Factors in ﬂ uencing the relationship between ﬂ uoride in drinking water and dental ﬂ uorosis: a ten-year systematic review and meta-analysis

The relationship between naturally ﬂ uoridated groundwater and dental ﬂ uorosis has received large attention from researchers around the world. Despite recognition that several factors in ﬂ uence this relationship, there is a lack of systematic studies analyzing the heterogeneity of these results. To ﬁ ll such a gap, this study performs a systematic review and meta-analysis to understand which factors in ﬂ uence this relationship and how. Selected studies were sampled between 2007 and 2017 from Web of Science, PubMed, Google Scholar and Scopus using keywords and Boolean operators. Results of the systematic review show that dental ﬂ uorosis affects individuals of all ages, with the highest prevalence below 11, while the impact of other factors (gender, environmental conditions, diet and dental caries) was inconclusive. Meta-regression analysis, based on information collected through systematic review, indicates that both ﬂ uoride in drinking water and temperature in ﬂ uence dental ﬂ uorosis signi ﬁ cantly and that these studies might be affected by publication bias. Findings show that ﬂ uoride negatively affects people ’ s health in less developed countries. The conclusions discuss policy tools and technological innovations that could reduce ﬂ uoride levels below that of the World Health Organization (WHO) ( < 1.5 mg/L).


INTRODUCTION
Fluorine ranks 13th in the world for abundance and constitutes 0.08% of the Earth's crust (Sananda & Biplab ).
Fluorine, a natural element, is easily soluble in water, soil and air, and being one of the most reactive chemical elements, does not exist on its own in the natural environment but rather as fluoride. Fluoride represents compounds that have ion F- (Pitzer ). Surface water is normally low in fluoride, with values lower than (1.5 mg/L), while groundwater can contain higher concentrations of fluoride depending on geological conditions (Sananda & Biplab ). Starting from the first water fluoridation studies in the USA in the late 1940s, fluoride has had a tremendous impact on the oral health of millions of adults and children (Blinkhorn & Mekertichian ). The reduction in dental caries ranged from 20% to 40% (Downer & Blinkhorn ). However, epidemiological evidence suggests that higher fluoride concentrations in drinking water above (1.5 mg/L) increases the risk of dental fluorosis, while progressively higher concentrations lead to increasing the risks of skeletal fluorosis (World Health Organization (WHO) ). Skeletal fluorosis is a slow, progressive, crippling condition, such as bone fracture resulting in health complications with overlapping manifestations in several other diseases (Gopalan et al. ). In early stages, dental fluorosis in humans is mainly identified through dental defects (Dhar & Bhatnagar ). Other conditions and diseases associated with excess fluoride intake include paralysis, respiratory complications and low blood pressure (Andezhath & Gosh ), while after chronic exposure, symptoms such as weight loss, anorexia, anemia and cachexia are very common (Ibrahim et al. ). In most circumstances, dental fluorosis occurs during tooth formation and influences dental enamel formation or mineralization and its structure (Blinkhorn & Mekertichian ). It can result in esthetic and functional problems depending on the severity of the condition (Sananda & Biplab ). For this study, we evaluated the impact on dental fluorosis from naturally fluoridated water while controlling for other possible factors related to this condition. Epidemiological studies that analyze the relationship between intentional water fluoridation and dental fluorosis are not the focus of this paper. We believe that it is useful to keep these two phenomena separated because of their diverse causes and different policies to reduce the risk of dental fluorosis.
Regarding the factors that may influence dental fluorosis, we investigated the influence of age on the severity of dental fluorosis in children exposed to drinking water, as well as other factors such as local temperature, rainfall and altitude indicated as important factors in determining the total concentration of fluoride in drinking water. Higher temperatures can increase the level of water intake above that usually consumed (Galagan et al. ; Firempong et al. ). On the other hand, the use of collected rainwater can reduce the use of unsafe groundwater (Susheela ; Edmunds & Smedley ).
Furthermore, it has been reported (Manji et al. ; Rwenyonyi et al. ) that altitude causes an increase in the prevalence and severity of dental fluorosis.
To understand the relationships between naturally fluoridated water and fluorosis, we proposed to answer the following research questions: • What is the relationship between naturally occurring water fluoride concentrations and the prevalence of dental fluorosis?
• Is this relationship the same in different geographic areas of the world where, according to the systematic review, this disease is more prevalent?
• What are the impacts of other factors, such as age, climate and altitude on the prevalence of dental fluorosis?
The remainder of this article is organized as follows. The methodology section will explain both the criteria used to select studies for the systematic review and the statistical approach used to perform the meta-analysis. The results section will discuss the qualitative and quantitative findings of the systematic review and meta-analysis, respectively.
Finally, in the conclusions, we will discuss possible policy implications that emerged from the study.

METHODS
It is widely recognized that systematic reviews and metaanalyses have become more important in many disciplines because they allow researchers and practitioners to be up to date on advances in the field, to identify research questions based on the results obtained in different contexts and to justify grants for further research (Cooper ; Littell ). To answer the research questions presented above and to achieve the objective of this study, we conducted both a systematic review and a meta-analysis. A systematic review is a process that uses systematic and explicit methods to identify, select and critically appraise relevant research. It collects and analyzes data from the studies included in the review (Moher et al. ; Gough et al. ; Mallett et al. ). The well-defined process of inclusion and exclusion of suitable articles is based on the identification of ad hoc criteria and promotes a confident approach to analyzing data not only qualitatively but also quantitatively bringing to fore empirically robust results via meta-analysis.

Search strategy and identification criteria
We During the systematic review, we detected papers written in Chinese, Spanish and Arabic languages. We do not consider these articles following the suggestions of Cochrane The motive for including the Dean and Thylstrup-Fejerskov indexes in the Boolean search is linked to the fact that many studies reporting information about dental fluorosis widely used the two indexes in their analyses. The choice of the proposed time span was motivated by the fact that fluoride concentration and variation can be related to climate, rainfall and temperature, and thus enlarging the timeperiod can introduce not only additional heterogeneity, but also information that does not necessarily reflect the current situation related to climatic changes. Table 1 shows the criteria that were used to select studies included in this systematic review and meta-analysis. One of the most important criteria for the initial selection of studies was the evaluation of the relationship between natural fluoride content in drinking water and the prevalence of dental fluorosis.
Sampling of studies and data extraction Figure 1 shows that initially, a total of 2,580 peer-reviewed papers were identified from the four databases as follows: Web of Science (970), Scopus (883), Pub Med (577) and Google Scholar (149), and one study was obtained by requesting it from the author. Information gathered at this stage on all articles was stored in Endnote. The first major elimination of duplicated copies was carried out on Endnote, where 796 studies were eliminated. For the remaining stages, two reviewers (MA and LG) analyzed the remaining 1,784 studies and independently applied the inclusion and exclusion criteria as indicated in Table 1. Disagreements were resolved through discussion and consensus with a third author (GN). At the end of the screening stage, 39 articles were included in this study. All 39 articles underwent a content check and qualified for systematic review. However, only 21 of the 39 articles qualified for meta-analysis (see Table 2), since 18 articles did not contain information on key variables, such as fluoride level in drinking water or the number of people affected by dental fluorosis. • Articles focusing on humans.
• Empirical, theoretical, review, clinical, academic papers and reports providing information from all countries around the world.
• Articles that focused on connections between natural fluoride content in drinking water and prevalence of dental fluorosis.
• Articles that focused on population that are long-term inhabitants.
• Articles providing information on fluoride content.
• Articles providing information on the sample size.
• Articles providing information on the number of people affected by fluorosis.
• Articles providing information on the age of participants.

• Articles published between
April 2007 and May 2017.
• Articles in other languages other than English.
• Articles on intentional water fluoridation.
• Articles on water defluoridation technologies.
• Articles that focused on animal dental fluorosis both livestock and laboratory research animals.
• Articles providing information on dental fluorosis caused by fluoride from soil and air.
• Articles providing information on dental fluorosis caused by industrial pollution.
• Articles providing information on dental fluorosis linked to other health issues e.g. children IQ, cancer, etc.  For the 21 articles selected for meta-analysis data of maximum and average annual temperature, annual rainfall and local altitude were additionally sourced. When not included in the studies, data were collected from https:// www.weather2visit.com/.

Statistical analysis
Meta-analysis refers to the use of statistical techniques in a systematic review to integrate the results of included studies.
Data from the 21 studies selected for the quantitative analysis showed dental fluorosis heterogeneity. We focused on random-effects meta-analysis of the dental fluorosis prevalence rate defined as the proportion of people affected by dental fluorosis. Binomial exact confidence intervals were used to contrast and summarize the results. The use of the random effect model is justified by the expectation that observed differences among the prevalence rates cannot be entirely attributed to sampling error, but also to other factors such as differences in the population under observation, publication bias etc. (Egger et al. ). Heterogeneity is quantified using the I 2 statistic that represents the percentage of total variation across all studies due to betweenstudy heterogeneity (Higgins & Thompson ). Usually, an I 2 value greater than 50% indicates significant heterogeneity. We also performed an independent meta-regression analysis to estimate the contribution of different study characteristics to heterogeneity within the selected sample of studies worldwide. A p-value <0.05 was considered statistically significant, unless otherwise specified. Statistical Data categorization was also conducted by subgrouping the prevalence rate by age, i.e. splitting the sample of

Systematic review results
All the studies included in this systematic review confirmed the higher prevalence of dental fluorosis with long-term consumption of elevated fluoride levels (>1.5 mg/L) in drinking water. One article confirms lower dental fluorosis prevalence rate with a lower level of fluoride in drinking water Besides the impact of naturally occurring fluoride content on dental fluorosis, the analysis of these studies shows that this condition is influenced by several factors that have been categorized into the following five relationships: 1. Age and prevalence of dental fluorosis; 2. Gender and prevalence of dental fluorosis; 3. Temperature, rainfall and altitude and prevalence of dental fluorosis; 4. Dental caries and prevalence of dental fluorosis and 5. Other dietary intake and prevalence of dental fluorosis.
Age and prevalence of dental fluorosis In 37 out of 39 papers selected for this systematic review, age was considered by all authors to be a key variable that is related to the prevalence of dental fluorosis. In the studies Prevalence of dental fluorosis, in most studies, display a positive linear relationship with greater age. Most of these studies were conducted in schools, and the age of children does not necessarily refer explicitly to a specific age, but rather to diverse age brackets ranging from 3 to 18 years of age.
Children in the teeth and body tissue growth phase (0-5 years) are more susceptible to dental fluorosis due to tissue, bone and teeth mineralization. This situation increases the susceptibility of the enamel to many forms of attack leading to discolored or mottled teeth.
Dental fluorosis in children also seems to be more prevalent and evident in permanent teeth than in primary teeth, further suggesting that in children over ten years of age, dental fluorosis is more visible and prevalent than those below the age of eight ( years old seem to be more affected where their condition deteriorates more rapidly compared to that of their female counterparts. In the oldest group, the male fluorosis prevalence rate is also found to be consistently higher than that of females, but these differences are not significant both for dental fluorosis and for urine samples tested in the An interesting concern is that some severe dental fluoro-

The meta-proportion results
The meta-analysis of the 21 selected studies starts by proposing in Figures 2 and 3 the forest plots related to the impact of age and fluoride level in drinking water on the prevalence of dental fluorosis. The forest plots display the results from each study as a square and a horizontal line, representing the intervention effect estimate together with its confidence interval. The analysis highlights a substantial level of heterogeneity for all the variables.
In Figure 2, the forest plot of the fluoride content in drinking water is presented. Most of the studies showed a level of fluoride between 1.5 and 3.0 mg/L. The forest plot also shows that the overall prevalence rate is 60% (0.60; 95% CI ¼ 0.44-0.75) and that increasing levels of fluoride in drinking water affect the higher occurrence of dental fluorosis. For values below 1.5 mg/L, which is the recommended value by Taghipour et al. (), the prevalence rate was 41% (0.41; 95% CI: 0.03-0.80). For values higher than 3.0 mg/L, the percentage of people affected by the dental fluorosis condition was statistically 92% (0.92; 95% CI: 0.84-0.99). However, the heterogeneity was significant (p < 0.001, I 2 ¼ 99.93%).
With respect to age, Figure 3 presents the pooled prevalence rates of the studies divided into three subgroups, under 11, between 11 and 15 and older than 15 years. The results indicate a higher prevalence rate for the first group.
However, in this case the heterogeneity is relevant (p < 0.001, I 2 ¼ 99.93%) as well.

The meta-regression results
Considering the relatively high heterogeneity shown in these studies, meta-regression analysis was conducted to explore the possible sources of this heterogeneity further. Table 3 shows the results of two meta-regression analyses, where the dependent variable is the prevalence rate of dental fluorosis of the 21 studies reported in Table 2.
Meta-regression 1 included only the fluoride level (FLUOR) in drinking water while meta-regression 2 includes also other regressors: two dummy variables DAFRICA and DAMERICA, respectively, with 1 if the observation is in a group of countries in Africa or America and zero otherwise. The two dummies were included due to the possible differences among the three groups of studies (Africa, America and Asia) which emerged from the preliminary analysis. In meta-regression 2, we also included the TEMP indicator calculated as a ratio between the maximum and average annual local temperature through which we want to consider possible increases in water consumption due to warmer temperatures. Temperature data were retrieved from the studies and when not available, from https://www.weather2visit.com/. Other models were estimated that included the rainfall and altitude variables. For brevity, these results were not reported due to the non-significance of the two variables.

Publication bias
Results presented in Table 4 suggest that the publication process of these articles and the scientific community seem to value information regarding the prevalence rate of fluorosis differently. In model 1, the beta parameter of the h-index is positive and statistically significant (p < 0.05); therefore, it is likely that the higher prevalence rate of dental fluorosis seems to influence more the acceptance rate of papers managed by high-profile journals. On the other hand, model 2 shows that the beta parameter of the number of citations is not significant, and thus this does not influence the citation behavior of the scientific community.

DISCUSSION AND CONCLUSIONS
In this study, we present the results of a systematic review and meta-analysis aimed at understanding how environ-  In parentheses () the standard errors, in parentheses {} the p-values of the t-statistics. The first is p-value under the t-statistic distribution, the second one is the p-value computed using a permutation test proposed in Higgins and Thompson (2004) to simulate the data under the null-hypothesis. Twenty thousand replications were used to compute the p-values. The Τ 2 is the estimate of the between-study variance and I 2 is the percent of residual variation due to heterogeneity.
For individuals older than 18 years of age, dental fluorosis is also associated with skeletal and non-skeletal fluorosis,   In parentheses () the standard errors, in parentheses {} the p-values of the t-statistics. The first is p-value under the t-statistic distribution, the second one is the p-value computed using a permutation test proposed in Tahir and Rasheed (2013) and Higgins and Thompson (2004) to simulate the data under the null-hypothesis. Twenty thousand replications were used to compute the p-values. The Τ 2 is the estimate of the between-study variance and I 2 is the percent of residual variation due to heterogeneity.
long-term climatic changes, the problem of dental fluorosis in these geographic areas will be exacerbated.
The significant positive relationship between temperature and fluoride in drinking water suggests that the use of innovative technologies cannot be neglected, because if temperature increases as a consequence of long-term climatic changes, the problem of dental fluorosis in these geographic areas will be exacerbated. Rainwater harvesting systems, for example the use of rainwater collection tanks, can help in reducing the consumption of unsafe groundwater (Susheela ; Edmunds & Smedley ).
However, recent research (Haque et al. ) showed that the performance of these systems can be negatively impacted by climate change conditions. Such situations mainly involve an expected worsening of drought periods in the future decades.
Meta-analysis also seems to indicate publication bias, and thus more attention should be paid to the publishing process to include in high h-index-ranked journals more studies reporting low and medium prevalence rates of dental fluorosis disease.
The negative impact of dental fluorosis on people's health is well recognized by many governments but not very much is done to tackle this issue affecting the life of millions of people in developing countries. The inadequate capacity of poor people to find alternative water sources is a challenge leading to high exposure to dental fluorosis in fluoride-affected regions. Improved water management systems can efficiently respond to naturally fluoride water impacts, but this solution is often overlooked in villages where people struggle to find better sources of water. Governments should invest in modern social water policy programs that can supply filtrated or clean water in remote villages where lack of alternatives has forced residents to utilize springs and other unconventional sources of water. However, to achieve such results, actions such as Prevention via innovative technologies and implementation of these policies appear to be a promising strategy to improve the quality of life of both children who are reported to undergo socio-psychological distress due to mottling and damage of teeth and adults who also face dental skeletal fluorosis. However, considering that the complexity of dental fluorosis is determined by the content of fluoride in drinking water, preventive results can only be achieved with more international cooperation and interdisciplinary research to tackle this problem from different angles.