modulate these proinflammatory signaling pathways by reducing the expression of proinflammatory biomarkers. However, no published studies investigated the effects of vitamin D and calcium supplementation on TLR4, TLR5, and NF-κB [as can be assessed by measuring the expression of phospho-IKKα/β (pIKKα/β; ref. 19)] in the normal rectal mucosa. Therefore, our goals were to generate preliminary data on the effects of supplemental vitamin D and calcium, alone and in combination, on the expression of these biomarkers in the normal rectal mucosa of patients at high risk for colorectal neoplasms. We examined pIKKα/β and TLRs expression in the crypt epithelium as well as TLRs expression in the stroma, which may have higher exposure to bacterial products due to its proximity to the lumen and includes hematopoietic and nonhematopoietic cells responsible for multiple important roles in immune responses and initiation of inflammation (20).

The participants in this study (“adjunct biomarker study”) were recruited from participants enrolled in a larger 11-center, randomized, placebo-controlled, partial 2 × 2 factorial chemoprevention clinical trial (“parent study”; Vitamin D and Calcium Polyp Prevention Study) testing the efficacy of supplemental calcium and vitamin D₃, alone or in combination, over 3 to 5 years on colorectal adenoma recurrence in colorectal adenoma patients (21). The parent study protocol and eligibility and exclusion criteria were previously published (21). Briefly, participants needed to be 45 to 75 years of age, have at least one colorectal adenoma removed within 120 days of enrollment with no remaining polyps after a complete colonoscopy, and anticipated to undergo a 3-year or 5-year colonoscopic follow-up examination (21). For participation in the adjunct biomarker study, additional exclusion criteria included medically inadvisable to cease aspirin use for 7 days, history of a bleeding disorder, and current use of an anticoagulant medication.

The study protocol for the biomarker study was also previously published (22). Briefly, patients for the biomarker study were recruited at 2 of the 11 clinical centers (South Carolina and Georgia) between May 2004 and July 2008, and agreed to undergo rectal biopsies at baseline and after 1 year of supplementation with the study agent. Participants were enrolled and consented near the end of the 3-month run-in period for the parent study without knowledge of the assigned treatment. One-hundred and five participants met final eligibility and had sufficient rectal biopsy tissue taken for biomarker measurements at both baseline and at the 1-year follow-up. The Institutional Review Boards at both clinical centers approved this research.

Baseline information was collected on each participant upon enrollment into the parent study, and included medical history, medication and nutritional supplement use, demographics, lifestyle, and diet (using the Block Brief 2000 food frequency questionnaire; NutritionQuest). Serum 25-hydroxyvitamin D [25(OH)D] and calcium concentrations were measured at baseline. Serum 25(OH)D concentrations were also measured at year one follow-up.

Following the 3-month run-in period, most subjects were randomly assigned to the following four treatment groups (4-arm randomization): placebo, 1,200 mg/d calcium supplementation (as calcium carbonate in equal doses twice daily), 1,000 IU/d vitamin D₃ supplementation (500 IU twice daily), and 1,200 mg/d elemental calcium plus 1,000 IU/d vitamin D₃ supplementation. Women who declined to forego calcium supplementation were randomized to calcium or calcium plus vitamin D₃ (2-arm randomization). Randomization was conducted with the use of computer-generated random numbers with permuted blocks and stratified by clinical center, sex, anticipated colonoscopic examination at 3 years or 5 years, and full factorial or two-arm randomization. All study staff, as well as the participants, were blinded to treatment assignments.

During the trial, participants agreed to refrain from taking additional vitamin D or calcium supplements, although personal daily supplements up to 1,000 IU vitamin D and/or 400 mg elemental calcium were permitted starting in April 2008. Bottles of study tablets were delivered to participants every 4 months. Telephone interviews were conducted every 6 months regarding participant adherence to the study treatment, illnesses, use of medications and supplements, and colorectal endoscopic or surgical procedures.

Biopsies were collected from normal-appearing rectal mucosa at baseline and at 1-year follow-up visits without any preceding bowel-cleansing preparation. Specific details of the rectal biopsy procedure and immunohistochemistry protocol were previously published (22). Briefly, six biopsies, approximately 1 mm thick, were taken from the rectal mucosa 10 cm above the external anal aperture. Biopsies were placed onto a strip of bibulous paper and immediately put in normal saline, oriented, transferred to 10% normal-buffered formalin for 24 hours, and then transferred to 70% ethanol. Within a week, the biopsies were processed and embedded in paraffin blocks. For each biomarker, five slides with three levels of 3-μm-thick biopsy sections taken 40 μm apart were prepared, yielding a total of 15 levels. Heat-induced epitope retrieval was performed in a Lab Vision PT Module device (Lab Vision Corp.) in 1x Citrate Buffer with a pH of 6.0 (ThermoScientific, TH 250-Premix). Next, slides were immunohistochemically processed in a DakoCytomation Autostainer Plus System (Agilent Dako) automated immunostainer, using a labeled streptavidin-biotin kit (ThermoScientific UltraVisionKit, TP-125-HL), by applying a rabbit monoclonal antibody against pIKKα/β (catalog no.: 2697P, dilution, 1:150; Cell Signaling Technology), a mouse monoclonal antibody against TLR4 (catalog no.: ab47093, dilution 1:325; Abcam), and a mouse monoclonal antibody against TLR5 (catalog no.: MAB6704, dilution 1:200; R&D Systems). We also used previously collected data on MIB-1/Ki-67 expression (catalog no.: M7240, dilution 1:350; Agilent Dako) in the normal colorectal mucosa to investigate whether baseline levels of colorectal epithelial cell proliferation are associated with expression of TLR4 or TLR5 in colorectal crypts. Baseline and follow-up biopsy slides for individual study participants were included in the same immunohistochemistry batch, and each batch included a balance of participants from each treatment group, plus positive and negative controls. The slides were coverslipped with a Leica CV5000 Coverslipper (Leica Microsystems, Inc.) and then digitized using the PanoramicScan 150 whole slide image scanner (3DHISTECH).

A quantitative image analysis method (“scoring”) was used to quantify biomarker expression in colon crypts. Digital images of baseline and follow-up slides for the 105 participants were reviewed using custom-developed software, CellularEyes (DivEyes LLC). Technicians blinded to treatment assignment selected only “scorable” hemicrypts (Fig. 1A), which were defined as one side of a crypt bisected from base to colon lumen surface, intact, and extended from the muscularis mucosa to the colon lumen. Once a scorable crypt was identified, the technician traced its borders using a digital drawing board. The program then divided the outlined area into 50 equally spaced segments of approximately average normal colonocyte width, and measured the background-corrected optical density of the biomarker labeling across the full length of the hemicrypt, as well as within each segment (Fig. 1B). All resulting data were automatically transferred into a MySQL database (Sun Microsystems, Inc.). For each biomarker, the technician identified and outlined scorable hemicrypts until at least 8 and at most 40 hemicrypts were scored per participant per visit.

To measure biomarker expression in the intercrypt stroma adjacent to a previously scored hemicrypt, another technician would locate the previously scored hemicrypt(s) and visually inspect whether the intercrypt width was sufficient for stoma scoring. If suitable, the technician would then outline the intercrypt stroma, excluding epithelial cells, muscle tissue, and staining artifacts (Fig. 1C). The technician would continue these procedures until a minimum of eight intercrypt stromal regions were scored for each patient visit.

To assess intrareader scoring reliability, reliability control samples previously analyzed by the technician were reanalyzed during the course of the trial. For pIKKα/β, 2 technicians scored 13 batches each, with intraclass coefficients ≥0.97, and an interclass correlation coefficient of 0.98 for the sole technician who scored pIKKα/β expression. For the sole technician who scored TLR4 expression, the intraclass correlation coefficients for crypt and stroma expression were 0.98 and 0.92, respectively. Three technicians scored TLR5 expression in the crypt with average intraclass coefficients of ≥0.98, and an average interclass coefficient of variation was 0.91. For the sole technician who scored TLR5 expression in the stroma, the intraclass correlation coefficient was 0.96.

We compared the baseline characteristics of the participants across treatment groups using the χ² test for categorical variables and ANOVA for continuous variables. To evaluate treatment effects, we assessed differences in pIKKα/β, TLR4, and TLR5 expression from baseline to year 1 follow-up between participants in the treatment group of interest and those in the comparison group using general linear mixed models. The models included as predictors the intercept, visit (baseline and year 1 follow-up), treatment group, and a treatment-by-visit interaction term. We evaluated changes in the expression of pIKKα/β (full length of the crypt only), TLR4 (in the full length of the crypt and stromal region), and TLR5 (in the full length of the crypt and stromal region) for the treatment groups that received (i) calcium relative to those that did not (“calcium versus no calcium,” excluding two-arm participants), (ii) vitamin D relative to those that did not (“vitamin D versus no vitamin D”), and (iii) vitamin D plus calcium relative to those that received only calcium (“vitamin D + calcium versus calcium”). In addition to evaluating biomarker differences in the full length of crypts and stromal regions, we evaluated changes in biomarker expression within the upper 40% of crypts (the canonical differentiation zone) and the lower 60% of crypts (the canonical proliferation zone), and the ratio of expression in the upper 40% to the whole crypt (ϕ_h). To assess changes in the balance of TLR4 and TLR5 relative to pIKKα/β, TLR4 and 5 to pIKKα/β ratios were calculated by dividing an individual's TLR4 and TLR5 expression by their pIKKα/β expression in the full length of the crypt.

Treatment effects were calculated on the ratio and difference scales according to the following: relative effect = [(treatment group follow-up)/(treatment group baseline)]/[(control group follow-up)/(control group baseline)]; absolute effect = [(treatment group follow-up) – (treatment group baseline)] – [(control group follow-up) – (control group baseline)]. A relative effect of 1.3 would indicate a 30% increase in biomarker expression in the treatment group relative to the control group. In all analyses of randomized treatments, participants were retained in their originally assigned treatment group, regardless of adherence to study treatment and procedures.

To assess potential confounders, identified by imbalances in their distribution across treatment groups at baseline (e.g., dietary fiber intake) or by inclusion into the study (e.g., study center), two additional models were run. The first model controlled for age, sex, and study center, whereas the second additionally controlled for smoking status (current/former vs. never), multivitamin use (yes vs. no), physical activity [metabolic equivalent of task (MET)-minutes, continuous], and total energy (continuous) and dietary fiber (continuous) intakes. Adjustment for these potential confounders did not materially affect the estimated treatment effects; therefore, only minimally adjusted results are presented.

Finally, we assessed whether pIKKα/β, TLR4, and TLR5 expression at baseline differed by categories of a priori–selected biologically plausible factors, including age, sex, smoking status, regular NSAID use, regular aspirin use, history of sessile serrated adenomas, body mass index (BMI), red/processed meat intake, and colorectal MIB-1 expression. Means, 95% confidence intervals (CI), and P values were calculated using general linear models adjusted for age, sex (by study arm), study center, and staining batch, where applicable. Proportional differences were calculated according to the following: [(comparison mean – reference mean)/reference mean] × 100%. P values for trend for categorical variables with more than two levels were calculated by treating the ordered categories as a continuous variable in the same general linear model.

All statistical analyses were conducted using SAS 9.4 statistical software. A two-sided P value ≤ 0.05 was considered statistically significant.

Selected baseline characteristics of the 105 adjunct biomarker study participants are presented in Table 1. The mean age of the participants was 59 years, 47% were male, 79% were white, and approximately half held a college degree or higher. Most of the participants were nonsmokers (92%), overweight (79%), and consumed less than 1 drink of alcohol per day on average. There were observed differences in physical activity and dietary fiber intake across treatment groups.

During the first year after randomization, 76% of the participants reported taking 80% or more of their study tablets. At 1 year after randomization, mean serum 25(OH)D concentrations had increased 11 (95% CI, 7–14) ng/mL and 12 (95% CI, 7–16) ng/mL in the vitamin D and calcium + vitamin D groups, respectively, relative to their reference groups.

The estimated treatment effects on the ratio and difference scales for pIKKα/β, TLR4, and TLR5 expression in the full length of crypts and stroma are shown in Table 2. Similar treatment effects were observed in the upper 40% of crypts (the canonical differentiation zone), the lower 60% of crypts (the canonical proliferation zone), and the ratio of expression in the upper 40% to the whole crypt (ϕ_h), and therefore these results are not discussed but are shown in Supplementary Tables S1 to S3, respectively. For each biomarker described below, we include a relative treatment effect on the ratio scale, which is interpreted as a percent change in biomarker expression in the treatment group relative to the control group. For example, a relative effect of 1.33 would indicate a 33% increase in biomarker expression in the treatment group relative to the control group.

In the calcium relative to the no calcium group, pIKKα/β expression increased an estimated 33% (P = 0.18; Table 2) in the full length of crypts. There was a minimal, statistically nonsignificant increase in pIKKα/β expression in the full length of crypts in the vitamin D + calcium relative to the calcium group, but no change in the vitamin D relative to the no vitamin D group (Table 2).

TLR4 expression increased an estimated 9% (P = 0.69; Table 2) in the full length of crypts and decreased 4% (P = 0.87; Table 2) in the stroma in the calcium relative to the no calcium group. In the vitamin D relative to the no vitamin D group, TLR4 expression decreased an estimated 9% (P = 0.60; Table 2) and 16% (P = 0.47; Table 2) in the full length of crypts and stroma, respectively. In the vitamin D + calcium relative to the calcium group, TLR4 expression decreased 10% (P = 0.62) in the full length of crypts, decreased 18% in the stroma (P = 0.50; Table 2), and statistically significantly increased an estimated 12% (P = 0.04; Supplementary Table S3) in the ϕ_h of crypts. The (pIKKα/β)/TLR4 ratio increased in all three crypt parameters for all three treatment groups (Supplementary Table S4), though not statistically significantly so.

In the calcium relative to the no calcium group, TLR5 expression decreased an estimated 6% (P = 0.85) in the full length of crypts and 27% in the stroma (P = 0.31; Table 2). TLR5 expression increased 8% (P = 0.79; Table 2) and 32% (P = 0.24; Table 2) in the full length of the crypts and stroma in the vitamin D relative to no vitamin D group. In the vitamin D + calcium relative to the calcium group, TLR5 expression decreased 5% in the full length of crypts (P = 0.87; Table 2) but increased 23% (P = 0.47; Table 2) in the stroma. TLR5 expression in the ϕ_h of crypts and the (pIKKα/β)/TLR5 ratio did not materially change in any of the treatment groups relative to their reference groups (Supplementary Tables S3 and S4).

Proportional differences of adjusted mean full-length crypt pIKKα/β, TLR4, and TLR5 and stromal TLR4 and TLR5 expression across categories of a priori–selected participant characteristics at baseline are presented in Table 3 (a full version of the table, with means and CIs, is included as Supplementary Table S5). Women, on average, relative to men had lower expression of pIKKα/β (−24.6%, P = 0.09), TLR4 (−30.2%, P = 0.07), and TLR5 (−7.9%, P = 0.78) in the full length of crypts, and of TLR4 (−13.1%, P = 0.58) and TLR5 (−14.1%, P = 0.57) in the stroma. Current smokers, relative to never-smokers, had lower expression of pIKKα/β (−15.2%, P = 0.15) and TLR4 (−10.1%, P = 0.10) in the full length of crypts and TLR4 (−22.6%, P = 0.06) in the stroma. Among those who were obese, relative to those who were normal weight, mean expression of all biomarkers was higher in the full length of crypts and the stroma and statistically significantly so for pIKKα/β (36% higher, P_trend = 0.04). Those with a history of a sessile serrated adenoma had TLR5 expression that was, on average, 94.9% higher (P = 0.01) in the full length of crypts and 49% higher (P = 0.06) in the stroma. pIKKα/β. TLR4 expression was also observed to be higher among those with a history of sessile serrated adenoma, though not statistically significantly so. Among participants who ate one or more servings of red or processed meats, mean pIKKα/β expression was 46% higher (P_trend = 0.05). Among those in the highest relative to the lowest tertile of MIB-1/ki-67 (a biomarker of cell proliferation) expression, mean pIKKα/β, TLR4, and TLR5 expression was 21% (P_trend = 0.32), 23% (P_trend = 0.29), and 10% (P_trend = 0.69) higher in the full length of crypts, respectively.

Our preliminary findings from this pilot study suggest that vitamin D and calcium supplementation, alone or in combination, do not statistically significantly affect pIKKα/β, TLR4, or TLR5 expression in the crypts or in the stroma of patients at high risk for colorectal neoplasms.

Both vitamin D and calcium are hypothesized to have anti-inflammatory properties as experimental studies found that 1,25(OH)₂D₃, the hormonally active form of vitamin D, as well as calcium, can suppress NF-κB–induced TNFα activity (15–17). However, the findings for pIKKα/β and TLRs in our small pilot study are not consistent with such an effect in the normal human colorectal mucosa. Human data regarding a systemic anti-inflammatory effect of these agents are inconclusive and come from clinical studies with small sample sizes, short duration, different vitamin D and calcium doses, and patient populations. Most importantly, no reported studies investigated these inflammation biomarkers in colorectal mucosa tissue (23–30).

We also investigated associations of a priori–selected patient characteristics with pIKKα/β, TLR4, and TLR5 expression at baseline. Our findings suggest that women, relative to men, may have lower pIKKα/β, TLR4, and TLR5 colorectal mucosal expression. Previous studies found that men had higher LPS-induced TNFα production (31) than did women, leading to a sex difference in intensity of immune response (32). The observed difference in TLRs could also be potentially explained by differences in the gut microbiome composition, which was linked to inflammatory response in the colon (33).

Contrary to our original hypothesis, current, relative to nonsmokers, had lower estimated pIKKα/β in the full length of crypts and TLR4 in both the crypts and stroma. There were only 8 current smokers, which did not allow for stable estimates, although directions of the associations were similar to those for former smokers, except for TLR5 for which they were in the opposite directions. Macrophages and monocytes express the α7 subunit of the nicotinic acetylcholine receptor (α7nAChR), and stimulation of these receptors by nicotine activates several signaling pathways that mediate the synthesis and release of proinflammatory cytokines (34, 35). As inflammation is associated with higher TLR4 and TLR5 expression, indicated by pIKKα/β, the inhibition of proinflammatory signaling is likely associated with lower expression. In addition, in animal models, mice exposed to side-stream smoking had altered gut microflora composition (36). A decrease in gram-negative and flagella-supported bacteria, which are recognized by TLR4 and TLR5, could further explain the lower biomarker expression among ever smokers.

Our finding that BMI is positively associated with baseline pIKKα/β, TLR4, and TLR5 expression is largely consistent with previous literature. pIKKα/β is a marker of inflammation, and obesity is considered to be a state of chronic low-grade inflammation. Being obese was also previously found to be associated with higher plasma concentrations of LPS (37), which is recognized by TLR4, and higher circulating concentrations of inflammation markers such as TNFα (24). Another possible explanation is the differences in the gut microbiome between obese and normal weight individuals (38), which could be due to differences in diet between the two groups (39). For example, relative to those fed a low-fat diet, mice fed a high-fat diet were found to have an increase in Enterobacteriaceae (40), a Gram-negative bacteria, and therefore increased expression of proinflammation cytokines and induction of TLR4 (40). Last, we found that TLR4, TLR5, and pIKKα/β expression tended to be higher among individuals with high cell proliferation in the colon as indicated by the baseline expression of MIB-1/Ki-67. This observation is consistent with our initial hypothesis that increased cell proliferation would be associated with a proinflammatory state in the colon.

Sessile serrated colorectal adenomas have a higher prevalence of BRAF mutations than do traditional colorectal adenomas (41). These mutations induce the MAPK pathway that can cause decreased apoptosis and increased proliferation (42). MAPK activation is also regulated downstream by the TLR pathway, which, for TLR5, is activated by binding with flagellin. Flagellin was found to activate several antiapoptotic mediator genes such as cIAP-1, cIAP-2, and A20 (43). We found that TLR5 expression was statistically significantly 95% higher in the crypts of patients with previously removed serrated adenomas relative to patients with no history of serrated adenomas.

We also found that consuming greater quantities of red and processed meat was associated with higher pIKKα/β expression, but not with TLR4 or TLR5 expression, in crypts or stroma. Possible explanations for this association may include the effects of heterocyclic amines present in the meat on activation of pIκBα and NF-κB signaling (44), as well as upregulation of TNFα production by heme (45, 46). Thus, the adverse effects of red and processed meat consumption on colorectal carcinogenesis may be mediated at least in part through the activation of the NF-κB signaling pathway.

This study had several limitations and strengths. The strengths include high protocol adherence by study participants, the automated immunohistochemistry and novel image analysis software to quantify crypt and between-crypt stroma biomarker distributions, and the consequent high biomarker scoring reliability. This study is the first randomized, double-blind, placebo-controlled clinical trial to report on the separate and combined effects of supplemental calcium and vitamin D₃ on pIKKα/β, TLR4, and TLR5 expression in the normal colorectal epithelium. The primary limitation of this study is the small sample size, which limited our power to detect small differences in biomarker expression and to conduct subgroup analyses (e.g., small number of current smokers). We measured biomarkers only in the rectal mucosa; therefore, treatment effects in other parts of the colon are unknown. Race and ethnicity was found to be an important contributing factor for the prevalence and location of large polyps and tumors in average-risk individuals (47). Most of our study participants were white, limiting our ability to detect differences in biomarker expression by race. In addition, we procured colorectal tissue samples from patients at baseline and after 1 year of treatment, so we could not investigate possible longer-term effects of calcium and vitamin D supplementation on the biomarkers. It is also unknown whether calcium and vitamin D affect human normal colon, precancerous, and cancerous tissues differently.

In conclusion, the results of this pilot adjunct biomarker trial provide no strong evidence that supplemental vitamin D, alone or in combination with calcium, have substantial effects on the expression of biomarkers related to NF-κB and TLR signaling pathways in the normal-appearing colorectal mucosa of colorectal adenoma patients. On the other hand, our findings of estimated differences in biomarker expression levels by various participant characteristics support continued investigation of potential modifiable factors that could affect inflammation pathways relevant to colorectal carcinogenesis.

No potential conflicts of interest were disclosed.

Conception and design: S. Ray, R.M. Bostick, R. Yacoub, V. Fedirko

Development of methodology: S. Ray, R.M. Bostick, R. Yacoub, V. Fedirko

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): R. Hodge, H.B. Mandle, S. Ray, A. Henry, R.M. Bostick, J.A. Baron, R.E. Rutherford, M.E. Seabrook, V. Fedirko

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): R. Hodge, H.B. Mandle, S. Ray, S. Tandon, F.A. Jahan, J.A. Baron, V. Fedirko

Writing, review, and/or revision of the manuscript: R. Hodge, H.B. Mandle, S. Ray, F.A. Jahan, R.M. Bostick, J.A. Baron, E.L. Barry, R. Yacoub, V. Fedirko

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M. Peterson, A. Henry, V. Fedirko

Study supervision: J.A. Baron, V. Fedirko

This study was funded by NCI, NIH R21 CA182752 to V. Fedirko, R03 CA184578 to V. Fedirko, R01 CA114456 to R.M. Bostick, and R01 CA098286 to J.A. Baron; Georgia Cancer Coalition Distinguished Scholar award to R.M. Bostick; the Franklin Foundation to R.M. Bostick. Pfizer Consumer Healthcare provided the study agents. The NCI, the Franklin Foundation, and Pfizer Consumer Healthcare had no influence on the design of this study; the collection, analysis, and interpretation of the data; the decision to submit the article for publication; or the writing of the article. We thank all study participants for their time and dedication to the study.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.