Main

Water fluoridation is a well accepted public health measure. It is safe, effective, cheap and requires no effort on the part of the general population to benefit. Yet few of the British population benefit from this simple measure, which would markedly reduce dental decay in 5-year-old children within 5 years.

Members of the dental profession see the problem as ineffective legislation. The Water Fluoridation Act received Royal Assent in 1985, and no new water fluoridation schemes have been implemented since then despite requests from many health authorities.1

Social deprivation is associated with a gradient of mortality and disease morbidity2 including the prevalence of dental caries.3,4,5 More recently the use of indices of material deprivation instead of social class, to identify children at risk of dental caries has been reported.6,7

Water fluoridation is also reported to be more effective in reducing dental caries in lower social classes or in social groups classified by lower income or poorer educational attainment, so can therefore reduce, but not eliminate, social inequalities.8,9,10,11 These findings have been confirmed using a number of different deprivation indices.12,13 However, a review of the literature shows a number of studies which show no differential benefit of fluoridation to more deprived children and some which show the opposite effect.

Two approaches have been used to control for socioeconomic deprivation while examining the effect of fluoridation on dental caries. The first is to categorise individual children by the occupation, income or educational attainment of the parents — a social class approach. The second uses an ecological design based on the home or school address of the individual to assign them to an area with known socio-economic characteristics. These are usually defined in indices of deprivation.

Social class

Two studies from Finland in the early eighties showed no extra benefit from water fluoridation to more deprived groups in either the deciduous or permanent dentition.14,15 However, the studies were based on convenience samples and the authors themselves state that the results should not be generalised, presumably because they recognised the selection bias of their study population.

A study from New Zealand reported that 'the relationship between fluoridation, socioeconomic status and caries was equivocal'.16 The report shows they sampled 1,661 5-year-old children of whom 923 (56%) were dentally examined and 709 questionnaires were returned with the reported fluoride history of the children. This is a response rate of 43% and no attempt was made to assess response bias to see how representative the responders were. However, the study has a more basic flaw. There were only 61 'non-fluoride' children categorised into the three socioeconomic groups and the authors stated this will 'place some limitation upon the conclusions to be drawn from the study'. Using the original figures reported in the study if we assign 20 to each of the three groups, the power of the comparison between groups is only 21%, suggesting a type 2 statistical error; ie if the study was repeated 100 times, 79 times it would on average yield no significant difference when in fact one existed.17

Another author contradicted his own earlier conclusions of more benefit to deprived children9 after reworking data and adding new data on treatment provision.18 In addition to obvious selection bias in relying on treatment data, the report has been criticised for using different measures of social class over time and ignoring cross border flow in school attendance. The large annual variation in treatment provided per child also casts doubt on the validity of using quantity of treatment as a proxy for disease.19

Indices of deprivation

A recent study used individual DMFS in 14-year-olds and Enumeration District Townsend indices for the analysis, and found that fluoridation differentially benefits less deprived children.20 The study differed from most others in the use of fibre optic trans-illumination (FOTI) for caries detection, the DMFS index, and a sub-optimal level of fluoride in the study area (Anglesey) of only 0.7 mgF/litre. However, the use of individual DMFS scores but Enumeration District Townsend scores for a multiple regression is not ideal as reflected by the very low correlation coefficients reported.

Dental decay is not uniformly distributed throughout the population and surveys of 5-year-olds have shown increased disease polarisation with 43.6% of 5-year-olds in Great Britain with experience of dental decay in 1995.21

A previous study which clearly showed a differential benefit to more deprived children used Jarman scores as a measure of deprivation.22 Jarman was designed to weight payments to general medical practitioners and has recently been introduced to weight capitation payments to dentists and is not an index of deprivation. This study also failed to take account of the unequal number of children in each ward.23 However the analysis was carried out at ward level, not at an individual level, so the mean ward dmft is for the children in a ward regardless of the number. A regression model weighted by ward may bias the results in favour of those with large numbers, which tend to be deprived wards. To answer this problem, both a weighted and unweighted regression should be reported.

The aim of the present study was to examine the effect on dental caries of social deprivation, at an electoral ward level as measured by the Townsend score, in populations with none, artificial or natural water fluoridation.

Method

This is an ecological study ie based on geographical areas not individual subjects, of electoral wards in the North and North West of England. The results from the National Health Service 5-year-old-children dental epidemiological surveys in 1991/2 and 1993/4, coordinated by the British Association for the Study of Community Dentistry, were used.21

The Townsend score is used to measure social deprivation and is the weighted total of four transformed and standardised 1991 census variables and can be used to rank areas in order of deprivation; a score of zero is the English average.24 The four variables are the percentages of: 1) economically active residents (males 16–64 and females 16–59) who are unemployed, 2) private households without a car, 3) private houses which are not owner occupied, 4) private households with more than one person per room (overcrowded households). An area with a larger score is more deprived than one with a lower (including negative) score.

As previously reported, the data were collected for electoral wards in three study areas.22 The first was non-fluoridated Salford and Trafford Health Authority in 1993/4. The second was fluoridated wards in Newcastle District Health Authority and eight wards in North Tyneside District Health Authority which had received consistent water fluoridation (1 mgF/litre) in 1993/4. These two study areas are in urban areas in the north of England. The third study area was naturally fluoridated wards in Hartlepool (1.2 mgF/litre fluoride) which is a more rural area. The results from Hartlepool came from the 1991/2 survey and were included as a positive control in this study. There was no significant difference in the mean dmft of 5-year-olds in Hartlepool between 1991/2 and 1993/4 (independent sample, t-test, t = 1.74, P = 0.08). The District surveys of 5-year-olds for this study used all children in state primary schools as the sample frame. The mean electoral ward dmft was calculated directly from the raw data, by using the home postcode of the subjects to allocate them to the appropriate electoral ward. A mean ward figure was then calculated from the individual dmft scores.

Multiple linear regression, both unweighted and weighted by the number of subjects per ward, was used to examine the relationship of the dependent variable (mean ward dmft of 5-year-olds) with independent variables, ward Townsend score, fluoridation status (adjusted or natural) and interaction variables calculated as the product of ward Townsend scores and fluoride status.13,20 When the Townsend index was coded as dummy variables, the effects were found to be approximately linear and it was therefore considered appropriate to enter it as a continuous variable in the multiple regression model. Inspection of the least squares regression lines for the non-fluoridated, artificially fluoridated and naturally fluoridated wards also support this linear relationship. In order to overcome possible problems of heteroscedasticity, robust variance estimates were calculated to estimate the standard errors of the regression coefficients. The White/Huber sandwich estimators were used, as they make no assumptions about the distribution of the error terms.25 A significance level of 0.05 was used.

The null hypothesis for this study was that mean ward dmft scores are independent of social deprivation and water fluoridation.

Results

In Salford & Trafford (n = 5137) and Newcastle (n = 3816) 86% of 5-year-old children on the school role were examined. In Hartlepool the figure was 83% (N = 1051). In the fluoridated wards in North Tyneside the denominator population is unknown although 959 children were examined and the estimated response rate for North Tyneside was 87%. Children who were not examined were absent from school on the day the dental examinations were carried out.

The Townsend scores for the 41 wards in Salford & Trafford ranged from –4.48 to 10.94. In the 34 wards in Newcastle and part of North Tyneside the range was –3.30 to 9.40 and in the 17 wards in Hartlepool from –5.3 to 8.02. The mean dmft scores ranged from 0.24 to 4.32 teeth affected in Salford & Trafford, 0.35 to 2.54 in Newcastle and North Tyneside, and 0.23 to 0.98 in Hartlepool.

Weighted and unweighted multiple linear regression analysis gave similar results and both showed statistically significant differences between both the slopes of the fluoridated and naturally fluoridated regression lines and the slope of the non-fluoridated regression line (Table 2). There were statistically significant differences in the slopes of the association between ward Townsend scores and dmft in the fluoridated and non-fluoride areas, confirming a differential benefit to more deprived electoral wards.

Table 2 Table 2
Full size table

The best fit linear regression lines for ward Townsend score and dmft in nonfluoridated Salford & Trafford in 1993/4, fluoridated Newcastle & North Tyneside in 1993/4 and naturally fluoridated Hartlepool in 1991/2 are plotted in figure 1.

Figure 1
figure 1

The mean dmft and Townsend score by non, artificially and naturally fluoridated electoral wards, with best fit regression lines

Full size image

The divergence of the best fit regression lines in figure 1 shows that the more deprived a ward in the UK, the greater benefit 5-year-old children derive from water fluoridation.

There is a 43% reduction in the mean dmft of 5-year-olds in fluoridated wards at a Townsend score of zero, the national mean score.

Discussion

The significant differences found between study areas led to the rejection of the null hypothesis. The results are similar to those previously reported using the Jarman scores as a proxy for deprivation and confirm that dental decay in 5-year-old children has a strong positive association with deprivation as measured by the Townsend score at an electoral ward level, regardless of fluoridation status.22 The study areas are all in the north of England and therefore the District Townsend scores are higher than the English mean score of zero. The electoral wards in each of the three study areas covered a wide range of socioeconomic status, from very deprived wards with a score of 10 or more, to affluent wards with scores of –5 or less. The range of Townsend scores in each study area varied from 15.42 in Salford & Trafford to 12.7 in Newcastle so were broadly similar in spread.

The National Health Service 5-year-old-children dental epidemiological surveys are routine surveys and the calibrated, dental epidemiologists were not blind to the likely fluoride status of the local water supply. However, they did not know the results would be used in this study. To reduce inter-examiner variability, the diagnosis of caries for these surveys is carried out visually at the D3 diagnostic threshold — decay extending into the dentine.26

The response rates were high and ranged from 83% to 87%. The effect of missing those children absent on the day of the examination on the mean ward level of tooth decay is unknown. However, children with high absentee rates may have high levels of tooth decay and this may underestimate the mean ward dmft to favour the null hypothesis.

The weighted mean District dmft reported nationally for Salford & Trafford27 was 2.48 and in Hartlepool28 was 0.81 which both lie within the 95% confidence intervals derived from the mean of the ward scores (Table 1). The reported figure for Newcastle and North Tyneside is not comparable as non-fluoridated wards were included in the published District score. The confidence intervals of the mean dmft do not overlap confirming a significant difference between the three study areas.

Table 1 Table 1
Full size table

The study did not consider the fluoride history of the children in that some will have moved from fluoridated to non-fluoridated areas and vice versa, potentially reducing any effect attributable to water fluoridation. Office of Population Census and Surveys migration data shows that 11% of children aged 1 to 4 years of age in Northern Region and 10.3% in the North West Region (11.5% in England) had a different address one year before the census.29 Despite this potential confounding which would again favour the null hypothesis, the results were still statistically significant.

The divergence of the best fit regression lines shown in figure 1 clearly shows that the more deprived an area, the greater benefit 5-year-old children derive from water fluoridation and the significance of the interaction variables in both the weighted and unweighted multiple regression confirm this. There is much supporting literature for the finding that water fluoridation is differentially beneficial to more deprived groups and close examination of the reports which contradict this finding are methodologically biased, of insufficient power or are statistically flawed as discussed earlier.

The data from Hartlepool are included as a positive control with natural fluoridation at a level of 1.2 mgF/litre. This is higher than the level recommended for artificial fluoridation. The 43% reduction in decay at 1 mgF/litre, fluoride at a Townsend score of zero, rising to 66% at 1.2 mgF/litre also indicates a dose response relationship. The shallow slope of both the artificial and natural fluoridation regression lines shows how fluoridation ameliorates the effect of social deprivation on tooth decay in 5-year-old children but does not eradicate the dental health gap.

What are the implications of the results for individual children? The number of children in each ward varied from 13 to 264 and the mean reduction in dmft of 43% may reflect a differential effect for individual children, with high caries individuals gaining more benefit. As the mean dmft increases so does the prevalence in a population, thus in a fluoridated area a larger proportion of the child population has the potential to benefit as well as more severely affected individuals. Further research is need to investigate the dental health benefits for individuals.

The results show the differential efficacy of fluoridation to more deprived areas at an electoral ward level, confirming that most of the British population who are still deprived of water fluoridation, continue to suffer from the pain, disfigurement, misery and embarrassment associated with poor dental health.30 Recent surveys show an increase in tooth decay in 5-year-olds which may reflect increasing levels of child poverty.22 Deprivation per se does not cause tooth decay and the causes of dental decay results from an inability or unwillingness to eat or buy a non-cariogenic diet, and behavioural risk factors such as high frequency of sugar intake and not brushing with a fluoride toothpaste.4

Even if water fluoridation was introduced more widely, we would still see a social class gradient in disease levels, though unlike many health promotion initiatives, on the evidence from this study, the 'dental' caries divide would become narrower. Opponents of water fluoridation argue that fluoride toothpaste is available as an alternative to water fluoridation for those who chose to use it. However, almost all of the toothpaste sold in the UK contains fluoride and we can safely assume there was no difference in the proportion of toothpaste used which was fluoridated between the two study areas. However, in England there is evidence of a North/South gradient in the volume of toothpaste purchased per head of population.31 Despite fluoridated toothpaste, water fluoridation still produced a 43% reduction in decay.

The polarisation of dental caries confirms the urgent necessity to institute water fluoridation as the most cost-effective primary preventive measure capable of bridging the 'dental caries gap' by improving the dental health of those who are socio-economically deprived.

The dental profession can advocate water fluoridation, secure in the knowledge that it is evidence-based and effective. Our results clearly show that today we have almost twice as much tooth decay in non-fluoridated areas because the population is denied water fluoridation.

This study was carried out by the North West Dental Public Health Resource Centre which is jointly funded by Health Authorities in the North West Region. The author extends thanks to Mrs Joyce Tadeusiak for her secretarial expertise, to Dr G. Whittle, Mr D. Evans, Mr D. Trotter and Mr J. Stewart for the provision of data for this study and also to Mr. K. Woods, Director of the Resource Centre.