Serum Salicylate Levels and Risk of Recurrent Colorectal Adenomas

Aspirin use has been shown to reduce the risk of colorectal neoplasia in randomized controlled trials and cohort studies (1–6), presumably through inhibition of COX-1 and COX-2 (7, 8) but also through other non-COX mechanisms (9–11). Salicylic acid (SA), the deacetylated metabolite of aspirin (acetylsalicylic acid; ASA), is readily detected in serum and urine of humans and is a measure of salicylates, whether from ASA use or from diet and endogenous sources. Salicylate levels are also measurable in individuals not taking aspirin or salicylate medications (9, 11–14). Diet is known to be a source of salicylates, including fruits, vegetables, and meat. The salicylate content of blood and urine was shown to increase following consumption of a variety of meals, both vegetarian and meat-based, ranging from southern Indian dishes to European ones (15, 16).

In plants, SA is synthesized through metabolism of benzoic acid. A synthetic form of benzoic acid, sodium benzoate, is a widely used a food preservative, and it also contributes to SA levels after ingestion. In individuals not taking aspirin, it is estimated that 20% of the variability in serum SA maybe derived from dietary sources (17). Understanding various endogenous sources of salicylates is an ongoing area of research (18, 19). These include certain colonic bacteria species that may have the ability to synthesize SA, and synthesis of SA via metabolism of hippuric acid, a metabolite of benzoic acid demonstrated to be formed endogenously in man (18) The associations of serum salicylate levels, in particular among nonusers of aspirin, in reducing risk of colorectal adenoma are not known.

The aim of our study was to determine whether serum salicylate levels among individuals with history of prior adenoma are associated with risk of colorectal adenoma in individuals not taking aspirin.

We performed a secondary observational analysis on data from Calcium Polyp Prevention Study, a randomized controlled clinical trial of supplemental calcium for the prevention of colorectal adenomas. The design and findings of the trial have been previously described (20, 21). In the trial, 930 subjects from 6 clinical centers, all with at least 1 histologically confirmed colorectal adenoma and no known polyps remaining in the large bowel, were randomized to receive placebo or 1,200 mg elemental calcium as 3 grams of calcium carbonate daily for 4 years. Recruitment for the trial began in 1988 and randomized treatment ended in 1996. All subjects completed several baseline questionnaires, including a food frequency questionnaire. Intakes of aspirin or other NSAIDs were not restricted. Although reports of aspirin (ASA) use were collected, the specific dose and frequency was not recorded. About 1 year after randomization, all subjects underwent blood draws and serum salicylic acid levels were measured as described in the following text. All subjects underwent colonoscopy at year 1 and 4 after randomization. While on treatment, questionnaires were sent to participants every 6 months regarding adherence to study treatment and use of medications, over the counter drugs, and nutritional supplements.

Serum SA measurements were performed on stored serum samples collected at the year 1 blood draw using a highly sensitive HPLC method described previously (9, 22). All measurements were performed blindly at the same laboratory facility.

ASA use during treatment was summarized by percentage of interval questionnaires reporting use. Individuals were considered non-users if none of the questionnaires completed before/up to the time of blood draw at year 1 reported ASA use and there was no reported use, or reported use in 25% or fewer of questionnaires, between year 1 and year 4. Subjects who reported ASA in more than 25% of questionnaires after the blood draw were considered ASA users.

Of the 930 randomized subjects, 832 completed follow-up colonoscopies between year 1 and 4 of the trial. Eight hundred and twenty-five subjects provided information on ASA use in periodic questionnaires. Adequate samples for SA measurements were available for 691 of these subjects, of which 459 were non-ASA users and were included in the analysis. To assess selection bias, we compared individuals with and without specimens available with regard to differences in baseline or outcome variables. All analysis was carried out using STATA version 11.0.

The primary endpoint of the study was adenoma recurrence after the year-1 examination up to and including the year 4 exam. We also evaluated occurrence of advanced lesions, defined as any lesion with the following features: tubulovillous components (25%–74% villous), villous adenoma (≥75% villous), advanced dysplasia or invasive cancer, and large adenomas (≥1 cm in diameter).

We evaluated the association of serum SA with risk of recurrent adenoma and advanced lesions among nonusers of ASA during the main risk period and also conducted analyses stratified by calcium treatment assignment. We evaluated the interaction between calcium intake and serum SA level on risk of recurrent adenomas and advanced lesions using standardized serum SA values as a continuous variable [SA/sd(SA)], as a 4-level variable based on quartiles, and as a dichotomous variable above versus below the median. We used generalized linear models for binomial distribution with a log link function to compute risk ratios, adjusted for potential confounders. Variables considered in the model included age, sex, center, follow-up time, and treatment group. Interactions were assessed using product interaction terms between treatment assignment and serum SA levels at year 1. Wald's tests were used to assess the statistical significance of association of serum SA and adenoma risk

Baseline characteristics of the trial participants included in the analysis are summarized in Table 1. Treatment groups were similar in demographic characteristics.

Of the individuals with adequate samples for SA measurement, 459 (61%) were nonusers of aspirin. Overall, the median SA level at that time was 1.05 μmol/L (IQR 0.21–2.21 μmol/L).

In generalized regression models, among nonusers of ASA, there was no association between serum SA and risk of adenoma or advanced adenoma over the next 3 years in crude or adjusted models (adjusted RR per standard deviation = 1.01 (95% CI 0.97–1.04) and 1.05 (95% CI 1.00–1.10) for all adenomas and advanced adenomas. Results were unchanged when serum SA was entered in the model as a 4-level variable based on quartiles (Table 2), However, the point estimates associated with quartile 2 and higher suggest that there may be a threshold protective effect. There was no interaction between calcium supplementation and serum SA levels on risk of adenoma or advanced adenoma (P for interaction 0.60 and 0.58 for all adenomas and advanced adenomas respectively).

Our study is the first effort to evaluate the relationship between serum salicylate levels and risk of recurrent adenoma. In this regard we had several notable findings. First, we are able to confirm that salicylic acid levels are detectable in human serum, in individuals with little or no reported regular intake of ASA-containing compounds. This is consistent with growing literature on dietary and endogenous sources of SA (9, 12, 22).

Benzoic acid is a natural constituent of plants and found in large amounts in fruits such as berries. (17) Patterson and colleagues have reported the conversion of benzoic acid to SA in humans (12). One of the main metabolites of benzoic acid, hippuric acid, which is metabolized to salicylate, may be also formed endogenously in man (18). Several factors likely contribute to serum SA levels, such as differences in dietary intakes of fruits, vegetables, benzoic acid, and red meats, variation in endogenous cellular production through metabolism of hippuric acid, or from de novo synthesis by colonic bacteria. There is evidence that the SA content of certain food dishes, such as vindaloo curry is comparable to low dose (75 mg) aspirin (11). Other factors such as genetic factors, polymorphisms affecting enzyme activities, and the interaction of genetic, dietary and environmental factors are currently being investigated. Recent evidence suggests that chronic consumption of low dose aspirin (75 mg) reduces long-term colorectal cancer incidence and mortality (23). Whether comparable chemoprotection can be achieved from SA derived from non-ASA sources is an important question.

We found no association between serum SA and risk of adenomas or advanced adenomas on colonoscopy performed 3 years later (or in a few instances in the interval between), and no modification by SA level of the effect of calcium in reducing risk of all recurrent adenomas or advanced adenomas among nonusers of ASA. The findings of a possible threshold effect when quartiles of serum SA were used in the model should be interpreted with caution, particularly because we did not see similar trend for advanced adenoma. We chose not to specifically test for the threshold effect, because post hoc nature of the test would invalidate it.

There is ample literature on ASA intake and reduction in risk of adenomas and colorectal cancer from observational studies (5, 24) and ASA supplementation as a chemopreventive agent for adenomas and advanced adenomas from randomized clinical trials (2, 6). However, in the absence of ASA use, our results suggest that serum SA levels, derived from dietary and endogenous sources may be insufficient to confer protection from risk of subsequent adenomas, or to demonstrate an interaction with calcium supplementation.

Our study has several limitations. There is a possibility of type II error, as the the risk estimates do suggest a threshold effect. Serum SA levels were not available in all 930 subjects, and the 691 subjects included in this analysis, although they did not differ in baseline or outcomes variables of interest, they may differ in other characteristics that could affect adenoma risk. We had no information regarding aspirin use specifically in the day before the blood draw, and it is possible some subjects thought to be off aspirin had actually taken the drug. We measured serum SA onetime only, although the literature supports that the measurements are reproducible (12). We measured SA in serum only, whereas the literature reports measurement of urinary salicylate levels. Urine salicylate levels may be a more useful measure of the body's SA exposure than serum SA because they reflect use over a longer time period. We did not have information on the timing of the blood draws in relation to the participants' last meals, which may influence serum SA levels. Strengths of our study include large sample size, randomization of calcium supplementation, and separate analysis of advanced adenomas.

Our data suggest a lack of an association of risk of adenoma occurrence with serum SA among nonusers of ASA. Further research would be desirable to confirm this, ideally with prospective, long-term designs and careful correlation of serum levels with proximate measurement of ASA intake.

No potential conflicts of interest were disclosed.

Supported in part by grants: NCI (CA37287 and CA23108) And VA Minneapolis Career Development Award (CDA-2; A. Shaukat)

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.