Summary

Har fluorid i drikkevand en betydning for caries hos børn ?

The aim of the present project was to describe the possibilities of linking data from health and environmental registers in relation to the question “Do we still see an association between fluoride in drinking water and caries in children”.

Data from the Danish Recording System for the Public Dental Health Services (SCOR), National Board of Health were linked to analyses of fluoride in drinking water. The project used data from SCOR, The Central Population Register (CPR), The Jupiter database (Geological Survey of Denmark and Greenland (GEUS), The Building and Dwelling registration (BBR), and The Tax registration (SKAT).

As part of the project technical and administrative problems with linking data from health and environmental registers by using data from Geographic Information Systems (GIS) were evaluated and described. Additionally, recommendations on measures to facilitate the linking process were given.

The first step was to collect and evaluate data from the included registers.

Data from children born in 1979, 1989, and 1999 were used for linking. Dental status at 5 and 15 years for these three classes was used for linking, since registration and reporting of dental status are statutory for these age groups.

Table 1 Study cohort defined by year of birth

	1979	1989	1999
Dental status at 5 years		X	X
Dental status at 15 years	X	X

The present study used two caries indices:

DMFS: Decayed, missing, and filled surfaces of permanent (indicated by capital letters) teeth
dmfs: Decayed, missing, and filled surfaces of deciduous (indicated by lower-case letters) teeth

The CPR register was used as reference, as data are considered complete regarding individual registration. Non-matching data to e.g. SCOR or incomplete data concerning residence/address were excluded from the final analysis, but used for analysis of drop-out rate.

Data linking was restricted to addresses that according to the BBR registration were connected to public water supply.

The following inclusion criteria were used for the individual waterworks to participate:

At least two values for fluoride analyses should be available in the period 1994-2004, if less than 5 analyses in the period, more than 5 years should separate the analyses
At least one value for catchment-amount should be available in the period 1994-2004
Coordinates (X,Y input to denote a space on earth) for the individual waterworks or associated residences should be present

It was not possible to link a specific waterworks to a specific address. The final result from the analyses of fluoride data in The Jupiter database was a digital map which separated the country into small polygons. To each polygon information on fluoride concentration in drinking water from a nearby waterworks was linked. By use of a simple GIS-operation fluoride information from the map could be transferred to all address-points with caries information. The resulting combination table constituted the basis for further statistical analysis.

The analyses of fluoride in drinking water did not show annual variation in fluoride concentration.

However, a considerably variation in fluoride concentration, due to geographical site of the waterworks, was found. On this basis an exposure variable was constructed for a given waterworks as the average of analyses over a ten-year period (1994-2004).

Data were drawn from the 4 cohorts in table 1, defined by year of birth and follow-up period. Each cohort was analysed first with all children participating in the cohort and next with children having the same address in the exposure period, where the fluoride exposure was assumed to be constant throughout the period. The cohort with 15 years follow-up used DMFS as outcome, and the cohort with 5 years follow-up, dmfs. Level of fluoride exposure was calculated from the fluoride concentration and duration of residence linked to the specific address.

A total of 67.261 persons born in 1989 were registered in The Central Population Register (CPR). Of these 64.223 (95.5 %) were registered in The Tax registration (SKAT), and 52.645 (78.3 %) were registered in the Danish Recording System for the Public Dental Health Services (SCOR). A total of 43.848 (65.2 %) were registered in SKAT, SCOR and BBR.

The present project found an inverse relation between fluoride concentration in drinking water and risk of caries (50 % reduced risk for DMFS > 1 and 60 % reduced risk for DMFS = = 6 at fluoride concentration > 1 mg/l) in permanent teeth in children born in 1979 and 1989.

The deciduous teeth in children born in 1989 had a 50 % reduced risk for caries (dmfs) at fluoride concentration > 1 mg/l in drinking water. The effect of fluoride in drinking water on dmfs was reduced to 25 % over the 10 year period (children born in 1999).

The present project showed that it was possible to link data from health and environmental registers within a reasonable time. Also, the project showed that linking this kind of data required registers with valid data including a large part of the population. The Danish registers showed to fulfil these criteria. Data analyses of large and complex data, as in this project, required specialist with detailed experience in databases and programming.

The historical part of the registers is essential in projects using register data as exposure indices. In the Building and Dwelling registration (BBR), old data are exchanged with new when updating data. This is inappropriate in projects linking data from registers and could give rise to erroneous exposure assessment.

This project used income data from SKAT as proxy for social status. It would have been preferable to use income data over a given period for a cohort, but data from SKAT did not allowed this. To do this, data from Statistics Denmark (DST) should have been used. These data are available only inside DST’s data network, and since the project group wanted to use GIS - and GIS is not in the research computers in DST - this could not be done.

It was not possible to generate a water supply map relating a specific waterworks to a specific address. Therefore, GIS was used for generating a theoretical calculated water supply map. This means that drinking water data do not exist at individual person level, and that the exposure data in certain areas, where fluoride concentration varies considerably even in small areas, are unreliable. The generation of the water supply maps gave several methodological problems and emphasized the need for a real water supply map based on substantial data from the waterworks. In this way data in the present project could be further optimized as only data from SCOR and SKAT were analysed at individual person level.

Such a database would also be very helpful in these circumstances where efficient alert has to be constructed to handle drinking water pollution and alternative drinking water supply.

In addition, investigations of relations between drinking water and its content (pesticides, microbial pollution, trace elements) in relation to human health as in the present project could be markedly improved and optimized by access to these data. The project group recommends initiation to build up such a database / digital map.

It is recommended that registers used for research should prepare standards with demands for data declaration, data update including creation of historical cohorts, principles for handing over, right of admission, and price setting. Request for feedback from researchers to data holders about errors and lacks should be initiated e.g. by filling in a report.

The present project group recommends that all data holders consider the research potential of their registers and guarantee access for the researchers to the registers. In this way the best data for environmental and health research projects would overall be secured and the best data for research in this area would be placed in Denmark.