Minutes

ITEM 1: APOLOGIES FOR ABSENCE/ANNOUNCEMENTS

2. Apologies for absence were received from COC member Professor Harrison, Professor Boobis (DH Toxicology Unit) and the DEFRA assessor Mrs Cameron.

Announcements

3. The Chairman welcomed Professor Elliot and Dr Key from Imperial College, Professor Jacobs (University Medical School, London) Professor Davies, Dr T England, Dr Pufulette (DH Toxicology Unit at Imperial College), Dr Gupta (DH), Ms Hind (DH) and Mr Strickland (Cabinet Office).

4. Members were reminded of the need to make any relevant declarations of interests before discussion of items.

ITEM 2: MINUTES OF THE MEETING ON 27TH JUNE 2002 (CC/MIN/02/2)

5. The minutes were agreed subject to minor editorial changes.

ITEM 3: MATTERS ARISING

3.1 Response to COC statement on the investigation of interaction between chemicals in the environment on the induction of cancer (CC/02/29)

6. The Secretariat has received a letter from Dr Zimmern (Director of the Public Health Genetics Unit, Cambridge) in which a number of initiatives were suggested where the COC statement on genotype and environment chemicals might be relevant. The Committee was asked to comment on points raised in the letter. In particular, the Committee considered suggestions that the UK Biobank Project and the proposed Genetic Research Parks may be used as resources for studying these types of interactions. A background note on the proposed UK Biobank Project was provided by the Research and Development Division of the Department of Health.

7. Members agreed that considerable planning had gone into developing the UK Biobank Project and that it was potentially a valuable prospective cohort study for investigating lifestyle factors and disease. However, members felt that the value of the project specifically with regard to studying interactions between genotype and chemicals in the environment was less clear, and this would be determined by the information gathered on exposures to chemicals at recruitment. Members agreed that it was important to obtain good information on exposure and felt that appropriate questionnaires would allow information to be gathered on lifestyle activities (eg smoking, drinking alcohol), but considered that this would not provide data on environmental exposure and might not adequately cover occupational exposure to chemicals. Members restated that a priori hypotheses were essential to allow for adequate and scientifically acceptable analyses of the findings.

8. Members discussed the use of biomarkers of exposure as an aid in assessing exposure to chemicals, however this would depend on the quality and storage of samples and care would be needed with potentially unstable biomarkers. It was also suggested that subjects should ideally be revisited to determine additional details of chemical exposure and possibly to obtain further blood samples, however the Committee noted that this was unlikely to be practical.

9. Members also suggested that such studies should ideally include information on phenotype variation, however the Committee is aware that it is unlikely such data would be available for the candidate genes selected. In addition, with the advancement of DNA sequencing techniques, many genes will be discovered in the future that would be highly relevant for this type of study, however appropriate storage of samples would be needed for retrospective analysis. The Committee did note that similar studies have used procedures whereby blood samples were fractionated to allow for diverse measurements, and this should be considered. Members also anticipated an additional problem, which involved the willingness by external laboratories to analyse the large number of samples generated. The Committee encouraged both projects to consider establishing their own analytical laboratories.

10. The Secretariat was asked to respond on behalf of the Committee. A draft letter would be circulated to members for agreement.

3.2 Malathion

11. The COC was informed that the COM had considered some new mutagenicity data at its meeting on 10th October 2002. The draft joint COM/COC statement had been sent to the data holder for comment. The draft joint COM/COC statement had also been sent to the Advisory Committee on Pesticides for initial consideration at the meeting on 28th November 2002. The Medicines Control Agency would consider the need for action when the joint COC/COM statement had been finalised.

ITEM 4: ALCOHOL AND BREAST CANCER

12. The minutes of this item now published - see paragraphs 35 - 62.

ITEM 5: REVISION OF COC GUIDELINES: 1ST DRAFT (CC/02/26)

13. The COC had considered mathematical models, use of potency estimates, use of threshold of regulation concept, and pragmatic minimum risk levels at its previous meeting (June 2002). Comments regarding these issues have been used in the writing of the draft Guidelines (Annex 1 of CC/02/26). The draft guidelines were written in such as way as to give a brief outline of hazard identification, hazard evaluation and exposure assessment, but the focus was on risk estimation.

14. Members considered the proposed use of the one-hit model to set upper and lower bounds of risk at low levels of exposure on the basis of human data from epidemiology. Members were advised that the intention was to use this very conservative approach to advise on the practicality of risk management options for specific air pollutants and there was no intention to publish risk estimates based on this approach. The primary objective would be to assess the cost of reducing air levels to the exposures associated with the upper bound estimate of risk based on the one hit model. The COC was content with the approach provided it was limited to chemicals for which there was good cancer epidemiology data and that risk estimates were used only as a guide when considering risk management options. Members felt it important to restate that extrapolation of risk estimates below the observed range was very problematic, as no model was completely satisfactory. Members suggested that this specific use of mathematical models could be addressed in a separate statement, rather than in the guidelines.

15. Members were generally content with the approaches and balance of information within the guidelines. When the recommendations were considered it was suggested that biomarkers of effect, in addition to biomarkers of exposure, should be researched to aid the extrapolation of low doses.

16. The Committee did not believe that modifications of the bioassay study design eg by optimising number of animals in dose levels, would present any practical way of improving knowledge of the dose response at low levels that would be of value in risk assessment. It was agreed that reference to this should be deleted.

17. Members considered a rationale for the use of 10,000 as a safety factor (Gaylor et al. (1999) Regulatory Toxicology and Pharmacology 29:151-157) in order to determine acceptable/tolerable levels of human exposure to both genotoxic and non-genotoxic carcinogens. Members felt that the rationale was not supported by scientific analyses of the data. Therefore, it was suggested that minimum risk levels should be determined on a case-by-case basis for genotoxins and a margin of safety approach should be used for non-genotoxins with the appropriate value again being set on a case by case basis.

18. Threshold of regulation as a method for setting regulatory minimum risk level for carcinogenic contaminants was discussed at the June 2002 meeting. Although, in general, members felt that the approach would have limited use, as the proposed exclusions covered the major classes of genotoxic carcinogens and also the non-genotoxic carcinogen TCDD, the Committee agreed to reconsider the threshold of toxicological concern (TTC) alongside the draft COC guideline document. Members suggested that the threshold of regulation method should not be used as a generic approach. However, as it is based on ranking by theoretical risk and exposure, it could be used for preliminary prioritisation of chemicals prior to the provision of carcinogenicity or mutagenicity data. The Committee was made aware of a report by the International Life Sciences Institute (ILSI) to refine the TTC approach, which should be available for the next COC meeting. Members agreed to review the TTC approach at this time.

19. Members were asked to comment on the evidence of non-linearity concerning exposure-response regarding 2-AAF and diethylnitrosamine in the appended papers (Williams (2000) Toxicologic Pathology 28:388-395). Members commented that the weight of evidence supported a linear dose response for DNA adducts after exposure to genotoxic carcinogens.

20. Members considered a population-based study, which investigated lung cancer risk associated with occupational exposure to asbestos (Gustavsson et al. (2002) American Journal of Epidemiology 155:1016-1022). Members observed the wide confidence intervals of estimates and agreed that many different models would fit these data. Members reiterated the lack of precision concerning mathematical modelling.

21. Members heard from a representative of the Medicines Control Agency, who suggested that the guidelines should make appropriate reference to The International Conference on the Harmonisation of Technical Requirements for the Registration of Pharmaceuticals for Human Use (ICH).

22. The Chair thanked members for their comments. It was agreed that the secretariat and the DH toxicology unit would use comments raised in the discussion, as well as written comments by members, to revise the COC draft guidelines on the evaluation of chemicals for carcinogenicity.

ITEM 6: POLY AROMATIC HYDROCARBONS: DIBENZO(a,l)PYRENE (CC/02/27)

23. In 1995, at the request of MAFF and the Department of Environment, the COC agreed a hazard-ranking scheme for the carcinogenicity of 25 polyaromatic hydrocarbons (PAHs). It was based on classification into one of 5 categories. The COC in 1995 had accepted the principle that the carcinogenicity of PAHs was additive. Advice on dibenzo(a,l)pyrene had not been requested in 1995. Since then, air pollution monitoring in the UK had detected the presence of dibenzo(a,l)pyrene in a number of samples. An assessment of the relative carcinogenic potency of dibenzo(a,l)pyrene compared to benzo(a)pyrene (claimed to be about 100 times) had been published in a recent Environmental Health Criteria Document (International Programme on Chemical Safety).

24. The COC was asked to consider the mutagenicity and carcinogenicity data on dibenzo(a,l)pyrene and to consider to what category of the COC hazard ranking scheme of PAHs benzo(a,l)pyrene could be assigned. Additionally the committee was asked whether the relative carcinogenic potency of dibenzo(a,l)pyrene compared with benzo(a)pyrene could be determined from the available data and in particular whether it was possible to make any statement on the relative potency by the inhalation route of exposure.

25. Members agreed that the in-vitro mutagenicity tests and information on in-vivo DNA adduct formation was consistent with dibenzo(a,l)pyrene being an in-vivo mutagen. Members also agreed that dibenzo(a,l)pyrene was carcinogenic in mice and rats. Dermal application to mice produced tumours at a number of sites (including the skin, lung, and malignant lymphoma of the spleen and malignant lymphoma with multiple organ involvement) and intraperitoneal administration to rats produced lung tumours. Intramammary instillation in rats resulted in mammary tumours. Dibenzo(a,l)pyrene also acted as an initiator in mouse skin carcinogenicity promotion assays. The COC therefore considered that dibenzo(a,l)pyrene should be assigned to group A of its hazard ranking scheme for PAHs. This category includes chemicals for which 'there is a high level of concern about a carcinogenic hazard for humans because the compound is an in-vivo mutagen and/or a multi-site carcinogen in more than one species'.

26. Regarding potency, the committee agreed that dibenzo(a,l)pyrene was a very potent genotoxic carcinogen and that potency varied depending on factors such as species, route of administration, dose and site of tumour produced. From the available data where a comparison could be made, members considered that the dibenzo(a,l)pyrene carcinogenic potency was likely to be in the range of 10-100 times more potent than benzo(a)pyrene depending on the tests system used. This was consistent with the available data and DNA adduct formation in-vivo. It was noted that the in-vitro data showed dibenzo(a,l)pyrene produced 6-9 times more DNA adducts than benzo(a)pyrene, but that it was less mutagenic in the Ames test.

27. Members were made aware of a suggestion that the coefficients from fitting tumour dose-response curves derived from intraperitoneal administration studies with mice could be compared, which would give a relative potency of about 20- 80 times for dibenzo(a,l)pyrene compared with benzo(a)pyrene.

28. Members agreed that there were insufficient data available on dibenzo(a,l)pyrene to draw any conclusions on the relative potency compared to benzo(a)pyrene by the inhalation route of exposure. The committee was made aware of an evaluation undertaken for Health Canada, which attempted to identify route specific relative carcinogenic potency of polycyclic aromatic hydrocarbons (PAHs). The committee agreed to reconsider the topic of the relative carcinogenic potency of PAHs at a future meeting. A statement on dibenzo(a,l)pyrene would be drafted for consideration at the next meeting.

ITEM 7: IMPLEMENTATION OF OST GUIDELINES: COC TEMPLATE AND MEMBER PROFILES (CC/02/28)

29. The COC considered the OST code of practice for Scientific Advisory Committees at its March 2002 meeting. The COC complies with most of the requirements of the code. The COC already publishes detailed statements, agendas, minutes and substantive background papers (where appropriate).

30. One area of future work concerned the development of a 'Membership Template', which covers the balance of expertise on the committee. A draft COC template and short examples of member profiles used by the sister committee the COT were attached as Annex 1 to CC/02/28 for consideration. The template gave a pictorial view of how the COC undertakes risk assessment and interacts with the COM/COT, other Government Departments and CMO.

31. The secretariat would only place member profiles on the COC website when agreed. It was suggested to introduce member profiles when the new COC committee is appointed in 2003. A finalised document would be placed on the COC website in due course.

ITEM 8: PAPERS FOR INFORMATION

32. The following papers were made available for information.

8.1 Do estrogens always increase breast cancer risk? By Hilakivi-Clarke et al (2002), Journal of Steroid Biochemistry & Molecular Biology (80), 163-174 (CC/02/31).

8.2 Preferential DNA damage and poor repair determine ras gene mutational hotspots in human cancer, by Feng et al (2002), Journal of the National Cancer Institute (94), 1527-1536 (CC/02/31).

ITEM 9: ANY OTHER BUSINESS

9.1 Reappointment of COC

33. The committee was informed that an advertisement for new members of the COC had appeared in the New Scientist.

ITEM 10: DATE OF NEXT MEETING

34. Date of next meeting 6th March 2003.

************************

ITEM 4: ALCOHOL AND BREAST CANCER

4.1 META-ANALYSIS (CC/02/24)

35. The COC had previously reviewed studies on the association of breast cancer and alcohol consumption up to 1995, and commissioned a further review from the Department of Health Toxicology Unit from 1995-1999. This work found no conclusive evidence that breast cancer is causally related to alcohol, but the significance for public health of even a weak association was such that this matter should be kept under review. Subsequently a meta-analysis had been commissioned, to obtain more definite information and to assist in the assessment of causality.

36. The committee heard a presentation by Dr Jane Key and Professor Paul Elliot from Imperial College Department of Epidemiology and Public Health on the final report of the meta-analysis project.

37. Dr Key noted that breast cancer is the most common cancer in women. The established risk factors were predominantly reproductive where there was little scope for prevention. She noted that the five previous meta-analyses of alcohol and breast cancer had all reported a small association (RR = 1 to 2) and there was a new reanalysis of cases from published and unpublished epidemiological studies recently published by the Oxford Collaborative Group on Hormonal Factors in Breast Cancer (British Journal of Cancer (2002), vol 87, 1234-1245). Dr Key agreed with COC that a small association, if causal, would have significant implications in terms of number of attributable cases from drinking alcohol.

38. The objectives of the meta-analysis undertaken by Imperial College had been to determine the magnitude of any association between drinking alcohol and primary breast cancer, to explore the dose-response relationship, to examine whether any association was related to specific beverages or to consumption of all alcoholic beverages, to explore possible heterogeneity, bias and confounding and to estimate the population attributable risk. All publications, in any language, between January 1st 1966 and 31st December 1999 were eligible. The results from studies were examined after data on study design and methods had been abstracted and reviewed independently by two members of the team. Duplicate studies were not included. A simple scoring scheme was used; suboptimal design (1), good design but insufficient control for confounding (2), good design and adequate control of confounding (3). Analyses were undertaken for least adjusted, age adjusted and multivariate adjusted data for all reports and those scoring 2 or 3 and finally multivariate design scoring 3. Dose response modelling used standardised exposures (converted to g/day), the mid-point estimates for consumption, and a linear model with a variable intercept and meta-analysis of dose-response using the random effects model.

39. A total of 254 papers were identified. Data from 87 were included in the meta-analysis which related to 77 unique studies and 2 combined studies. Difficulties in assessing the measurement of alcohol intake and the definition of non-drinkers were acknowledged. The number of studies that provided data that could be included in the ever versus never analysis was 66 and was based on 54,653 cases. Using all these studies and least adjusted odds ratios, the relative risk associated with drinking alcohol was 1.14 (95%CI 1.08-1.21). But combining multivariate odds ratios from studies scoring 2 or 3, the relative risk associated with drinking alcoholic beverages was 1.20 (95%CI 1.11-1.29). There were 34 studies that contributed to this analysis. The use of variable intercept dose-response modelling allowed for the presence of light drinkers/exdrinkers in the referent group and also avoided the assumption that if a linear dose-response relationship existed then it would be linear through the origin. There were 45,791 cases (56 studies) that could be included in the least adjusted pooled analysis, but the results quoted were for the analysis of most-adjusted odds ratios from studies scoring 2 or 3, and that only includes 38,710 cases (41 studies). It was reported that the relative risk of drinking 2 drinks/day compared to 1 drink/day was 1.11 (95% CI 1.07-1.16). There was no evidence for a stronger association with any particular beverage.

40. Dr Key noted that analysis for the heterogeneity had included data collected before and after disease onset, hospital and community controls, evidence in pre and post menopausal studies and influence of nationality of study population. It was noted that there was a significant difference in relative risk between studies using hospital or community controls but the association between drinking alcohol and breast cancer was still statistically significant in studies using either of these control groups. Meta-regression analyses suggested there was no evidence for significant differences between pre-and post menopausal women, but the possibility of publication bias or misclassification couldn’t be excluded from the analysis. Ever versus Never analyses indicated a slightly higher relative risk for retrospective compared to prospective analyses, and for country of origin (Europe versus USA and Canada). Overall, the analyses performed did not explain all of the potential heterogeneity between studies.

41. A Population Attributable Risk for England had been calculated using Cancer Statistics for England (1998) and information on drinking patterns from 1993 and 1998. Assuming causality and an ethanol content of 9.5g in an “average” drink, the PAR based on 1998 data was 7.9% (ie the fraction that could be prevented if drinking had been reduced to a very light level). A further analysis suggested that if women reduced their alcohol intake to 2 units/day then the number of breast cancer cases would be reduced by around 500.

42. Dr Key summarised the main points of the study. This was the largest and most comprehensive meta-analysis. The use of simple scoring system allowed examination of study quality, further analyses had been undertaken to examine for bias and confounding. It had been established that all analyses provided significant positive associations. The investigators had acknowledged that the definition of non-drinker, use of mid point estimates of alcohol consumption and aggregate (study) data instead of individual had limited the evaluation. However overall, it was felt that the association was not a chance finding.

43. Members agreed that the Imperial research team had undertaken an excellent piece of work and encouraged publication.

44. Members queried the consideration of potential differences between pre- and postmenopausal women. Members noted there was some evidence that the effects of alcohol in women were age related. It was noted that the data in table 11 (in Annex 1 to CC/02/24) suggested a significantly higher risk for postmenopausal women compared to premenopausal women. Dr Key replied that the analysis in Table 11 was an ecological analysis using the proportion of study subjects who were premenopausal as a covariate and is not directly interpretable and suggested that this table was removed from the document. Examination of table 10 in Annex 1 to CC/02/24 showed no statistically significant differences between pre- and postmenopausal women when data from all relevant studies (n=57) and those considered to be of adequate quality (scores 2 and 3, n=54 studies) were compared.

45. The Committee agreed that the evaluation of dose-response was difficult, particularly at higher levels of alcohol drinking where there was comparatively fewer data available. Members agreed the rationale stated for adopting a variable intercept model. It was noted that the available mechanistic data supported the possibility of a threshold for carcinogenesis.

46. The Committee considered that the estimate of Population Attributable Risk (PAR) was potentially one of the most important outcomes of the meta-analysis with regard to presentation of the public health significance of the analyses. The estimated value based on 1993 intake data was 7.04% The estimate using 1998 data was 7.9%. The range of estimates based on the seven analyses was 2.59% to 12.09%. For the smallest estimate of PAR, results from the seven dose-response analyses were evaluated and the smallest lower 95% confidence interval in the PAR calculation was used. For the largest estimate of PAR, the biggest upper 95% confidence interval was used. Members noted the apparently large range for the estimate of PAR. Members queried whether the inclusion of non-drinkers would significantly affect the result. Members were informed that analyses of using the variable (non-drinkers excluded) and zero intercept models (ie non-drinkers included) suggested there was no significant alteration in PAR when non-drinkers were excluded.

47. Members recalled their evaluation of Ellison et al (Am J Of Epidemiology vol 154, (8), 740-747, 2001) and considered that a further evaluation of the data to consider the influence of length of follow-up in prospective studies would be useful.

48. Members were concerned that there were no data to assess the potential for binge drinking on risk of breast cancer and agreed this should be noted in the statement.

49. The Chair asked Members to consider the evaluation published by the Oxford Collaborative Group (British Journal of Cancer, vol 87, 1234-1245, 2002) which had been forwarded separately to members. It was noted that the Oxford Collaborative Group had access to individual data from 58,515 women, including some from unpublished studies. They had been able to determine median intakes of alcohol. The dose response data reported in table 2 suggested some evidence for a threshold below a median intake of 8 g/day. Overall it was felt that the Oxford Collaborative group had considered virtually the same cases as the Imperial team. However it was unclear why the number of non-drinkers was 36% in the Oxford Collaborative study whereas the Imperial team had reported the percentage of non-drinkers to be 28.6%. The estimate of PAR reported by the Oxford Collaborative group was 4% compared to 7.9% for the Imperial research team. Members considered that further examination of apparent differences in PAR estimates between the two groups was important. Areas for consideration included the rates of breast cancer assumed, the relative risk estimates used, and data on intake of alcohol. Overall it was unclear from the data presented to the Committee how the calculation of PAR had been undertaken by the Oxford Collaborative Group.

50. Members considered the approach used to determine Cumulative Incidence of breast cancer per 100 women with age at 2, 4 and 6 drinks per day was a useful method of presenting risks (cf in figure 5 of the Oxford Collaborative Group evaluation). However it was felt that the confidence intervals for some of the determinations, particularly at high levels of drinking would have been very wide.

51. The Committee agreed it was important to obtain some further details from the Oxford Collaborative Group before deriving any conclusions.

4.2 Update on mechanisms (CC/02/25*)
*(CC/02/33 - Supporting papers for CC/02/25)

52. The COC reviewed the available published evidence regarding mechanisms by which alcohol might induce breast cancer in 1995 and 1999. Members reached a number of conclusions in 1999 after receiving advice from Professor Howard Jacobs (Emeritus Professor of Reproductive Endocrinology, University College Medical School, London) on the endocrinological aspects. Overall the available data had suggested a plausible mechanistic link between consumption of alcohol and breast cancer mediated via an effect of alcohol on hormones. The interpretation of these data was particularly complicated and difficult; for example, the influence of confounding effects of other possible breast cancer risk factors such as obesity, use of oral contraceptives and hormone replacement therapy and their potential interaction with drinking alcoholic beverages needed to be considered carefully. The Committee had agreed that further epidemiological work should consider a number of sub-groups, ie pre menopausal women who either used or did not use oral contraceptives and postmenopausal women who had or had not taken HRT. The Committee agreed in 1999 that there were insufficient data available to describe a threshold of action for alcohol-induced elevation in oestrogens.

53. The COC looked at a number of options for potential research to investigate the effect of alcohol on oestrogens; ie use of stored samples from the Health Survey for England and a scoping study for an intervention study. None of these suggestions were taken forward because of logistical problems in designing a study, potential ethical problems in administering alcohol at the recommended safe drinking limit for at least 1 month and possibly up to 3 months, and estimated costs of the study.

54. The DH Toxicology Unit (based at Imperial College of Science Technology and Medicine) had prepared a paper, which presented a review of studies published after the previous review in 1999. The authors highlighted evidence to support the view that alcohol induced hyperinsuliaemia and increased Insulin-like Growth Factors (IGFs) which subsequently induced an increase in breast tissue density through increased cell division. It was noted that there were additional studies to support an association between drinking alcohol and effects on oestrogen metabolism. There was considerably less data for a number of suggestions such as alcohol induced suppression of melatonin and aromatisation of androgens to oestrogens.

55. The Chair asked Professor Jacobs for any comments on the new data. Professor Jacobs made two observations on the available data. Firstly, in order to validate the oestrogen induction of breast cancer hypothesis, many investigators had tried to differentiate between oestrogen receptor positive and negative tumours, since evidence for raised oestrogen levels and oestrogen positive tumours would support the oestrogen induction hypothesis. Professor Jacobs noted that although the data on oestrogen receptor positive tumours was inconsistent, many studies had used outdated radiolabel methods for detection of oestrogen receptors. Overall he thought the evidence supporting the suggestion that drinking alcohol could induce breast cancer by raising oestrogen levels was plausible. He noted that a potential discrepancy related to whether the risk of breast cancer associated with oestrogen changed with age. He suggested that drinking alcohol might have a number of effects on biochemistry which could act together to raise the risk of breast cancer. Professor Jacobs considered the evidence to support the association between drinking alcohol and raised IGF1 levels and lowered levels of IGF-binding proteins was biologically relevant (Byrne C et al, Cancer Research, vol 60, 3744-3748, 2000). He suggested that an effect of drinking alcohol on liver biochemistry could be associated with both raised oestrogen and IGF1 levels.

56. Members agreed with Professor Jacobs that the data on receptor status and breast cancer was very confusing. It was suggested that one uncertainty related to how drinking alcohol could induce a number of biochemical effects in the liver which resulted predominantly in a significant increase in risk of breast cancer. Members were aware that the causal association between drinking alcohol and head and neck cancers was based predominantly on the consistent evidence for a large dose response in epidemiological studies using a variety of approaches. However there was no adequate mechanistic explanation for alcohol induction of head and neck cancers.

57. Members agreed that evidence to support the association of alcohol with increased oestrogens had been documented in a number of studies (Martin C et al The American Journal on Additictions, vol 8, 9-14, 1999 and Muti P et al, Cancer Epidemiology, Biomarkers and Prevention, vol 7, 189-193, 1998) not previously reviewed by COC. Members reviewed the cross sectional study by Verkassalo PK et al (Cancer Causes Control vol 12, 47-59, 2001) and agreed that sufficient numbers of premenopausal (636) and postmenopausal (456) had been included. The results were inconsistent with the data previously reviewed by the committee in 1999 and suggested an effect of cigarette smoking but not drinking alcohol on levels of oestrogens. These results were not consistent with the available epidemiological data on breast cancer.

58. Members reviewed the study by Dorgan JF et al (J Natl Cancer Inst, vol 93, 710-715, 2001) in postmenopausal women and agreed that a satisfactory cross over design had been used for this intervention study, although there were some reservations regarding potential compliance of study participants. Members considered that the study fulfilled many of the criteria the COC had discussed when considering a proposal for a scoping study in November 2000. It was noted that there was some evidence for a small increase in oestrone sulphate and dehydroepiandrosteone sulphate (DEHA sulphate) following the consumption of 15 g or 30 g ethanol/day over an eight week period. There was no effect on oestradiol (free or bound) in this study. Members agreed the data supported a small effect of drinking alcohol on adrenal output of hormones. This study suggested the effect of drinking alcohol on hormone levels was milder than the cross sectional studies previously reviewed by the COC in 1999. It would be important to assess all the evidence when reaching conclusions in a statement.

59. Members considered that the evidence supporting other proposed mechanisms was preliminary and no conclusions could be drawn. A number of observations of interest were noted. Thus the significantly increased relative risk in pre-menopausal women lacking GSTM1 and GSTT1 genes who drank alcohol (Odds Ratio = 5.3, 95%CI 1.0-27.8) was considered interesting with regard to the generation of hypotheses to study (Park SK et al, Pharmacogenetics, vol 10, 301-309, 2000). Members considered that further data on background incidence of tumours in ICR mice were required before an assessment of the study published by Watabiki et al (Alcohol Clinical and Experimental Research, vol 24, 117S-122S, 2000) but agreed that the administration of 10% and 15% ethanol solutions in the drinking water for 25 months would have resulted in excessively high doses.

60. The Committee agreed that it was appropriate to consider that drinking alcohol had a number of biochemical effects in the liver which may have resulted in changes to oestrogen and IGF levels which over a prolonged period of time, ie decades, could induce breast cancer. Members felt that both of these suggested mechanisms would potentially have a threshold with regard to induction of breast cancer. It was also possible that overt liver toxicity at high levels of drinking could reduce IGF production and thus lead to reduced stimulation of breast cell division.

61. The Committee discussed whether a conclusion could be reached with regard to causality and agreed that the new meta-analysis and information from the Oxford collaborative group indicated that the association between drinking alcohol and breast cancer was not a chance association. It was agreed that there was evidence for some residual confounding but overall the data suggested that many of the Bradford-Hill criteria had been fulfilled. There were still uncertainties regarding mechanism and whether the available evidence from studies of mechanisms could explain all of the features of the association between drinking alcohol and breast cancer (in particular the shape of the dose-response curve). The Committee felt it would be prudent to use a precautionary approach to risk assessment and asked that the estimates of PAR be further considered at the next meeting, where it was hoped further information from the Oxford Collaborative Group would be available. The secretariat were asked to prepare a draft statement for the next meeting but to clearly identify sections where additional information was required before definite conclusions could be reached.

4.3 Maternal alcohol intake in rats and Breast cancer in offspring (CC/02/30)

62. Members reviewed the preliminary data available from Dr Cabanes and colleagues (from Department of Oncology, Georgetown University, Washington, DC, USA) and agreed that a significant effect of in-utero alcohol exposure on tumour multiplicity had been reported but agreed that no conclusions on this study could be reached in the absence of a full report.

************************

ACTIONS

ITEM	ACTION	WHO
4. Alcohol & breast cancer	Draft statement	Secretariat
5. Revision of COC Guidelines	Draft guidelines	DH Toxicology Unit

Top

copyright: © | updated: 5 January 2005