Observational studies suggest that nonsteroidal anti-inflammatory drugs (NSAIDs) reduce the risk of esophageal adenocarcinoma, but it is not known at what stage they may act in the esophageal inflammation-metaplasia-adenocarcinoma sequence. In an all-Ireland case-control study, we investigated the relationship between the use of NSAIDs and risk of reflux esophagitis, Barrett's esophagus, and esophageal adenocarcinoma. Patients with esophageal adenocarcinoma, long-segment Barrett's esophagus and population controls were recruited from throughout Ireland. Esophagitis patients were recruited from Northern Ireland only. Data were collected on known and potential risk factors for esophageal adenocarcinoma and on the use of NSAIDs, including aspirin, at least 1 year before interview. Associations between use of NSAIDs and the stages of the esophageal inflammation-metaplasia-adenocarcinoma sequence were estimated by multiple logistic regression. In total, 230 reflux esophagitis, 224 Barrett's esophagus, and 227 esophageal adenocarcinoma and 260 population controls were recruited. Use of aspirin and NSAIDs was associated with a reduced risk of Barrett's esophagus [odds ratio [OR; 95% confidence interval (95% CI)], 0.53 (0.31-0.90) and 0.40 (0.19-0.81), respectively] and esophageal adenocarcinoma [OR (95% CI), 0.57 (0.36-0.93) and 0.58 (0.31-1.08), respectively]. Barrett's esophagus and esophageal adenocarcinoma patients were less likely than controls to have used NSAIDs. Selection or recall bias may explain these results and the results of previous observational studies indicating a protective effect of NSAIDs against esophageal adenocarcinoma. If NSAIDs have a true protective effect on the esophageal inflammation-metaplasia-adenocarcinoma sequence, they may act early in the sequence. (Cancer Res 2006; 66(9): 4975-82)
- esophageal adenocarcinoma
- Barrett's esophagus
- reflux esophagitis
- nonsteroidal anti-inflammatory drugs
- gastrointestinal cancers: other
- descriptive, risk factor, and methodologic studies
Gastroesophageal reflux disease is a commonly reported medical condition caused by reflux of acid and/or bile into the distal esophagus. In some people, chronic gastroesophageal reflux leads to Barrett's esophagus, a condition in which the normal stratified squamous epithelium lining the distal esophagus is replaced by columnar epithelium of intestinal type. Compared with the general population, Barrett's esophagus patients are 17 to 125 times more likely to develop esophageal adenocarcinoma ( 1– 5), a cancer that has been increasing rapidly in many western societies over recent decades ( 6– 10).
Evidence is emerging that nonsteroidal anti-inflammatory drugs (NSAIDs), including aspirin, may prevent several cancers ( 11). NSAIDs inhibit cyclooxygenase (COX) enzymes COX-1 and COX-2. COX-1 is constitutively expressed in many human tissues, whereas COX-2 can be induced in response to cytokines, growth factors, and mitogens. Inhibition of COX-2 restores apoptosis ( 12, 13), decreases cell growth ( 13), and/or decreases proliferation ( 14) and inhibits angiogenesis ( 15). Most research to date has focused on colorectal cancer, with studies reporting that both aspirin and non-aspirin NSAIDs may be protective ( 16– 19). Studies have also indicated that NSAIDs may prevent the development of premalignant colorectal adenomas ( 20, 21). Several studies have reported reduced risks of all esophageal carcinomas in NSAID users ( 11, 22– 27) and NSAIDs, including aspirin, have been associated with a reduced risk of esophageal adenocarcinoma ( 23, 28, 29). One study has indicated that esophageal adenocarcinoma patients are less likely to use non-aspirin NSAIDs compared with Barrett's esophagus patients ( 30), and a recent study reported that Barrett's esophagus patients who used NSAIDs had a lower incidence of esophageal adenocarcinoma ( 27), but, as yet, no research has investigated whether NSAIDs protect against Barrett's esophagus or specifically examined at what stage they may act in the esophageal inflammation-metaplasia-adenocarcinoma sequence.
Using an all-Ireland case-control study, the Factors Influencing the Barrett's Adenocarcinoma Relationship (FINBAR) study, we set out to investigate the relationship between the use of NSAIDs and reflux esophagitis, Barrett's esophagus, and esophageal adenocarcinoma to shed light on the stage at which NSAIDs may exert a protective effect. In doing so, we hoped to determine at which stage chemoprevention with NSAIDs might be most effective.
Materials and Methods
The FINBAR study commenced in Ireland in March 2002 and continued until December 2004. The study included three groups of subjects: (a) patients with esophageal adenocarcinoma, (b) patients with long-segment Barrett's esophagus, and (c) normal population controls, all recruited from both Northern Ireland and Republic of Ireland. From September 2004 to July 2005, a group of reflux esophagitis patients were recruited from Northern Ireland only.
Esophageal adenocarcinoma cases (aged ≤85 years) were patients with a histologic confirmation of adenocarcinoma within the esophagus. In situ cancers were not included. Cases from Northern Ireland were identified from electronic pathology records from all pathology laboratories within the province. Republic of Ireland cases were identified from the main hospitals involved in the diagnosis and treatment of esophageal cancer, including St. James's Hospital (Dublin, Ireland), which is a national referral center for esophageal cancer. All available relevant clinical and histologic records, including surgical and radiological reports, were reviewed by L.A.A., S.J.M., J.M., and a pathologist to confirm that the tumor was located in the esophagus and to assign tumors into two groups: (a) esophageal tumors (including tumors encroaching on the esophagogastric junction) and (b) junctional tumors (tumors involving the esophagus, esophagogastric junction, and gastric cardia). There was insufficient evidence available to classify the topography of four tumors; these were included in the overall analyses but excluded from the subgroup analyses.
Barrett's esophagus patients were eligible for inclusion if ≥3 cm of typical Barrett's mucosa were seen at endoscopy, and the presence of specialized intestinal metaplasia was confirmed by histologic examination of biopsy specimens. Incident and prevalent cases were included. Patients with dysplasia on histologic examination were not included. In Northern Ireland, cases of Barrett's esophagus were initially identified from pathology reports gathered from throughout Northern Ireland. Endoscopy note review was necessary in most patients to confirm the length of the segment of Barrett's esophagus because length was infrequently recorded on the pathology report. In the Republic of Ireland, clinicians in the Dublin and Cork areas sent details of Barrett's esophagus patients who met the inclusion criteria to the research personnel.
Reflux esophagitis patients were patients diagnosed in Northern Ireland with macroscopically visible erosive esophagitis at upper gastrointestinal endoscopy. Erosive esophagitis was defined as evidence of mucosal breaks or erosions within the esophagus (grades 2-4 in the Savary Miller/Hetzel-Dent classification or grades B, C, or D in the Los Angeles classification were included). Patients were excluded if they had Barrett's esophagus visible endoscopically or at biopsy, if they had macroscopic or histologic evidence of infection, such as candidiasis, if they had documented dysmotility, or if there was gastric outlet obstruction.
Eligible control subjects were adults without a history of esophageal or other gastrointestinal cancer or a known diagnosis of Barrett's esophagus. Northern Ireland controls were frequency matched by sex and 5-year age band to the distribution of esophageal adenocarcinoma patients and selected at random from the General Practice Master Index (a province-wide database of all persons registered with a general practitioner). Republic of Ireland controls were frequency matched (within 5-year age and sex strata) to the distribution of esophageal adenocarcinoma patients. They were selected at random from four general practices (two urban and two rural) in the Dublin and Cork areas chosen by the researchers to reflect the urban/rural distribution of esophageal adenocarcinoma patients in the Republic of Ireland.
All subjects underwent a structured computerized interview with trained interviewers. Information was collected on NSAID, aspirin, and paracetamol use, symptoms of gastroesophageal reflux, history of hiatus hernia, peptic ulcers, esophagitis, and Barrett's esophagus, weight, height, smoking history, education, occupational type (manual/nonmanual), and alcohol consumption. Anthropometric measures (height, weight, waist, and hip circumference) were taken at the time of interview. Frequent gastroesophageal reflux was defined as symptoms of heartburn and/or acid reflux occurring at least once weekly (>50 times per year) >5 years before interview. Subjects were classified as having gastroesophageal reflux if they responded positively to either of the following questions: (a) Have you ever had frequent heartburn (i.e., a burning or ache behind the breastbone that is not due to heart problems) not including the last 5 years? or (b) Have you ever had frequent acid reflux (i.e., a bitter taste of stomach acid, which has come up to the back of your throat) not including the last 5 years? Body mass index (BMI) 5 years before the interview date was calculated by dividing self-reported weight in kilograms by current height in meters squared. Current smoking status was defined as smokers who had smoked at least one cigarette daily for ≥6 months 5 years before interview date. Previous smokers were those smokers who had quit smoking >5 years before interview date. People who had never smoked, who had smoked <100 cigarettes in their lifetime, or who had smoked <1 cigarette daily for ≥6 months were defined as never smokers. Alcohol consumption was measured using the European Prospective Investigation of Cancer food frequency questionnaire used in the Norfolk area of England ( 31). This validated questionnaire was modified for the Irish population by including foods and beverages commonly consumed in Ireland as identified in the North/South Food Consumption Survey ( 32). Alcohol consumption (including stout beer, lager, wine, and spirits) 5 years before interview date was defined as the number of grams of alcohol consumed daily calculated by summing the grams of alcohol contained on average in a glass of wine, one pint of beer, or one measure of spirits.
Subjects were asked to recall if they had ever taken any prescription or over-the-counter aspirin or non-aspirin NSAID at least once weekly for ≥6 months (Appendix A). Cards with generic and brand names were provided for the subjects to aid memory recall. Subjects responding affirmatively were asked to provide their age or the date when they first started taking these medications regularly, when they stopped taking the medications on a regular basis, and the total duration of use and the frequency that the medications were taken. Medication usage is reported ≥1 year before interview date. As patients may have ceased using these medications before the diagnosis because of symptoms, a lagged reference date of ≥5 years is also reported. Duration of use reports the total length of time that individuals took the medications before the reference dates of 1 or 5 years.
Statistical methods. Statistical analyses were done using STATA 8.0. In bivariable analyses, χ2 tests were used to examine differences in categorical variables between groups, and t tests were used to compare continuous variables. Unconditional maximum-likelihood multinomial (polytomous) logistic regression analyses were undertaken to examine the associations between the use of aspirin and non-aspirin NSAIDs ≥1 or 5 years before interview and case/control status adjusting for potential confounders. In this way, the odds of reflux esophagitis, Barrett's esophagus, or esophageal adenocarcinoma patients having used NSAIDs/aspirin were compared with the odds of control subjects having these exposures. Odds ratios (OR) with 95% confidence intervals (95% CI) were calculated. Reflux esophagitis patients, who were only recruited from Northern Ireland, were compared with Northern Ireland controls, as there were prescribing differences between Northern Ireland and the Republic of Ireland. Analyses are shown adjusted for potential confounders, including sex, age at interview, job type (manual/nonmanual work), education (years full-time), BMI (5 years before diagnosis/interview), smoking (current, ex, never), alcohol consumption (g/wk), and location (Northern Ireland, Republic of Ireland), except reflux esophagitis analyses. Further adjustment was also made for gastroesophageal reflux symptoms (never, ever) and upper gastrointestinal tract (GIT) disorders, including hiatus hernia, peptic ulcers, esophagitis, and Barrett's esophagus, in the esophageal adenocarcinoma analyses. Analyses were done for non-aspirin NSAIDs and aspirin and also stratified by reporting of symptoms of gastroesophageal reflux 5 years before interview date. Differences between esophageal and junctional subgroups were explored.
Ethical committee approval for the FINBAR study was obtained from the Research Ethics Committee of the Queen's University (Belfast, Northern Ireland), Clinical Research Ethics Committee of the Cork Teaching Hospitals, and Research Ethics Committee Board of St. James's Hospital. Ethical approval was separately obtained from Queen's University for the recruitment of reflux esophagitis patients.
A total of 227 esophageal adenocarcinoma patients (131 esophageal, 92 junctional, and 4 unclassified), 224 Barrett's esophagus patients, and 260 population controls were recruited as part of the FINBAR study. Also included were 230 reflux esophagitis patients from an esophagitis study carried out in Northern Ireland. The characteristics of each group of subjects are presented in Table 1 .
Participation rates. Overall, 355 esophageal adenocarcinoma patients were suitable for inclusion. Of these, 49 patients died before contact was made, 42 patients were not contacted, as their clinician, general practitioner, or relative did not give permission, and 37 patients themselves declined to participate. The participation rate of eligible, alive patients was 74.2%, and the overall response rate was 63.9%. The participation rates of Barrett's esophagus and reflux esophagitis patients and control subjects were 82.4%, 68.7%, and 41.8%, respectively.
NSAID use in the FINBAR study. Barrett's esophagus patients were substantially less likely than controls [fully adjusted OR (95% CI), 0.53 (0.31-0.90)] to have used aspirin at least once weekly for ≥6 months ≥1 year before interview date ( Table 2 ). When use of aspirin ≥5 years before interview was examined, the OR remained decreased but ceased to be statistically significant [OR (95% CI), 0.69 (0.38-1.26)]. It is noteworthy that adjusting for gastroesophageal reflux symptoms and upper GIT disorders strengthened rather than attenuated observed associations. An inverse association was also observed between aspirin use ≥1 year before interview and esophageal adenocarcinoma risk [fully adjusted OR (95% CI), 0.57 (0.36-0.93)]. This association was somewhat weaker for aspirin use ≥5 years before interview [OR (95% CI), 0.75 (0.44-1.27)]. As for Barrett's esophagus, the adjustment for gastroesophageal reflux symptoms and upper GIT disorders did not attenuate the strength of the observed associations.
The adjusted ORs for the use of non-aspirin NSAIDs ≥1 and ≥5 years before interview in Barrett's esophagus and esophageal adenocarcinoma patients ( Table 2) showed an even greater reduction than for use of aspirin, although, as fewer subjects were taking these drugs than aspirin, the 95% CIs were wider, and only use ≥1 year before interview in Barrett's esophagus patients was statistically significantly reduced [OR (95% CI), 0.40 (0.19-0.81)]. As for aspirin, the adjustment for potential confounders strengthened the associations.
The use of NSAIDs, including aspirin, was associated with a reduced risk of both esophageal and junctional tumors, but the association was slightly stronger in the esophageal tumor group ≥1 year before interview date [OR (95% CI), 0.47 (0.28-0.81) and 0.57 (0.32-1.01), respectively]. However, ≥5 years before interview date, there were no significant associations in the esophageal tumors [OR (95% CI), 0.92 (0.51-1.66)] and junctional tumors [OR (95% CI), 0.68 (0.34-1.35)].
The relationship between duration of NSAID use and risk of Barrett's esophagus and esophageal adenocarcinoma was examined by categorizing duration of use of all NSAIDs 1 year before interview into <5 or ≥5 years. Both use for <5 or ≥5 years were statistically significantly associated with reduced risk of Barrett's esophagus and esophageal adenocarcinoma ( Table 3 ). The ORs for <5- or ≥5-year use were of similar magnitude in Barrett's esophagus patients, with lower ORs for <5-year use in esophageal adenocarcinoma patients compared with ≥5-year duration.
Most subjects used NSAIDs on a daily basis for ≥6 months; daily use was reported by 86% of Barrett's esophagus patients, 87.9% of esophageal adenocarcinoma patients, and 90.8% of controls. Daily or more frequent use of all NSAIDs was statistically significantly associated with reduced risk of both Barrett's esophagus and esophageal adenocarcinoma ( Table 3). The ORs for less frequent use of NSAIDs in Barrett's esophagus and esophageal adenocarcinoma were of similar magnitude to those for more frequent use, but, because of small numbers, the 95% CIs were large and included unity.
Patients with symptoms of gastroesophageal reflux may avoid using NSAIDs because of symptom exacerbation. It is conceivable therefore that all or part of the inverse relationships seen between NSAID use and Barrett's esophagus and esophageal adenocarcinoma within the FINBAR study result from avoidance of these medications. However, within FINBAR, NSAID use was more common in controls with gastroesophageal reflux symptoms (22.3%) than those without symptoms (16.6%). In addition, when the analyses of the relationships between all NSAID use and Barrett's esophagus and esophageal adenocarcinoma were stratified by whether subjects reported gastroesophageal reflux symptoms, the observed ORs for NSAID use ≥1 year before interview in both Barrett's esophagus and esophageal adenocarcinoma did not differ between subjects with and without symptoms ( Table 4 ). The pattern was less obvious for NSAID use ≥5 years before interview because of smaller numbers. The associations remained in non–gastroesophageal reflux sufferers, suggesting that the inverse relationship with NSAID use may not be fully explained by avoidance of these drugs in symptomatic Barrett's esophagus and esophageal adenocarcinoma patients.
NSAID use in the esophagitis study. Because reflux esophagitis patients were only recruited in Northern Ireland, the analyses of the relationship between NSAID use and reflux esophagitis used only Northern Ireland population controls. The number of subjects taking non-aspirin NSAIDs was small, and the analyses were limited to all NSAIDs and aspirin only. In fully adjusted analyses, the ORs were reduced slightly in patients using the medications ≥1 year before interview [OR (95% CI), 0.75 (0.45-1.26)] and ≥5 years before interview [OR (95% CI), 0.93 (0.52-1.69)]. For aspirin use, the associations were as follows: (a) ≥1 year [OR (95% CI), 0.80 (0.46-1.39)] and (b) ≥5 years [OR (95% CI), 1.14 (0.62-2.09)].
Paracetamol use. There were no significant relationships between paracetamol use and Barrett's esophagus or esophageal adenocarcinoma compared with controls [OR (95% CI), 1.09 (0.63-1.91) and 0.82 (0.45-1.52), respectively].
In this all-Ireland case-control study, previous use of aspirin and non-aspirin NSAIDs was lower among Barrett's esophagus and esophageal adenocarcinoma patients than among age-matched and sex frequency-matched controls. Adjusting for gastroesophageal reflux symptoms and upper GIT tract disorders strengthened rather than attenuated the associations, which were similar in subjects reporting and not reporting a history of gastroesophageal reflux symptoms. Use of NSAIDs was less clearly associated with the risk of reflux esophagitis, although wide 95% CIs could not rule out an inverse relationship. Paracetamol use was not associated with a reduced risk of Barrett's esophagus or esophageal adenocarcinoma. These data are consistent with a protective effect for NSAID use against Barrett's esophagus and esophageal adenocarcinoma, but other explanations are possible.
A recent meta-analysis of studies published before 2001 investigating NSAID and/or aspirin use in esophageal cancer patients reported a 33% reduced risk of esophageal adenocarcinoma ( 23). In the large U.S. case-control study, Farrow et al. ( 29) reported that aspirin use was associated with a 63% reduced in risk of esophageal adenocarcinoma 1 year before the diagnosis but that former use was not associated with a reduced risk of esophageal adenocarcinoma. A similar pattern was observed for the use of non-aspirin NSAIDs, which reduced the risk of esophageal adenocarcinoma by 43%. There remained a strong protective relationship between NSAIDs and esophageal adenocarcinoma in patients who reported no history of gastroesophageal reflux symptoms, suggesting that the effect was not explained by avoidance of NSAIDs among gastroesophageal reflux sufferers.
In a study using the United Kingdom General Practice Research database, Lindblad et al. ( 22) reported that current users of non-aspirin NSAIDs had a nonsignificant 41% reduced risk of esophageal adenocarcinoma [OR (95% CI), 0.59 (0.29-1.22)]. Among aspirin users, there was a 25% reduced risk of esophageal adenocarcinoma; however, none of these analyses were statistically significant, and the authors suggested that upper GIT disorders may distort the associations. The authors indicate several limitations of their study, including difficulty with case ascertainment, possibility of tumor misclassification, inadequate information relating to potential confounding variables, and that over-the-counter NSAID drugs were not included.
A recent study, using data from the Seattle Barrett's esophagus program, reported a reduced incidence of esophageal adenocarcinoma in Barrett's esophagus patients currently using NSAIDs, suggesting that NSAIDs may help prevent the progression from Barrett's esophagus to esophageal adenocarcinoma ( 27). It was suggested that the effects of NSAIDs are short-term and do not last after discontinuation of NSAID use. The FINBAR study findings suggest that NSAIDs may act before the formation of Barrett's epithelium. The Seattle Barrett's esophagus project was not able to assess this possibility, as they did not have a normal control group for comparison.
The strengths of the FINBAR study include its population-based design, size, tumor classification by a clinical panel, inclusion of over-the-counter NSAIDs, inclusion of Barrett's esophagus and reflux esophagitis patients, and adjustment for appropriate confounders. It is important to note that the Barrett's esophagus and reflux esophagitis patients and controls were frequency matched to the age and sex distribution of the esophageal adenocarcinoma cases. Therefore, Barrett's esophagus and reflux esophagitis patients may not be representative of the entire population of patients with Barrett's esophagus and reflux esophagitis, respectively, although the comparisons with controls and esophageal adenocarcinoma patients should remain valid.
This is one of the few studies to report NSAID use in histologically verified esophageal adenocarcinoma patients ( 11, 29), as most studies to date report use in all esophageal cancers ( 11, 24– 26). Although every effort was made to exclude tumors originating in the gastric cardia, some may have been included in the analyses if they extended into the esophagus. We compared patients with esophageal tumors (including tumors encroaching on the esophagogastric junction) with those that have tumors straddling the esophagogastric junction (junctional tumors) and found that the ORs for overall NSAID use were reduced in both groups and significantly so in the esophageal tumor group.
A potential limitation of the study was that the dose of NSAIDs taken was not recorded. COX-2 is thought to be inhibited by high-dose aspirin only, and several colorectal cancer studies have reported that the protective effects conferred by aspirin are only apparent in subjects taking 300 mg aspirin/d ( 18, 21). Low-dose aspirin does not seem to be protective of colorectal cancer ( 18, 21, 33) but may moderately protect against recurrent adenomas ( 34). If some patients included in the FINBAR study were on lower doses of aspirin (e.g., 75 mg) as recommended for the prevention of cardiovascular disease, then the protective effect of higher doses may be underestimated. Occasional users who used NSAIDs less than once weekly for ≥6 months were included in the “never user category.” This may have led to some attenuation of the risk estimates, but it is unlikely that occasional users would benefit with regards to cancer prevention.
Recall bias is a possibility within this study, as data on NSAID use were collected by interview with study participants rather than from medical or pharmacy records. However, it is very unlikely that patients were aware of the possibility of a protective effect of NSAIDs against esophageal adenocarcinoma or Barrett's esophagus, and it was one of several medication-related hypotheses to be examined in the study, making it unlikely that the data were systematically gathered differently between patients and controls. Using the interview method also made it possible to collect information on over-the-counter NSAID use. Paracetamol use was not significantly associated with a reduced risk of Barrett's esophagus or esophageal adenocarcinoma, and this may indicate that the reduced risk is specific to NSAIDs.
Another potential limitation of this study was the low response rate among controls and the possibility of selection bias. People who are in regular contact with health services (e.g., with cardiovascular disease or arthritis) may be more likely to participate as controls in medical research, such as the FINBAR study, and may have higher use of aspirin or other NSAIDs. On the other hand, these patients may be less likely to participate as controls because of disease-related disability. Controls were similar to the general population with regards to symptoms of gastroesophageal reflux and BMI. In a study from Bristol, 15.6% of people ages 20 to 59 years had weekly symptoms of heartburn ( 35) compared with 18.4% of controls within this age range in the FINBAR study. The mean weight and the proportion who were obese among the controls were also similar to those seen in the all-Ireland Food Consumption Survey (1997-1999; ref. 36). Mean BMI in males aged 51 to 64 years in the survey was 27.6 kg/m2 (SD, 3.6) and 28 kg/m2 in FINBAR (SD, 4.5). Similarly, 20% of men were obese (BMI >30 kg/m2) in the Food Consumption Survey compared with 22% of FINBAR male controls. These data suggest that the controls reflect the population from which they were drawn. However, in the 2001 Northern Ireland Health and Social Well-being Survey, 6 23.6% of males ages ≥55 years were current smokers, 53.1% were ex-smokers, and 23.3% were never smokers compared with 18.9%, 48%, and 31.1% in FINBAR controls, respectively, indicating that never smokers may be overrepresented among FINBAR controls. There is no comparable population-based data of the prevalence of prescription and over-the-counter NSAID/aspirin use in Ireland. However, an inverse association between NSAIDs and esophageal adenocarcinoma has been reported in one other case-control study, which had a better response rate among controls than the FINBAR study, and was undertaken in a different health care setting ( 29), suggesting that selection bias is unlikely to fully explain our findings.
Apart from recall and selection biases, there are three possible explanations for the lower odds of exposure to NSAIDs in esophageal adenocarcinoma and Barrett's esophagus patients compared with controls. These are (a) avoidance of NSAIDs because these drugs aggravate symptoms of either gastroesophageal reflux or early cancer, (b) contraindication or perceived contraindication to the use of the drugs in Barrett's esophagus or esophageal adenocarcinoma patients with a previous history of upper GIT tract conditions, and (c) a true protective effect against Barrett's esophagus and esophageal adenocarcinoma.
Patients taking NSAIDs may experience more frequent reflux ( 35). If patients stopped using NSAIDs for this reason, this may explain some of the apparent protective effects observed. However, when we adjusted for gastroesophageal reflux symptoms, the associations were strengthened not attenuated. The associations also remained in patients who had reported no history of gastroesophageal reflux symptoms >5 years before diagnosis/interview, suggesting that avoidance because of gastroesophageal reflux symptoms is unlikely to explain the findings. Esophageal adenocarcinoma patients may have avoided using NSAIDs because of early symptoms associated with their cancer, which may explain why there seems to be a stronger protective association >1 year before compared with >5 years before interview date.
NSAIDs inhibit COX-1, which is responsible for maintaining the integrity of the gastroesophageal mucosa, and are associated with an increased incidence of upper GIT symptoms ( 37– 39). Aspirin and other NSAIDs are well known to have damaging effects on the upper GIT ( 37– 39). Patients with symptomatic Barrett's esophagus may avoid using NSAIDs because of medical advice or contraindications. We adjusted for upper GIT disorders in the analyses and also for Barrett's esophagus in the esophageal adenocarcinoma analyses. The associations were strengthened after adjustment for upper GIT disorders, suggesting that for esophageal adenocarcinoma patients, at least, the avoidance of these medications because of contraindications does not fully explain the associations.
The data suggest that both aspirin and non-aspirin NSAIDs may prevent Barrett's esophagus and esophageal adenocarcinoma. The strength of the associations was not stronger in the esophageal adenocarcinoma patients than in the Barrett's esophagus patients, suggesting that the protective effect may occur early in the inflammation-metaplasia-carcinoma sequence before the development of Barrett's esophagus. It is also not possible to rule out the possibility that NSAIDs exert their effect before the diagnosis of esophagitis.
Very little research has been reported about the use of NSAIDs and the prevention of Barrett's esophagus. In a rat model, celecoxib, a selective COX-2 inhibitor, reduced the incidence of columnar-lined esophagus compared with the control group. It was also associated with a milder form of esophagitis in the animals. The authors suggested that celecoxib may prevent metaplasia in the esophagus by suppressing esophagitis ( 40). However, most research to date has focused on the effect of NSAIDs on the metaplasia-dysplasia-carcinoma pathway. One study of Barrett's esophagus patients reported that histologic, flow cytometric, and genetic abnormalities were less frequent in those who used NSAIDs ( 41).
The mechanism whereby NSAIDs/aspirin may confer a protective effect against Barrett's esophagus and esophageal adenocarcinoma is likely to be through the inhibition of COX-2. NSAIDs inhibit COX-2, which can be activated by proinflammatory cytokines, including interleukin-1β (IL-1β) and tumor necrosis factor-α (TNF-α), following mucosal injury. Overexpression of IL-1β has been reported in patients with esophagitis, Barrett's esophagus, and esophageal adenocarcinoma ( 42), and overexpression of TNF-α has been reported in patients with Barrett's esophagus ( 43). Exposure of the distal esophagus to acid reflux and components of bile reflux has been associated with the activation of COX-2 ( 44– 46). It has been suggested that there is an increase in COX-2 expression before there is any visible histologic evidence of esophagitis ( 45). Overexpression of COX-2 has been shown in patients with reflux esophagitis, Barrett's esophagus, dysplasia, and esophageal adenocarcinoma ( 14, 44, 47– 54), and expression of COX-2 has been reported to be higher in Barrett's esophagus patients compared with reflux esophagitis patients ( 47). COX-2 inhibitors have been shown to reduce the rate of proliferation in Barrett's esophagus cell lines ( 14), and it has been suggested that COX-2 expression occurs early on in the development of Barrett's esophagus ( 50).
As stated, there are several possible interpretations for the apparent protective association of NSAIDs against Barrett's esophagus and esophageal adenocarcinoma, which result from the inherent limitations of observational studies. Intervention trials are currently under way, including a trial using aspirin as a supplement to proton pump inhibitor therapy in Barrett's esophagus patients ( 55). This type of trial is superior to observational studies, such as the FINBAR study, and the results are eagerly awaited although unlikely to be available for some time. If NSAIDs prevent esophageal adenocarcinoma, it is possible that the stage of protection may be earlier than the stage of intervention of these trials. Intervention trials at an earlier stage in the inflammation-metaplasia-carcinoma sequence (possibly in patients with reflux esophagitis or gastroesophageal reflux) may be warranted.
Appendix A: Medications
|Askit powders *|
|Brand name||Generic name|
↵* No one reported taking these medications at least once weekly for ≥6 months.
Grant support: Ireland-Northern Ireland Co-operation Research Project grant sponsored by Research & Development Office (Belfast, Northern Ireland) and Health Research Board (Dublin, Ireland) and by Ulster Cancer Foundation (Belfast, Northern Ireland). Further funding was obtained for the esophagitis study from the Northern Ireland Research & Development Office Clinical Fellowship.
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.
We thank the study participants and their families for their contributions; clinicians who were contacted throughout the study period and their secretaries for administrative support; research team, including Siobhan Reynolds, Majella Gallagher, Carol Anderson, and Martin McAnaespie, and Dr. Damian McManus with the help in classifying the tumor sites; and Northern Ireland Cancer Registry and National Cancer Registry Cork for their support and involvement in the research.
- Received November 29, 2005.
- Revision received February 13, 2006.
- Accepted March 2, 2006.
- ©2006 American Association for Cancer Research.