The excessive consumption of alcohol-containing beverages, or ACBs, moderately increases risks of cancers of the mouth, pharynx, larynx, esophagus and liver.1 ACBs also interact with cigarette smoking to increase greatly the risk of several of these cancers.2 However, alcohol (ethanol) is not carcinogenic in animals,3 and only its prolonged abuse in humans is related to hepatocellular carcinoma, or HCC. Indeed, the causal nature of the ACB-HCC association has been questioned.4
It is unlikely that the use of mouthwashes that contain alcohol increases the risk of developing oropharyngeal cancer.
The finding that ACBs strongly enhance the effect of smoking in producing cancer is puzzling in view of the weak or absent carcinogenic activity of ethanol itself. To explain this, researchers have suggested that the ethanol in ACBs acts by altering the surface of cells or mucosal tissues of the oral cavity. This alteration may increase exposure to, or facilitate the action of, carcinogens in cigarette smoke, food or ACBs themselves. Researchers also have suggested that the carcinogenic effect of high consumption of ACBs on the oral cavity may be mediated in part by a systemic mechanism.5 The relationship between ACBs and HCC may be even more complex, involving liver damage and an increase in DNA transcription errors during cell regeneration. Additional factors may play a role (for example, the severe dietary deficiencies so prevalent among alcohol abusers).
 |
GENERAL CONSIDERATIONS
|
|---|
In 1979, a combined case-series report and case-control study raised concern that use of alcohol-containing mouthwash, or ACM, also causes oropharyngeal cancer, or OPC.6 We review that report below, as well as all of the other available epidemiologic studies. In addition, we describe four general considerations that bear on epidemiologic studies of the ACM-OPC relationship.
First, relative risks, or RRs, for OPC up to 5 commonly are seen for smoking7 and up to 15 for the synergistic effect of smoking and drinking.2 In addition, evidence exists8–11 that mouthwash use is elevated among drinkers and especially among smokers. The effect of this pattern of mouthwash use is to make it difficult or impossible to eliminate the confounding effects of smoking and drinking from the evaluation of the mouthwash-OPC relationship.
Second, the difficulty of controlling for ACB use also is increased because ACB consumption may be underreported by subjects in epidemiologic studies. Underreporting of a confounder precludes fully controlling its effects,12 and may even create a false-positive result.13 Underreporting of ACB use will lead to inadequate control of the effects of ACBs and of the ACB-smoking interaction in any evaluation of mouthwash use and OPC.
Third, although all of the studies reviewed attempted to evaluate the ACM-OPC relationship, only four included information about the alcohol content of the mouthwash products evaluated. Thus, many of the findings reviewed here relate to mouthwash use, not to ACM use. However, since most mouthwashes contain ethanol, the available literature does address the ACM-OPC relationship to a reasonable degree.
Fourth, an ACM-OPC relationship seems implausible because ACM use involves low exposure to alcohol. We are not aware of data that allow a comparison of the topical exposure of the oral cavity to alcohol resulting from ACM use with that from ACB use. Intuitively, one or two uses per day of an ACM that contains 25 percent alcohol, with each use lasting up to 30 seconds, seem unlikely to produce a topical exposure to alcohol equivalent to that from abuse of ACBs. At most, such exposure from ACM use might be considered similar to that from one or two drinks of ACBs per day. This level of ACB consumption probably does not increase the risk of OPC.
Even if the topical exposure of the oral cavity to alcohol from ACM use were considered equivalent to that from a higher level of ACB consumption, the biological credibility of a causal relationship between ACM and OPC still would be weak. The carcinogenic agent in ACBs may not be ethanol, and may not be present in ACM. Also, some evidence exists that it is ingested alcohol, not topical exposure, that plays a major role in causing OPC.5
|
OVERVIEW OF EPIDEMIOLOGIC STUDIES
|
|---|
Nine epidemiologic studies relating to mouthwash use and OPC have appeared in the literature since 1979. Three included some positive results and six had results that were entirely negative. Eight of the studies are reviewed in chronological order. The ninth, a large and broadly based study, has been seen as providing some support for a causal relationship between ACM and OPC. We review it in detail in the next section of this report.
In 1979, Weaver and colleagues6 first suggested a link between ACM and OPC. In a series of 200 patients with OPC, they identified 11 people who did not smoke or consume ACBs. However, 10 of the 11 used mouthwash, including nine who used a product containing about 27 percent alcohol. While a case series such as this raises concern, it cannot be evaluated with the criteria of causality appropriate to a formal investigation.14
The report6 also described a case-control study composed of the 200 patients with OPC and 50 general surgical patients serving as controls. These results were positive only when the case group was restricted to the original 11 patients. Although no overall RR was provided, the authors reported that the case-control study results were not statistically significant. Thus, the case-control study does not appear to add to any possible concern raised by the case-series report.
In 1983, Blot and colleagues9 recontacted subjects in an earlier case-control study of OPC among women in the southern United States.15 The original investigation had focused on use of tobacco and ACBs and did not address mouthwash use. It identified 255 case subjects (237 of whom were interviewed) and 502 control subjects (410 were interviewed). For all forms of OPC combined, the study reported RRs typically of 2.0 to 4.0 among women who dipped snuff, about 3.0 for women who smoked and about 5.0 for women who consumed alcohol.
The 1983 follow-up study9 was designed to address the ACM-OPC relationship. It was based on a five-minute telephone interview with 206 case subjects and 352 control subjects (or their next of kin), or 81 and 70 percent, respectively, of the original subjects. The findings were mixed, with an overall RR of only 1.2 for mouthwash use. The authors found no association between mouthwash use and OPC among tobacco users. However, and curiously, among women who used no tobacco, the RR of OPC for mouthwash use was 1.9 (not significant). Several of the findings were inverse (for example, an RR of 0.5 for users of full-strength [undiluted] mouthwash when based on information from proxy interviews) or inconsistent (for example, the RR for the same exposure was 4.7 when based on information from the subjects themselves). The authors found no consistent dose-response relationship. The alcohol content of the products used was unknown, as were the reasons for mouthwash use. Findings for mouthwash use were not adjusted for use of ACBs, and no data were presented for non-drinkers. The study provides little support for an association between mouthwash use and OPC.
In 1983, Wynder and colleagues8 reported a hospital-based investigation of 571 patients with OPC and 571 control subjects. The study results were negative for mouthwash use and OPC among men. Among women (157 case subjects with OPC, 157 control subjects), the crude data indicated a moderate association (RR of 2.8) between daily mouthwash use and OPC. However, the results showed no relationship between OPC and duration of mouthwash use. No information was available about the alcohol content of products used or whether mouthwash was used to conceal tobacco or alcohol odors on the breath. Control of potential confounders eliminated any association between duration of use of mouthwash and OPC among women. The authors suggested that underreporting of ACB use or causes of OPC associated with edentia (that is, poverty, smoking, alcohol abuse) had produced the positive finding in the crude data.
In a 1997 study, the relative risk of developing oral epithelial dysplasia varied inversely with the percentage of alcohol in the mouthwash used.
In 1985, Mashberg and colleagues16 evaluated the ACM-OPC hypothesis among 95 men with OPC and 913 men serving as control subjects. All had been patients in a New Jersey Veterans Administration, or VA, hospital. The authors found a weak, inverse association between mouthwash use and OPC. RRs for users were 0.8 in the overall data and 0.9 after controlling for smoking or ACB consumption. Among mouthwash users, the authors found an inverse association between OPC and the alcohol content of the mouthwash used; the RR was 0.6 for users of the product with the highest alcohol content. This study is useful but its generalizability is limited. VA hospital patients typically are heavy smokers and consumers of ACBs, as was evident in this study’s control group. These high consumption patterns do not invalidate this study because case and control subjects were selected from the same hospital. Nonetheless, the absolute risk of OPC may be so high among VA hospital patients because of their smoking and drinking that a modest increase in risk from mouthwash use, if such were to exist, might be missed.
In 1986, Young and colleagues17 described a multihospital-based case-control study of 317 OPC cases. The study included 306 control subjects who had a cancer of the head and neck "not thought to be related to tobacco use" or cancer of the larynx. The RR associated with mouthwash use was 1.0 among men and 0.5 among women. Findings were similar for cancer of the mouth and for oropharyngeal and hypopharyngeal cancer. No information was available about the alcohol content of the products used. The study results were distinctly negative, but issues may be raised about the study’s design (for example, combining the first control group, which had a presumably typical smoking pattern, with the second control group [cancer of the larynx], which probably had a high level of smoking) or its interpretation (for example, the study’s negative findings for mouthwash use despite the fact that neither smoking nor drinking was controlled).
The 1989 study by Kabat and colleagues18 was a multihospital-based investigation of 125 case subjects and 107 control subjects. It was a follow-up to the 1983 study by Wynder and colleagues,8 and so was restricted to women. The authors found no association between mouthwash use and OPC, but no information was available regarding the alcohol content of the products used. This study provided an important finding for interpreting all studies of mouthwash use and OPC. Moderately strong associations (RRs for OPC of 2.6 to 3.2) existed among women who used mouthwash to disguise breath odors of tobacco or alcohol. However, the RR was only 0.7 or 0.8 among women who used mouthwash to conceal food odors or for other reasons. The findings of the 1983 study by Wynder and colleagues8 and of this study by Kabat and colleagues18 indicate that positive results in overall, or crude, data are potentially confounded and, especially, that controlling for the reasons for mouthwash use is important.
In 1997, Morse and colleagues19 described a case-control study of "oral epithelial dysplasia" among 127 people identified from two large oral pathology laboratories. The 127 control subjects were pair-matched to the case subjects for age, sex and referral source to the pathology laboratory. People with oral epithelial dysplasia have an increased risk of developing OPC, and an earlier report from this investigation19 had shown both smoking and ACBs to be important risk factors for oral dysplasia. In the 1997 report, the authors examined eight variables describing mouthwash use and the alcohol content of the products used. Overall findings were negative, as were those for all eight variables. In fact, as Mashberg and colleagues16 had found, the RR varied inversely with the percentage of alcohol in the mouthwash used. The findings were unchanged after the authors controlled for smoking and ACB use. The results of this study are convincingly negative; they show no relationship between mouthwash use and oral epithelial dysplasia.
In 2001, Winn and colleagues20 reported findings from a large (342 case subjects, 521 control subjects) population-based study of OPC among people diagnosed in Puerto Rico between December 1992 and February 1995. This study appears to have been generally well-conducted, but the authors acknowledge that they were unable to "evaluate the accuracy of reporting of tobacco, alcohol or mouthwash use." They found no association between mouthwash use and OPC. Both the crude and adjusted RRs were 1.0, and there was no evidence of a dose-response effect for any of several measures of mouthwash use. Findings were positive for mouthwash use (RR = 2.8) among nonsmokers who abstained from alcohol. The authors considered these subjects to be the most likely to demonstrate any actual effect of mouthwash use on OPC. Such a consideration is speculative at best, especially in view of the strong interaction between alcohol abuse and smoking in the production of OPC. Perhaps more important, the extreme imprecision of the RR of 2.8 (95 percent confidence interval, 0.8–9.9) indicates that chance is a highly credible explanation for the findings.
We have described eight studies. The results of two of these were reported as positive, but one of these studies6 may have been subject to selection bias, while the findings of the other study9 were sufficiently inconsistent as to appear unreliable. The results of the study by Winn and colleagues20 were essentially negative, but the study included a nonsignificant positive finding among people who neither smoked nor drank alcohol. The results of two other studies were positive in their crude data, but became negative when the authors controlled for confounding by risk factors for OPC8 or for reasons for mouthwash use.18 The results of two of the studies were clearly negative, but the studies may have had limitations that make them unpersuasive17 or that limit the generalizability of the findings.16 The results of another study19 were convincingly negative, but they related to epithelial dysplasia and not directly to OPC.
|
THE NATIONAL CANCER INSTITUTE STUDY
|
|---|
Investigators at the National Cancer Institute, or NCI, and at four cancer registries in the United States conducted a large, population-based case-control study of OPC. The purpose of the study was to describe in quantitative terms the effects of known and strongly suspected causes of OPC. The study also attempted to identify new causes of the disease. Cases were diagnosed between January 1984 and March 1985. Control subjects younger than age 65 years were selected via random-digit dialing of telephone numbers, while older control subjects were selected from rosters of the Health Care Financing Administration (Medicare). The authors identified 1,485 case subjects and 1,668 control subjects. We consider two of the publications from this study.
The first report21 was based on interviews with 1,114 case subjects and 1,268 control subjects, about 75 percent of the subjects identified. It focused on smoking and ACB consumption, and both variables were found to be independently and synergistically associated with OPC. These associations were statistically significant. An exceptional finding was the absence of increased risk of OPC among wine drinkers, except among subjects who consumed 30 or more drinks of wine per week. Even for this group, RRs were much lower (about 2.0) than they were (about 10.0) for subjects who consumed 30 or more drinks per week of other ACBs. This finding suggested to the authors that "it is possible that ingredients in alcoholic beverages other than ethanol are involved" in increasing OPC risk.21
A subsequent publication by Winn and colleagues in 199111 primarily addressed findings regarding mouthwash use. In comparison with the first report, the number of case subjects was reduced by 22.3 percent (from 1,114 to 866), while the number of control subjects was reduced by only 1.5 percent (from 1,268 to 1,249). After adjusting for tobacco use, ACB consumption and other factors, the authors found that the RR of OPC associated with "regular use" of mouthwash at least one year before the interview was 1.4 for men and 1.6 for women; each finding is marginally statistically significant. Dose-response relationships were evaluated for the age at which mouthwash use was started, as well as for duration and frequency of use. None of these variables was consistently associated with risk of OPC, either among men or women. The study results suggested higher RRs with earlier age at beginning mouthwash use and with use of products containing 25 percent or more alcohol. However, use of diluted mouthwash was not associated with lower risk.
The authors also presented results for several conditions suspected of being associated with OPC.11 An increased RR was found only for leukoplakia. The authors found a reduced RR of 0.6 for OPC among men and women who reported having had dental radiographs.
Several aspects of the NCI’s 1991 report11 warrant mention. First, although the study is large, the information about mouthwash use is limited and may have been available only for self-selected subjects. Only 866 (58 percent) of all the case subjects identified are included in the analyses relating to mouthwash use, while the comparable figure for control subjects is 1,249 (75 percent). The major reason for losses among the case subjects is the exclusion of proxy interviews, but there were nonrespondents among both case and control subjects. The authors stated that the exclusion of proxy interviews resulted in little change in the findings regarding mouthwash use. However, even a little change could be meaningful for RRs that are only 1.4 and 1.6 after exclusions of 25 to 42 percent of subjects.
It is widely believed that the early signs and symptoms of oropharyngeal cancer often are treated by the affected people themselves.
Second, a "regular user" of mouthwash included a person whose use could have occurred solely, or largely, within just a few years before the interview (excluding the immediate one-year period) (not one year before the diagnosis). This is problematic, because some of the mouthwash use reported by patients may have been stimulated by the early signs or symptoms of an incipient OPC. It is widely believed that the early signs and symptoms of OPC often are treated by the affected people themselves.
Third, information was not available regarding subjects’ reasons for mouthwash use. This may reflect the fact that subjects were interviewed before the 1989 publication by Kabat and colleagues,18 which indicated that the reason for mouthwash use was potentially a crucial confounding factor.
Fourth, the RR of 0.6 for OPC associated with having had dental radiographs is equal in strength to a positive RR of 1.7. That is, the inverse association with radiographs is slightly stronger than the overall RR of about 1.5 for mouthwash use. This inverse or "protective" association between dental radiographs and OPC almost certainly is due to the higher economic status of the control subjects than that of the case subjects. Surprisingly, however, the association persisted after controlling for education and other factors. The reason for the persistence of this non-causal association is unclear, but it indicates that, despite its strengths, the NCI study can produce noncausal associations with OPC as strong as, or stronger than, those reported for use of mouthwash.
|
A REASSESSMENT OF THE NATIONAL CANCER INSTITUTE STUDY
|
|---|
After publication of the report by Winn and colleagues,20 an oral pathologist reviewed the histologic appearance of the subjects’ diseases. He found that among men, the histologic appearance of 38 (6.6 percent) of the tumors was nonmucosal, while among women, the histologic appearance of 37 (12.6 percent) of the tumors was nonmucosal. The 75 nonmucosal cases consisted of 11 adenocarcinomas, 12 adenoid cystic carcinomas, 17 mucoepidermoid carcinomas, 13 sarcomas and 22 cases of Hodgkin’s disease and other lymphomas.
Dr. William Blot, one of the NCI investigators, provided us with a copy of the data set, which included information about the relevant variables for the mucosal and the nonmucosal cases. Hereafter, we refer to the mucosal cases as true cases of OPC and the other cases as pseudo cases. We have repeated the study’s major analyses for all cases combined, as well as separately for the true and pseudo cases. We included all control subjects of the appropriate sex in each of the three analyses. Both smoking and ACB consumption were controlled using the same statistical model as that used in the original analyses.
We describe the comparison of findings for the true and pseudo cases as a specificity analysis. It is based on the assumption that a real cause of OPC will be associated more strongly with true OPC than with the heterogeneous conditions that make up the pseudo cases. There is no reason to expect the pseudo diseases to have any significant part of their etiology in common with one another or with OPC. We evaluated this assumption first by comparing the true and pseudo cases with respect to their demographic features and their associations with smoking and ACB consumption. For reference, the analyses include "all" cases, but our focus was on a comparison of the true cases with the pseudo cases.
Table 1
shows that the true and pseudo case groups are quite different with respect to their demographic profiles, especially in their associations with smoking and ACB use.
TOP
ABSTRACT
GENERAL CONSIDERATIONS
