INTRODUCTION

The cationic diamine putrescine and its triamine derivatives spermidine and spermine play an essential role in tumor growth, thereby serving as a rationale for targeting the inhibition of intracellular polyamine biosynthesis for the treatment and prevention of cancer (1–4). Relative to adjacent normal tissue and tissues undergoing normal rates of turnover, these polyamines are present at high concentrations in tumor cells (5–7). Pharmacologic targeting of intracellular polyamines via the inhibition of ornithine decarboxylase, which is the rate-limiting enzyme that converts ornithine to putrescine, has been extensively investigated as a drug-based strategy to suppress tumor growth (1). Although of limited efficacy for advanced cancers (8), an irreversible inhibitor of ornithine decarboxylase (oral difluoromethylornithine) has demonstrated growth supressive activities, on colorectal adenomatous polyps [i.e., premalignant precursor of colorectal cancer (CRC)] in animal (9) and human (10) studies. Thus, pharmacological studies have established evidence of a growth-promoting role of intracellular polyamines in early colorectal tumorigenesis and as a target for chemoprevention in high-risk populations. Polyamines, from de novo synthesis, the diet, and products of gut microbiome metabolism, are made available to cells via active transport (11, 12). These extracellular sources of polyamines are essential for normal mammalian embryogenesis (13) and play a key role in the normal development and maintenance of the gastrointestinal tract (14). Polyamines are essential to normal cellular homeostasis (15). 

Inhibition studies showed that polyamines have important roles in the macromolecular synthesis of proteins and nucleic acids, regulation of gene and protein expression, and regulation of ion channels. More recent system-wide studies identified polyamines as immunomodulatory molecules and as important regulators in adult stem cell populations (16). The ability of probiotics to extend the life span in mice depends on bacterial-derived putrescine to inhibit colonic cell senescence (17). These findings are consistent with a progrowth effect on healthy colonic cells. Despite the recognized role of polyamines in both normal cellular homeostasis and intestinal tumor growth, there have been no studies, to our knowledge, that tested the association between dietary polyamines and human CRC. The development of American (18) and Swedish (19, 20) dietary polyamine databases has facilitated the estimation of polyamine intake from foods for application to population studies. With these instruments, daily intake of polyamines (primarily from produce, fermented, and aged foods and beverages) has been estimated at w300 mmol in the United States (4, 18). With the use of the American database, we reported that dietary polyamine intake above the population median (.289 mmol/d) was associated with a significant 30% increase in odds of developing colorectal adenomatous polyps in patients with a history of forming colorectal adenomas (i.e., high-risk patient population for CRC) (4). Similarly, Raj et al. (21) reported that high polyamine intake reduced the efficacy of difluoromethylornithine when combined with sulindac for the prevention of colorectal adenomas in highrisk polyp formers enrolled in the difluoromethylornithine and sulindac trial previously mentioned (10). 

Despite these promising observations in patients at above-average risk of developing CRC, to our knowledge, no studies have investigated whether dietary polyamine intake influences carcinogenesis in a general population or in those with average risk of CRC. Therefore, we sought to test the hypothesis that dietary polyamines increase risk of CRC and CRC-specific mortality in an average-risk cohort. In addition, because dietary polyamines may act as tumor promoters only in a high-risk background (4), we explored whether risk factors that increase the likelihood of having initiated colorectal epithelial cells (e.g., family history of CRC and older age) modify the effect of dietary polyamines on CRC. With the use of data from the WHI (Women’s Health Initiative), which is a longitudinal cohort of postmenopausal women with dietary data, adjudicated outcomes, and extended follow-up, we provide the first study to our knowledge of dietary polyamine exposure and CRC risk in average-risk women. 

 Thus, risk of CRC death was analyzed only in participants who had invasive, Surveillance, Epidemiology, and End Results-staged (24) CRC (local, regional, or distant but not in situ) by using a competing risk model to calculate a sub-HR, 95% CI, and P value as outlined by Fine and Gray (25) and adjusted for the stage at diagnosis and all model 3 covariates. Statistical models were adjusted for potential confounders of the polyamine-CRC risk association or strong predictors of CRC on the basis of previous literature. Model 1 was adjusted only for age (continuous) (4, 26). Model 2 was adjusted as for model 1 and for total estimated dietary energy, household income (,$10,000– $19,999, $20,0000–$49,999, $50,000–$99,999, or .$100,000), smoking pack-years (continuous) (27, 28), recreational physical activity (continuous metabolic equivalent task hour per week) (29), race-ethnicity (non-Hispanic white, black or African American, Hispanic or Latino, or other) (30), hormone-replacement therapy use (ever or never) (31), history of colonoscopy or sigmoidoscopy (yes or no) (32), and BMI (in kg/m2 ; continuous) (33). Model 3 was adjusted as for model 2 and for continuous dietary exposures including total fiber and calcium because they are present in foods with a high polyamine content (4, 18, 19) and are related to CRC risk (34, 35). 

Arginine (36–38), dietary fat (37, 39), and folate equivalents (40) were also added to model 3 because of their putative effects on polyamine metabolism, presence in high-polyamine foods (4, 18–20), and association with CRC risk (37, 41–44). The association between CRC and dietary putrescine alone and, separately, spermidine plus spermine was evaluated under model 3 conditions. Schoenfeld’s residuals were tested, and there was no evidence that proportional hazards assumptions were violated. Likelihood ratio tests were conducted to analyze the overall effect of adding dietary polyamines (quintiles) to each Cox model. Potential effect modifiers of the dietary polyamine–CRC association were tested by using a likelihood ratio test that compared model 3 to a model that included an interaction term. The following potential modifiers were evaluated because they are indicative of an initiated colorectal epithelium where polyamines may have a different effect than in a noninitiated colorectum (4, 13, 21, 45, 46): personal history of colorectal polyp removed, family history of CRC, older age, overweight or obese status, not using aspirin, and low fiber intake. In addition, arginine (21, 36, 47) and folate and folic acid (4, 40, 48) are involved in endogenous polyamine metabolism and, thus, have the potential to modify exogenous polyamine uptake and use. Further, the gut microbiome is responsive to BMI (49, 50) and fiber (51) and is believed to produce and use polyamines (17, 52, 53). Fiber, arginine, folate, and folic acid amounts were categorized into above (i.e., high intake) compared with below (i.e., low intake) median intakes to test for an effect modification.

 In ecological studies, the Mediterranean diet, which includes high citrus fruit intake, is associated with protection against the development of diseases including CRC (55). In experimental animal models, highpolyamine diets delayed aging (56), and putrescine was shown to mediate the effect of probiotics on longevity in mice through direct inhibitory effects on colonic cell senescence (17, 53). In contrast, low-polyamine diets were shown to delay wound healing (57) and inhibit growth (13). In humans at high risk of CRC (adenomatous polyp formers), we previously showed that dietary polyamines increased risk of adenomatous polyp recurrence (4). Similarly, mice initiated with a chemical colorectal carcinogen showed an increase in the severity of premalignant colorectal aberrant crypt foci and adenomatous polyps when exposed to a high-polyamine diet (45, 46, 56). However, these carcinogen-treated animals had fewer aberrant crypt foci in the intestinal tract when they kept receiving low-polyamine diets (13, 46), and a high-polyamine diet before carcinogen treatment decreases tumor incidence (56). Overall, our observations are consistent with the notion suggested by the literature that the effect of dietary polyamines is to promote health in low-risk colorectums, and polyamines may promote carcinogenesis in colorectums that experienced an initiating event.

 Although the multiple subgroup analyses reported should be interpreted cautiously, our findings suggest that dietary polyamines may be particularly beneficial in lowering CRC risk in women with normal BMI and in those who consumed higher than average fiber in this population (.15.63 g/d). Fiber (58) was not protective against CRC in the WHI cohort overall, but fiber intake has been shown to reduce risk of CRC in previous studies (35). Obesity has been consistently linked with CRC risk, more so in men (33), and the effect of polyamines could be dependent on body size (49, 50). Thus, these observations also support a differential effect of dietary polyamines dependent on background CRC risk. Although the exact mechanisms by which BMI or fiber act with polyamine intake in a protective fashion are unknown, we speculate that one mechanism may be through effects on the gut microbiome. Specifically, BMI (49, 50), fiber (51), and polyamines (17, 52, 53) all interact with the gut flora, which is a major secondary source of polyamine exposure to the colonic epithelium (11, 12). This plausible mechanism of action, and others, warrants exploration in both high- and low-risk CRC model systems to better understand by what mechanism exogenous polyamines affect gut function and tumorigenesis. In contrast with our previous findings for adenoma outcomes (4), we did not show evidence that women with a history of colorectal polyp removal had increased risk of CRC from dietary polyamine exposure. However, the WHI cohort did not differentiate between histories of colorectal polyps compared with adenomas. 

In addition, we failed to identify significant associations between polyamines and CRC risk in participants with some highrisk characteristics (e.g., family history of CRC or advanced age). These results may reflect differences related to past screening behaviors in these groups for which we lacked sufficient detail before enrollment in WHI. We could not discount the possibility that our findings were biased because of residual confounding because women of lower socioeconomic status are more often in the lowest quintile of dietary polyamine consumption. The potential for misclassification of a dietary exposure, the presence of correlated nutrients, and the lack of more-precise biomarkers of polyamine intake were important limitations in this and other similar studies of dietary factors (18). The polyamine content depends on the ripeness, bacterial activity, and age of foods (4, 18, 59), which makes estimation challenging. For example, the considerable overlap of polyamine-, fiber- and folate-rich foods may preclude a conclusive distinction of individual effects of any of these dietary exposures in the observational setting. This possibility raises the prospect that previous findings of associations between dietary folate and fiber with CRC may have been confounded by the biological effects of polyamines or that our study suffered from residual confounding. Nonetheless, the large sample size, quantity of measured variables, adjudicated CRC outcomes, and long follow-up period in this study were considerable strengths.

 In conclusion and in contrast to our original hypothesis, we showed no evidence for a positive association between dietary polyamines and CRC risk. Instead, this study provides the first evidence, to our knowledge, that diets high in polyamines may decrease risk of CRC in average-risk postmenopausal women, particularly in those with other CRC risk–lowering habits such as consumption of higher fiber diets and maintenance of a normal body weight. This observation is consistent with emerging evidence of beneficial effects of polyamines on longevity and gastrointestinal health and argues for differential effects by risk status, thereby warranting additional study on the potential health benefits of polyamine-rich foods. The authors’ responsibilities were as follows—AJV, RBW, MLN, CAT, and PAT: designed the research; AJV, ELA, and BCW: conducted the statistical research and analyses; RBW, MLN, and CAT: provided access to WHI data sets; AJV, ELA, CAT, and PAT: wrote the manuscript; and AJV: had primary responsibility for the final content of the manuscript. None of the authors reported a conflict of interest related to the study