Fluoridation Review - Minutes of Meetings

Advisory Panel Meeting - 4 May 2000

Chair: Professor Trevor Sheldon, Department of Health Studies, University of York.

Members of the Advisory Panel present: The Earl Baldwin of Bewdley, House of Lords, Dr. Iain Chalmers, UK Cochrane Centre, Dr. Sheila Gibson, Glasgow Homeopathic Hospital, Dr. Alan Glanz, Department of Health (observer), Ms. Sarah Gorin, Help for Health Trust, Professor MA Lennon, University of Liverpool, Dr. Peter Mansfield, Director of Templegarth Trust, Professor JJ Murray, University of Newcastle, Mr. Jerry Read, Department of Health, Dr. Derek Richards, Centre for Evidence-Based Dentistry, Professor George Davey Smith, University of Bristol

Members of the Review Team present: Dr. Ivor Chestnutt, Dental Public Health Unit, Cardiff, Professor Jos Kleijnen, CRD, University of York, Dr. Marian McDonagh, CRD, University of York, Dr. Elizabeth Treasure, Dental Public Health Unit, Cardiff, Ms. Penny Whiting, CRD, University of York, Mr. Paul Wilson, CRD, University of York

1. Apologies for absence

Ms. Pamela Taylor

2. Opening and Introductions

Professor Sheldon (Chair) made the opening remarks and the panel was introduced.

3. Review of draft report

The chair lead a discussion of each section of the review. Specific comments of panel members are listed below.

General Comments

Incautious language throughout (IC)
Opportunity to comment after this version (JM) - yes (TS)
Share unpublished papers with the panel (ML)
Panel Chair (TS) to read future versions before being distributed to panel (IC)
Make objectives consistent throughout document (IC, EB, agreed by panel)

4. Background

High level evidence only - why toxicology and animal studies (PM & EB) were excluded, give an example of errors using poor evidence (IC)
Ethical and legal issues - not part of review (SG)
1ppm commonly considered optimal (GDS) - maybe restate in relevant sections (TS)
Remit provided by the government and did not included economic aspect (EB)
State that there has not been a complete systematic review of the area (EB)
Discuss what fluoride compounds are used in the artificial fluoridation of water and what occurs naturally (PM)
State what was not included; refer to study design descriptions in glossary. Add more on why studies were excluded (i.e. cross-sectional). Include more on how levels of evidence were assigned. (TS)

5. Methods

Not enough methodological background to explain levels of evidence (A, B, C) (TS). Stress that inclusion criteria were set before studies were looked at (GDS)
Discuss observer bias and importance of blinding (likely to increase caries ascertainment in non-F areas, and fluorosis) (PM) as part of levels of evidence
Heterogeneity - make clear the difference between statistical heterogeneity and other heterogeneity. Does not make sense not to pool based on Q statistic (did actually pool anyway by using the meta-regression). Should show pooled statistic, Q statistic and how this is changed by the meta-regression (GDS)
Objectives do not match the purpose, especially objective 2 (EB & PM)
Ian Chalmers - heterogeneity statistics first?
IC said that the 'literature searches' (Section 3.1) should really just be called 'Searches' since not all of the data was published

6. Results

General

Remove +/- from results tables (IC and general discussion)
Do a sensitivity analysis by removing unpublished data
Put Forest plots sooner in the results
Sort all tables on validity (IC)
Remove vote counting from below the tables (IC)
Remove any sentences that appear to draw conclusions or make inferences (PM/EB)
Clarify what is meant by positive and negative associations (IC), investigate other possible ways of describing association
Negative effects should not include the word 'health' (ML)

Objective 1

Check/take into account co-linearity of age-teeth type (GDS)
Eruption time delay not taken into account (SG). This needs to be noted somewhere in the review (general discussion) - in confounding factors
Check year for Gray on forest plot (GDS)
Do a meta-analysis before the regression analysis (GDS)
Put Hardwick on the graph (IC and TS)
Discuss observer bias (likely to increase caries ascertainment in non-F areas) (PM)
Are before-after studies subject to confounding? If baseline caries risk is similar in all groups then does this mean that they are not subject to confounding, and what confounding factors could act during study period? (ML)
Note reasons for starting and stopping fluoridation (GDS) - may be possible confounding factor. Highlight those that are natural experiments (ET)
Use whether study was blinded as variable in meta-regression (ML)
Meta-regression should incorporate baseline risk of caries
Do univariate regression analysis (GDS), add table which shows the results of this and how much of the heterogeneity each study accounts for
Talk through the meta-regression more (TS)
Take about constant as it's not right (GDS and TS)
State that first paragraph of meta-regression addresses objective 2 (IC). Remove not stated tooth data as missing data (GDS). Separate out permanent and primary teeth analysis
Publication bias - debate as to what the funnel plot showed. Look at this section in more detail (general discussion)
Validity score should be included in both meta-regressions (IC)

Objective 2

State that there were no specific studies for objective 2 (general discussion)
Value for DMFS for Hardwick is wrong (ML)

Objective 3

The difference between relative and absolute difference (GDS)
Re-interpret results - the conclusions drawn from the data at the moment don't reflect the data (GDS)
In age-groups other than 5 years this association is not seen (GDS)
Define urban-ordinary (IC)

Objective 4

Fluorosis

Cross reference to Appendix A for fluorosis indices (TS)
DDE score may measure enamel opacities other than fluorosis (ET)
Move section on linear relationship/logs to below the graphs, and explain further (TS)
Odds for unit changes in fluorosis tables (GDS)
Change 'suffering from' fluorosis to 'with' (IC)
Remove methods of fluoridation from the model - misleading as only kept in model based on the 'not stated' category (GDS)
NNH - make equivalent NNT for caries (IC)
Investigate why we 0.4 was taken as the lower limit - investigate whether could take lower value e.g. 0.1 (PM, S. Gibson, EB)
Repeat regression analyses taking only the lower values (<1.5ppm) (GDS)
The fluorosis results need to be considered in the presence of other sources of fluoride
Look at the systematic review on toothpastes (IC) - either JK or ET will contact author
Section 4.5.1.3. - how studies were selected for inclusion on graphs - enough time for fluoride to have effect, only areas that would have had F toothpaste etc (ET)
Need to interpret regression analysis as an exploratory analysis - trust the plots and see if regression agrees. (GDS)

Bone studies

Move forest plot earlier
Use men and women not male/female (IC)
Use ethnicity throughout not race (GDS)
Try and trace Sowers to find study which associated individual fluoride intake
Change publication bias section to say that effect sizes are similar and so cannot detect publication bias. If Sowers is not locatable then this should be mentioned under this section

Cancer studies

Soften emphasis on blinding and explain how studies could be blinded (IC)
Mention thyroid cancer (IC)
Try and get more confidence intervals (GDS)
Provide definitions of summary measures in the text (GDS)
Indicate scales where appropriate (GDS)
Rank tables by outcome measure (IC)

Other negative effects

One of the Rapaport studies did make some adjustment for confounding this needs to be changed in the text (SG)
Discuss limitations of Rapaport study (ML)
Based on the results of the Down's syndrome and other congenital malformations/infant mortality studies suggesting some possible evidence of an association with water fluoride levels should we look at animal studies (EB). The general consensus was that the studies were too weak to provide enough evidence to go back to animal studies

Objective 5

The interpretation of the fluorosis regression is not correct change this to say that the results showed no difference between artificially and naturally fluoridated water (GDS)

7. Discussion

Put discussion of each objective at the end of each section and make each objective a separate chapter (JM agreed by general consensus)
Suggest future research include personal fluoride exposure measurements (PM)
Highlight that fluorosis was the only adverse effect to show association with water fluoride (PM)
Remit of the review - EB asked about whether the remit should be broadened. After discussion with CRD, DOH and panel members it was agreed that the workload involved in the current remit was already big enough that if it got any bigger the quality would be compromised and the review process and practical aspects would result in significant errors
The general consensus was that the discussion section was decided to be too long and repetitive. Several sections will be removed or shortened
EB suggested that the language referring to the overall quality of studies, particularly with respect to objective 4, was too kind
Section 5.3 The ability of the evidence to answer the objectives. IC said that the word 'address' rather than 'answer' should be used
EB said that some of the language in section 5.3 was too vague
JM suggested that the general discussion section start with section 5.4 Limitations of this review
ML offered to help get the Polish papers translated. CRD will pursue getting the Greek paper translated
TS and GDS suggested removing 'vote counting' from the discussion section
The terms 'reducing caries' and 'preventing caries' were discussed. The difference between these should be made clear, i.e. if reducing refers to DMFT and preventing refers to % caries free (IC)
Policy statements will be taken out of sections not directly referring to other factors to be considered
The first sentence under section 5.6 Other factors to be considered (The scope of this review is not broad enough to answer independently the question 'should fluoridation be undertaken in the UK?'), was considered to be too proscriptive. It was suggested that it be removed or moved, or softened
Section 5.6.2 PM disagreed with the sentence referring to comparable total fluoride exposure in two areas considered comparable in other ways. It was agreed to remove this sentence. PM also suggested that his study be referenced in this section
S Gibson, PM and EB said that the fluorosis level quoted in the discussion (10%) was incorrect and should be 12.5%
The implications for practice and research sections were discussed at length. The suggestion was to shorten these somewhat, and just clearly state where the gaps in knowledge are
JR sought replacement of 5.9.1 to 5.9.3 with a list of the unmet research needs identified from within the remit of the study, e.g. : caries effects adjusted for confounding factors like those noted in 4.22 and sugar consumption, exposure to other sources of fluoride etc effect on prevalence of caries across social classes long term effects on caries and overall health different effects between natural and artificial fluoridation impact of different levels of fluoridation on dental fluorosis association with congenital diseases, infant mortality and SIDS

8. Time scale

Next draft 19th May
Referees comments by 5th June
1 July to DOH
Publication end of July/early August
Do not send current version to external referees; wait for version of 19th May
Website version - remove comments reply box. Put up statement that updated version will appear on 19th May and comments can be submitted on this version after this date

9. Meeting closure (approximately 4pm)

Return to the Fluoridation Review home page.