News and Articles on Health. Medical research topics.

Because there is evidence to support an association of CD44 and CD44 variant isoforms with HNSCC progression, it is natural to consider a putative role for CD44 receptors in head and neck dysplastic lesions. Increased CD44 and CD44 variant isoform expression is observed in most dysplastic lesions of the head and neck. Nonetheless, there are conflicting reports regarding CD44 status as a predictor of clinical outcomes in laryngeal and oral dysplasia. Osteopontin (a ligand with reported specificity for the CD44 v6 isoform) and CD44v6 are both overexpressed in laryngeal dysplasia and laryngeal SCC, and their levels correlate with degree of dysplasia, disease-free survival, and recurrence. Plasma osteopontin levels also correlate clinically with a worse response to chemoradiation treatment. CD44v6 expression was reported to correlate with developing both dysplastic and malignant oral epithelia. In contrast, loss of CD44 expression has been observed in adjacent nontumor epithelium in tongue cancer. Loss of CD44v7-8 was associated with worse prognosis in oral cavity dysplasias associated with betel quid chewing. Thus, regarding the role of CD44 variant isoforms in head and neck dysplasia, a few studies point to the importance of CD44 v6 and its ligand osteopontin in the development of dysplastic lesions and progression to HNSCC, but the role of other CD44 variant isoforms is uncertain.

Interaction of HA and CD44 Promotes EGFR-Mediated Pathways to Increase HNSCC Growth, Survival, and Chemoresistance

The epidermal growth factor receptor (EGFR) is expressed in nearly all HNSCCs, and its overexpression is associated with poor prognosis and diminished survival. Pharmaceutical agents targeting EGFR and EGFR signaling pathways have recently been introduced for the treatment of advanced HNSCC.33 The concept that EGFR and EGFR signaling might be linked to CD44 was first suggested for cervical carcinoma cell lines, and subsequently was demonstrated for breast cancer and glioblastoma cell lines. CD44 colocalized with EGFR in both cervical carcinoma clinical specimens and cell lines, and increased CD44 expression resulted in enhanced EGFR activity. CD44 also colocalized with both EGFR and ERBB2 in metastasizing breast cancer cells. CD44 and EGFR interact in glioblastoma cells, and this interaction resulted in enhanced HA-mediated phosphorylation of extracellular signal regulated kinases 1 and 2 (ERK1 and ERK2). Additionally, co-expression of CD44 and EGFR in glioma cell lines differentially up-regulated the HA-induced expression of a number of genes associated with cellular invasion and proliferation, including TIMP metallopeptidase inhibitor 1 (TIMP1) and MYC (alias c-Myc). Because of the apparent importance of the EGFR pathway in HNSCC, our group investigated and reported a novel linkbetween HA-CD44 interaction and EGFR signaling in HNSCC.

CD44-EGFR-Mediated Oncogenic Signaling

Analysis of anti-CD44-mediated immunoprecipitates from HNSCC cell lysates indicates that HA/CD44 interaction leads to CD44-EGFR complex formation. HA treatment was shown to promote phosphorylation of EGFR, activating the EGFR tyrosine kinase (TK) and promoting downstream EGFR-mediated pathways including Ras, RhoA, Rho kinase, and phosphatidylinositol-3 (PI-3) kinase signaling. The CD44-EGFR complex also associates with leukemia-associated Rho-guanine (LARG) nucleotide exchange factor to effect downstream signaling through ras and RhoA. Ras activation by HA-mediated CD44-LARG-EGFR complex formation promotes Raf-1 phosphorylation and MAPK activation. The stimulation of EGFR and downstream EGFR signaling activity by HA treatment was similar to epidermal growth factor stimulation. Wang et al38 reported that HA/CD44-mediated activation of EGFR signaling leads to MAPK phosphorylation/activation that results in increased HNSCC tumor cell growth; however, the HA-mediated increased cell growth was blocked by treatment with inhibitors of EGFR and MAPK.

CD44 v6-containing isoforms also appear to promote tumor progression. Transfection of CD44 v6 converted nonmetastatic rat carcinoma cells into metastatic cells, and co-injection of anti-CD44 v6 antibody into these same cells suppressed their metastatic behavior. The CD44 v6 splice variant was also found to stimulate sustained increased mitogen activated protein kinase (MAPK) levels and subsequent downstream Ras signaling, resulting in increased tumor cell proliferation.

Because of evidence linking CD44 v6 as a marker for HNSCC tumor proliferation and metastasis, a phase I clinical trial was conducted to evaluate anti-CD44 v6 antibody therapy for the treatment of incurable HNSCC, with a 10% response rate reported.23 Administration of anti-CD44 v6 antibody to a CD44 v6 isoform-expressing HNSCC cell line resulted in decreased cell proliferation and increased cisplatin sensitivity.20 Using the same anti-CD44 v6 antibody, immunohistochemical analysis of HNSCC tissue specimens indicated that CD44 v6 isoforms were preferentially expressed in metastatic lymph nodes, and strong expression of CD44 v6 in primary tumors was significantly associated with advanced T status, perineural invasion, and shorter disease-free survival. Although these studies are small, the association of CD44 v6 isoforms with clinically advanced HNSCC and the intriguing results of anti-CD44 v6 therapy in a phase I clinical trial suggest a need for further study of the role of this CD44 variant isoform in HNSCC.

Although all CD44 isoforms share HA binding domains, certain CD44 variant isoforms, such as CD44 v10, exhibit significantly reduced affinity for HA binding. It is thought that the reduction in HA-mediated cell adhesion in tumor cells expressing CD44 v10 may be the earliest event in the onset of tumor migration and invasion. The unique structure of CD44 v10 may also cause constitutive activation of CD44-cytoskeleton interactions that induce tumor cell migration and invasion. CD44 v10-containing isoforms have been reported to promote tumor progression in breast and renal cell carcinoma. Studies have revealed that CD44 v10-transfected breast tumor cells display higher migration/invasion potential, produce higher levels of basic fibroblast growth factor and interleukin-8, and exhibit more potent tumor growth potential than parental control cells. Treatment of a CD44 v10 isoform-expressing HNSCC cell line with anti-CD44 v10 antibody resulted in decreased tumor cell proliferation and increased cisplatin sensitivity. The same anti-CD44 v10 antibody was used in the analysis of HNSCC clinical tissue specimens, demonstrating that CD44 v10 isoforms were preferentially expressed in metastatic lymph nodes. In addition, strong expression of CD44 v10 in primary tumors was significantly associated with distant metastasis, failure of radiotherapy, and shorter disease-free survival. Further studies to establish the interaction of the ligands involved in CD44 v10 signaling in HNSCC are needed to determine its role in HNSCC metastasis and treatment resistance.

Although we acknowledge that the recent scientific literature is without broad consensus, support for a link between CD44 variant expression and HNSCC progression is bolstered by the recent description of three CD44 variants associated with HNSCC progression in both clinical specimens and in in vitro studies. Furthermore, the identification of CD44 as a putative cancer stem cell marker, as well as a marker for chemoresistance, is also consistent with the notion that certain CD44 variant isoforms may be contributory to HNSCC progression.

The mechanism of action for most CD44 variant isoforms in solid malignancies is not well understood. Several CD44 variant exons are known to contain binding domains for various growth factor ligands. The v3 exon is known to contain important glycosaminoglycan (GAG) attachment sites, and it has been postulated that tumor cells with CD44 receptors exhibiting these GAG sequences are involved with heparin binding growth factors such as basic fibroblast growth factor, vascular endothelial growth factor (VEGF), and heparin binding epidermal growth factor. Thus, CD44 v3-containing isoforms appear to be capable of promoting tumor cell proliferation in an HA-independent manner, through the presence of additional growth factor binding sites. The CD44 v3 isoform is expressed in the metastatic breast tumor cell line Met-1, in which it has been shown to bind VEGF, suggesting that this isoform may promote breast tumor-associated angiogenesis. CD44 v3 also appears to be linked to several other tumorigenic molecules, promoting tumor cell migration and the invasive tumor cell phenotype. In particular, CD44 v3 colocalizes with the active form of matrix metalloproteinase-9 (MMP-9) in Met-1 cells, promoting degradation of the ECM to facilitate tumor cell invasion. This isoform also up-regulates cytoskeleton function, through ankyrin, to activate the membrane-associated actomyosin contractile system and mediate tumor cell migration.

The role of CD44 v3 isoforms in HNSCC progression is highlighted by studies showing an association of v3-containing isoforms with HNSCC growth, migration, and MMP expression. Transfection of a CD44 v3 isoform into a nonexpressing HNSCC cell line also resulted in significantly increased tumor cell migration, but not proliferation. Treatment of a CD44 v3 isoform expressing HNSCC cell line with anti-CD44 v3 antibody decreased in vitro proliferation and increased cisplatin sensitivity. Using the same anti-CD44 v3 antibody, immunohistochemical tissue analysis revealed that CD44 v3 isoforms were preferentially expressed in metastatic lymph nodes. Additionally, strong CD44 v3 isoform expression in primary tumors was significantly associated with advanced T status and positive lymph nodes, but did not correlate with disease-free interval. In summary, there are several in vitro and histopathological studies suggesting that CD44 v3 isoforms are involved in HNSCC progression behaviors. To verify these findings, further research is needed to elucidate the mechanisms of CD44 v3 signaling.

Because of evidence linking CD44 variant isoform expression with tumor progression in a variety of solid malignancies, multiple groups have studied the role of CD44 variant isoforms in HNSCC. Nonetheless, the role of CD44 variant isoforms in HNSCC remains controversial. The literature is conflicting regarding the significance of the under- or overexpression of CD44 variant isoforms in HNSCC.

Whereas some studies have found a correlation between increased CD44 variant expression and HNSCC progression, other studies have reported no correlation or negative correlation. Reategui et al18 described a novel CD44 v3 isoform in both tissue and soluble form that correlated with HNSCC status. Wang et al19, 20 reported that CD44 v3-containing isoforms were associated with HNSCC lymph node metastasis and advanced T status, CD44 v6-containing isoforms were associated with perineural invasion and shorter survival, and CD44 v10-containing isoforms were associated with distant metastasis and failure of radiotherapy. Kawano et al found correlations between CD44 v6 expression in HNSCC and tumor volume, lymph node metastasis, and shorter survival. Others, however, have reported that down-regulation of various CD44 variant isoforms correlated with a worse prognosis. Kanke et al4 reported that down-regulation of CD44 v2 correlated with poorer differentiation and shorter overall survival, whereas down-regulation of CD44 v6 correlated with a higher rate of cervical metastasis. Other groups, including Van Hal et al6 and Herold-Mende et al, found no correlation between CD44 splice variants and any clinicopathologic variables, and these authors concluded that CD44 variant isoforms do not play a role in HNSCC progression.

There are several possible explanations for these discrepant results. First, different studies have used different antibodies, making comparisons across reports from different research groups difficult. Certain CD44 variant domain epitopes may become hidden and not recognized by some antibodies because of post-translational changes (eg, glycosylation) that alter the three-dimensional conformation of the protein. In addition, assessment of immunostaining positivity is dependent on what region of the tumor is examined, and it has been reported that there are often large areas within the tumor that do not stain for many CD44 isoforms. Tissue heterogeneity appears to be another important factor in the conflicting results, with the specific region of the tumor that is examined apparently having high importance in the determination of CD44 expression levels. CD44s and CD44 variant receptors appear to be concentrated at invasive fronts of HNSCC tumors, but have decreased expression in other parts of the tumor.21 It has been hypothesized that the major CD44 ligand HA, which is present at high levels in the ECM surrounding HNSCCs, is responsible for the stimulation of CD44 to promote migration, invasion, and metastasis. Hyaluronan has been reported to stimulate both CD44 and lymphatic vessel endothelial hyaluronan receptor (LYVE-1) to promote nodal metastasis in HNSCC.

All isoforms of the CD44 membrane receptor share a common ligand-binding region for hyaluronan (HA), a glycosaminoglycan component of the extracellular matrix (ECM). Recently, HA has been studied with regard to its ability to promote various CD44-mediated signaling pathways. HA/CD44 signaling has been linked to solid tumor progression, including invasion, metastasis, and chemoresistance. In cancer cells, HA interaction with CD44 promotes multiple signaling pathways that influence tumor cell progression behaviors in a variety of solid tumors. Because all CD44 isoforms are capable of HA interaction, HA-mediated signaling may involve CD44s or variant isoforms.

An accumulating body of evidence indicates that HA and CD44 signaling play an important role in HNSCC tumor progression and chemoresistance. Here we review the recent scientific literature regarding the role of HA-mediated CD44 signaling in HNSCC.

CD44 Variant Isoform Expression Is Associated with HNSCC Progression

All CD44 isoforms contain a conserved extracellular binding domain for HA, along with common cytoplasmic domains capable of interacting with cytoskeleton proteins to activate cell signaling pathways. The ability of CD44 receptors to provide a direct link between the ECM and the cytoskeleton, coupled with their ability to interact with a multitude of signaling kinases, explains how one family of molecules is able to mediate diverse cellular functions.

CD44 isoforms are expressed to varying degrees in both normal and tumor cells, suggesting a normal cellular function for many of the different CD44 receptor isoforms. How the same CD44-mediated signaling pathways can promote both normal and malignant cellular behaviors could be explained by several different mechanisms. For example, the regulation of CD44 signaling includes variable N- or O-linked glycosylation of the CD44 extracellular domain that differentially modulates the interaction of cytoskeletal proteins such as ankyrin with the cytoplasmic domain. It is thought, however, that the most important mechanism regulating CD44 signaling is through alternative exon splicing leading to expression of CD44 variant isoforms. In cancer cells, the exon-splicing mechanism can result in overexpression or novel expression of certain CD44 variant isoforms. The insertion of variant exons into the extracellular component of the CD44 receptor has been shown to alter the signaling properties of the CD44 receptor by providing additional binding domains for molecules other than HA. The insertion of variant exons may also result in changes to the HA-binding affinity of the CD44 isoform, further altering its signaling behavior.

Overexpression of several CD44 variant isoforms has been associated with tumor progression, suggesting that these CD44 isoforms may have unique signaling properties. In colon cancer, CD44 v3 has been shown to promote invasion and resistance to apoptosis, and CD44 v6 has been associated with metastasis and shorter disease-free survival. In lung cancer, there is preferential CD44 variant expression in squamous cell carcinoma and bronchioalveolar carcinoma, in which the v5 and v6 variants appear to promote metastasis. Numerous reports have shown that CD44 variants promote breast cancer progression, including the association of CD44 v3-containing isoforms and breast cancer metastasis.

Head and neck squamous cell carcinoma (HNSCC) is an aggressive malignancy that may involve the oral cavity, pharynx, larynx, and paranasal sinuses. The mechanisms of tumor progression underlying the clinical behavior of HNSCC remain unclear. CD44 comprises a family of transmembrane receptors that can give rise to multiple CD44 variant isoforms. Hyaluronan (HA), a major extracellular matrix component is the primary ligand for CD44 receptors. HA and CD44 signaling play an important role in HNSCC progression. Several CD44 variant isoforms (including v3-, v6-, and v10-containing isoforms) are associated with advanced disease, possibly through unique growth factor interactions with binding domains in the inserted variant regions of the cytoplasmic domain of CD44. In HNSCC, HA mediates the formation of a complex including CD44 and the epidermal growth factor receptor (EGFR) which is overexpressed in a large proportion of HNSCCs. Downstream effectors under EGFR regulation are activated, promoting promote cell growth and tumor survival. The leukemia-associated Rho-guanine nucleotide exchange factor (LARG) also associates with CD44 and EGFR to promote several Ras and RhoA pathway effectors, leading to cell migration, growth, and tumor survival. The secretion of matrix metalloproteinases, necessary for tumor cell invasion, is also regulated by these HA/CD44-mediated pathways. Finally, EGFR-mediated pathways play major roles in the HA/CD44 promotion of chemoresistance in HNSCC. Understanding HA/CD44-mediated signaling pathways may lead to improved treatment of HNSCC.

Head and neck squamous cell carcinoma (HNSCC) is a malignancy that may involve the oral cavity, pharynx, larynx, and paranasal sinuses. It is the sixth most common cancer worldwide. Advanced HNSCC is an aggressive disease, associated with major morbidity and mortality. The 3-year survival rate for patients with advanced-stage HNSCC treated with standard therapy is only 30%–50%.1 Resistance to standard therapy continues to be a limiting factor in the treatment of HNSCC. Nearly 40% to 60% of HNSCC patients subsequently develop locoregional recurrences or distant metastases.There is a great need to clarify the mechanisms of tumor progression underlying the clinical behavior of HNSCC.

CD44 comprises a family of transmembrane receptors found on many different benign and malignant cells. The human gene (CD44) contains 19 exons. Up to 10 exons (primarily exons 6 through 14) may be alternatively spliced to give rise to multiple variant CD44 isoforms (eg, CD44 v3, CD44 v6, CD44 v10, etc) that, along with standard CD44 (CD44s; ∼85-kDa isoform with no variable exons), make up the CD44 class of receptors. Alternative splicing and post-translational modifications are tightly regulated and permit expression of multiple different CD44 isoforms. The many splicing possibilities of the variable exons of CD44 could in theory give rise to a vast number of CD44 variants, although relatively few have been described to date. Various CD44 variant isoforms are differentially expressed in normal and malignant cells, and confirmation of CD44 isoform expression in HNSCC is well documented, in both tissue specimens and established cell lines.

Adenoma recurrence is already widely accepted as a surrogate end point for colorectal cancer. However, our data on initial adenoma, taken together with our earlier findings in this cohort, which demonstrated an inverse relationship between use of aspirin and colorectal cancer, provide additional evidence to substantiate the protective effect of aspirin on colorectal neoplasia. In our previous study, the relative risk for cancer after 20 years of consistent aspirin use was 0.56 (CI, 0.36 to 0.90). The influence of duration on risk for colorectal cancer (but not adenoma risk) seems plausible because there may be a dwell period of a decade or longer within which an adenomatous polyp becomes invasive cancer. Our results also demonstrate that optimal chemoprevention requires substantially higher doses of aspirin than those previously considered. These findings may have clinical implications related to potential adverse events associated with prolonged use of aspirin. Gastrointestinal bleeding, the most common major toxicity of aspirin, occurs in approximately 1 in 100 people taking aspirin over at least 2 years, and considerable evidence suggests that this toxicity, as well the severity of bleeding, is strongly dependent on increasing dose. In a large, randomized trial of aspirin in the secondary prevention of cerebrovascular events, the annual rate of upper gastrointestinal bleeding was 0.6% in patients taking 1200 mg/d, 0.3% for those assigned to 300 mg/d, and 0.1% for those taking placebo. Another rare but potentially devastating adverse consequence of aspirin use, intracerebral bleeding, may also be related to increasing dose. In our cohort, participants tak ing the highest dose of aspirin (_2 tablets/d) had a 2-fold higher risk for subarachnoid hemorrhage. Treatment of large numbers of healthy persons to prevent colorectal cancer in a small percentage requires a high standard of safety as well as efficacy. For example, the U.S. Preventive Services Task Force presently recommends routine use of low-dose aspirin as primary prophylaxis against myocardial infarction only in those at high risk for coronary heart disease. Thus, it has been suggested that any agent requires a safety profile that meets or exceeds that of low-dose aspirin before it can be recommended for widespread use in colorectal cancer prevention.

Moreover, 2 cost-effectiveness analyses concluded that present colorectal cancer screening strategies were generally superior to daily use of 325 mg of aspirin. Although the overall cost-effectiveness of aspirin improves if cardiovascular benefits are incorporated, our findings suggest that the optimal doses for cardiovascular and neoplasia prevention differ considerably. Because our study is observational, the results are not as definitive as an intervention trial designed to directly examine the effects of daily administration of low-dose, standard-dose, or 2 standard-dose aspirin tablets on subsequent risk for initial adenoma, as well as adverse outcomes in a healthy population. However, such a trial is probably not feasible given the need for a large number of patients to examine relatively rare outcomes and even more infrequent toxicities. Moreover, the cost of surveillance colonoscopy would be prohibitive, and ethical concerns about adverse consequences related to high doses of aspirin may be considerable. In summary, our results demonstrate that both shortand long-term aspirin use reduces the risk for adenoma in an average-risk population. However, chemoprevention efforts using aspirin may require substantially greater doses of aspirin than those currently recommended for the prevention of cardiovascular disease. The potential of regular use of moderate aspirin doses to prevent colorectal neoplasia necessitates further evaluation. The risk–benefit profile must be thoroughly considered, especially when compared with other potential strategies.

Because most women in our study used aspirin for reasons other than cardiovascular prevention, it is likely that most women were consuming standard 325-mg tablets. However, to address potential secular changes in the pattern of consumption, we specified both low-dose and standard-dose tablets in subsequent questionnaires. A limitation of our study is that we cannot directly assess the relationship of risk to more refined categories of dose (for example, number of milligrams) or more frequent categories of use (for example, daily). Nonetheless, we believe that the number of tablets used per week is a relevant measure of a range of aspirin intake that has been previously associated with other dose-related outcomes in this cohort. Moreover, any misclassification of dosage would have tended to bias the observed dose relationship to the null. Because adenomas are essentially diagnosed only through endoscopy, a potential limitation of a populationbased analysis is differential exposure to endoscopy, resulting in a bias in ascertaining our outcome.

To minimize this detection and selection bias, we restricted our cohort to patients who underwent endoscopy during the study period, an approach that is consistent with other observational analyses and clinical trials assessing the risk for adenomatous polyps. Although we acknowledge that this restriction may limit the generalizability of our findings to the remaining cohort, our previous analyses of colorectal cancer in the larger cohort compared with analyses of adenoma in the endoscopy cohort have yielded complementary results. Furthermore, because most adenomas are asymptomatic, discontinuation of aspirin due to symptoms attributable to adenomas or differential bleeding is unlikely to explain our results. Most important, our findings remained unchanged despite stratification by the indication for endoscopy, including bleeding. Because we assumed that many screening endoscopies performed in this cohort were sigmoidoscopies rather than colonoscopies, we restricted our analysis to adenomas of the distal colon and rectum to minimize detection bias. Therefore, women with only proximal adenomas were classified as non–case participants.

Nonetheless, we believe that our results can be generalized to proximal adenomas. In a secondary analysis in which we defined cases as proximal adenomas without distal findings, we observed a similar effect of aspirin on the risk for proximal adenomas. Moreover, a recently completed intervention trial demonstrated comparable effect for aspirin on proximal and distal adenomas, and our previous analysis in this cohort observed a similar effect for aspirin on the risk for proximal and distal cancer. Our results may have potential implications for population- based colorectal cancer prevention strategies using aspirin. Because existing screening efforts using endoscopy and polypectomy are costly, invasive, and not widely used, there is substantial enthusiasm for aspirin chemoprevention of colorectal cancer.

Although low-dose aspirin is sufficient to inhibit colonic prostaglandins and constitutively expressed cyclooxygenase- 1 isoenzymes in anucleated platelets, higher doses are needed to inhibit the cyclooxygenase-2 inducible isoenzymes in nucleated cells, such as those found in the colonic epithelium. The cyclooxygenase-2 pathway seems to be more directly relevant to colorectal neoplasia, because expression of cyclooxygenase-2 but not cyclooxygenase- 1 is upregulated in colorectal adenoma and carcinoma compared with normal colonic mucosa. Furthermore, increasing experimental data suggest that aspirin works through several other potential antineoplastic mechanisms that are maximized at higher doses. Consequently, our findings seem to be biologically plausible and support current understanding of aspirin’s known mechanisms. Our data support an effect of aspirin use in a primary prevention cohort that has a lower risk for adenoma than patients with a history of adenoma or cancer enrolled in recurrence trials. Overall, about 5% of our participants received a diagnosis of distal adenoma compared with up to 45% diagnosed with either recurrent proximal or distal adenoma in the aspirin intervention trials. Our proportion of cases is also lower than that of other average-risk populations. This may be due to the lower median age of our cohort, a higher proportion of women of premenopausal status, and a lower prevalence of other adenoma risk factors among potentially health-conscious nurses.

Furthermore, our main analyses focused on distal adenomas, and we required strict pathologic confirmation of all cases. Hence, women with only proximal adenomas or adenomas that were endoscopically ablated or not retrieved for pathologic examination were not classified as case-participants. In addition, self-report of polyps may have led to underascertainment. However, studies of other population-based cohorts have confirmed the validity of self-reports of lower endoscopy, and a parallel study of a similarly educated cohort of health professionals has validated our method of self-report of polyps. Most important, misclassification or underascertainment of adenoma cases would have probably diluted the protective benefit we observed. The strengths of our study include its prospective design, repeated assessments of aspirin use, detailed data on potential confounders, long-term follow-up, and consistently high response rate. Since the participants were all nurses, the accuracy of self-reported aspirin use is likely to be high. Moreover, because aspirin use was recorded before endoscopy, errors in recall of aspirin use would not be influenced by knowledge of the presence or absence of adenoma. Hence, nondifferential misclassification of aspirin dose or frequency would have tended to attenuate rather than exaggerate a true association.

In this large prospective cohort study, regular aspirin use (defined as _2 tablets/wk) was associated with a statistically significant 25% reduction in the risk for sporadic colorectal adenomas in an average-risk population. Of note, the greatest reduction in risk was observed at doses greater than 14 standard tablets per week. Moreover, a similar dose–response relation was observed among regular short-term (_5 years) and long-term (_5 years) aspirin users. In contrast, increasing duration of aspirin use was not significantly associated with risk after adjusting for aspirin dose. Controlling for other known or suspected risk factors for colorectal adenoma and cancer did not alter these findings. Although our study was limited to women, previous reports have also demonstrated an association between reduced risk for colorectal adenoma and aspirin use in men. Results from 2 intervention trials of patients with a history of adenoma or a history of colorectal cancer have established a chemopreventative influence of aspirin on recurrent adenomas after 3 years. However, these studies examined limited doses and yielded conflicting data. The trial of patients with previously documented colon cancer demonstrated a statistically significant reduction in incidence of colorectal adenomas in patients randomly assigned to standard-dose aspirin. In contrast, the trial of postpolypectomy patients did not observe any reduction in adenoma incidence in a group randomly assigned to standard-dose aspirin but did observe a moderate benefit in a group randomly assigned to lowdose aspirin. The results of several earlier studies are consistent with our findings that higher doses of aspirin may be needed for the optimal prevention of colorectal adenomas.

The Physicians’ Health Study, a randomized, placebocontrolled trial, found a relative risk of 0.86 for adenoma among men randomly assigned to low-dose aspirin (325 mg of aspirin every other day). With only 263 total cases, this did not achieve statistical significance but was consistent with our results in the low-dose categories. In a recent French trial that assessed adenoma recurrence, the rate of adenoma recurrence at 1 year was 25% among patients randomly assigned to receive 300 mg of daily soluble aspirin compared with 35% among patients randomly assigned to receive 160 mg per day. However, this difference also did not achieve statistical significance because both aspirin groups had included only 126 patients. A posthoc analysis of a dietary intervention trial found a substantially lower incidence of recurrent adenomas among participants who reported use of more than 1 standard aspirin tablet per day compared with those who used lower doses. In a nested cohort study conducted in the United Kingdom, 300 mg of aspirin per day was associated with a statistically significant reduction in adenoma risk, whereas no benefit was seen with daily doses of either 75 or 150 mg. Finally, a randomized trial of the cyclooxygenase- 2 inhibitor celecoxib demonstrated that high, but not standard, doses significantly reduced adenoma burden in patients with familial polyposis.

We also examined whether aspirin dose influenced both short-term (_5 years) and long-term (_5 years) users. For both regular short-term and long-term users, the dose relationship was statistically significant. Among women who reported using more than 14 tablets per week, the multivariate relative risk for adenoma was 0.36 (CI, 0.19 to 0.71; P _ 0.001 for trend) for short-term users and 0.52 (CI, 0.38 to 0.72; P _ 0.001 for trend) for long-term users. To test whether duration of aspirin use altered the association between aspirin dose and adenoma risk, we added to the multivariate model the cross-product interaction term of aspirin dose and duration of aspirin use (each measured continuously). However, no statistically significant interaction seemed to be evident between aspirin dose and duration of use (P _ 0.2 for interaction). We also assessed the effect of duration of aspirin use on adenoma risk. After we controlled for other adenoma risk factors, we observed a statistically significant reduction in risk, even after 5 or fewer years of use, and progressively greater reductions in risk with longer use. The most substantial reduction was seen in women who had used aspirin for more than 20 years (P _ 0.001 for trend). However, when we adjusted for the number of tablets taken per week, the effect of duration of use was no longer statistically significant (P _ 0.2 for trend).

The effect of regular aspirin use was not modified by age, the presence of a family history of colorectal cancer, or use of postmenopausal hormones. Regular aspirin use was associated with a multivariate relative risk for adenoma of 0.71 (CI, 0.60 to 0.85) among women younger than 60 years of age and 0.78 (CI, 0.66 to 0.92) for those 60 years of age or older. In addition, the multivariate relative risk for regular aspirin use was 0.73 (CI, 0.57 to 0.92) for women with a family history (first-degree relative) of colorectal cancer and 0.76 (CI, 0.66 to 0.87) for those without a family history. Finally, stratified by postmenopausal hormone use, the multivariate relative risk for regular aspirin use was 0.73 (CI, 0.61 to 0.88) among women who never used hormones, 0.79 (CI, 0.62 to 1.01) among women who were past users of hormones, and 0.71 (CI, 0.57 to 0.88) among women who were current users of hormones. Because many of the screening endoscopies in this cohort were sigmoidoscopies, we limited our primary analyses to distal colorectal adenomas. Nonetheless, we examined the possibility that aspirin may have a differential effect on proximal adenomas by performing a secondary analysis in which the outcome of interest was defined as proximal adenoma without distal synchronous adenoma. Among 351 cases of proximal adenoma without distal adenoma, regular aspirin use remained inversely associated with adenoma risk (multivariate-adjusted relative risk, 0.81 [CI, 0.65 to 1.01]). We also considered the possibility that concurrent use of other nonsteroidal anti-inflammatory drugs may have influenced the relationship between aspirin and adenoma risk. Because data on nonsteroidal anti-inflammatory drugs were not consistently collected until 1990, we restricted the cohort to the 11 360 women who underwent endoscopy after 1990. Among these women, the multivariate relative risk for adenoma in regular aspirin users was 0.78 (CI, 0.67 to 0.92). After additional adjustment for regular use of nonsteroidal anti-inflammatory drugs, the risk for adenoma in regular aspirin users was not materially altered (relative risk, 0.74 [CI, 0.63 to 0.88]).