Concentration, distribution and probabilistic health risk assessment of exposure to fluoride in drinking water of Hormozgan province, Iran

Herein, the health risk assessment of exposure to fluoride in drinking water of southern province of Iran was performed with a definite and probabilistic approach. The Monte-Carlo simulation and sensitivity analysis were used to explain the impact of risk and uncertainty upon estimations. The results indicated that fluoride concentration was in the range of 0.1–0.9 mg/L with an average of 0.454 ± 0.205 mg/L, and distribution function followed the normality. Moreover, the highest fluoride concentration was observed in the central and northern regions of the province. In the definitive method, hazard quotient (HQ) associated with fluoride in drinking water were lower than 1 (HQ < 1). Conducting Monte-Carlo simulation illustrated that the potential non-carcinogenic risk for children in the 95th percentile exceeded the safe limit of 1, indicating a potential non-carcinogenic in this age group. Sensitivity analysis showed that fluoride concentration and water consumption rate had the most effect in the model. Therefore, consumption of water and foods containing fluoride along with excessive consumption of tea could increase the risk for human health. The results suggested that a continuous monitoring of fluoride in water resources besides proper dietary regime for inhabitants of this province should be taken into consideration.


Introduction
Due to increasing public concern about water resources and the environment, the importance of groundwater and surface water conservation has become apparent. Issues related to water supply, water quality, and the destruction of aquatic environments have nowadays become among the most attractive issues (Peters and Meybeck 2000;Yousefi et al. 2017). Heavy metals, pesticides, nitrate, and fluoride are instances of overriding pollutants in drinking water sources. Nath et al. 2018;Kiefer et al. 2019). Small amounts of fluoride (F) are naturally present in water, air, soil, and living organisms. As a result, humans are exposed to fluoride through food consumption, drinking water, skin contact, and respiration (Badee et al. 2014;Yousefi et al. 2018;Mukherjee et al. 2019;Badeenezhad et al. 2020). Fluoride concentration in natural waters differs from 0.1 to 10 mg/L (Edmunds and Smedley 2013). Different factors such as geological settings, chemical characteristics of groundwater, porosity of aquifer materials, pH, climate condition and well depth have an influence on fluoride concentration (Hossain and Patra 2020).
Based upon World Health Organization (WHO) guidelines, the maximum concentration level (MCL) of fluoride in drinking water is 1.5 mg/L (WHO 1993;Xiao et al. 2015;Badeenezhad et al. 2019;Dehghani et al. 2019). Nowadays, the problems of exposure to high fluoride concentrations, especially in children, have caused numerous health problems. Fluoride helps maintain strong bones, but during intake of high contents of fluoride, it can cause discoloration and tooth decay, osteoporosis, and damage to the kidneys, nerves, and muscles (Li et al. 2001;Harrison 2005;Fordyce et al. 2007;Badeenezhad et al. 2021a, b).
Recent studies have reported high and low levels of fluoride in groundwater resources in some parts of Iran. The regions with an anomalous content of fluoride in drinking waters in Iran mainly located in the northwest, central, south and southeast of Iran (Battaleb-Looie et al. 2012;Mohammadi et al. 2017;Dehbandi et al. 2018;Naderi et al. 2020). The study results of Mirzabeigi et al. showed that the average concentration of fluoride was 2.92 mg/L (range: 0.9-6 mg/L). Also, in half of the villages, F concentration was higher than the standard recommended by the WHO (1.5 mg/L) (Mirzabeygi et al. 2018). Also, Ghaderpour et al. showed that the average fluoride concentration in urban and rural areas of Khorasan Razavi province were 0.74 and 0.59 mg/L, respectively, which were lower than WHO recommended value (Ghaderpoori et al. 2018). These observations have raised concerns about the level of fluoride in drinking water (Saeed et al. 2020). Fluorosis is currently a major issue in more than 25 countries worldwide, and in many regions, the incidence of fluoride poisoning is staggering (Ali et al. 2019). According to the latest proclamation, over 200 million people are at risk of fluorosis, which raises the alarm globally [13]. The issue of long-term intake of highfluoride drinking water and related fluorosis has been reported in different countries such as India, China, Tanzania, Mexico, Argentina, South Africa and Pakistan (Selinus et al. 2005;Ayoob and Gupta 2006;Radhika and Charan 2017;Kashyap et al. 2020;Rashid et al. 2020). Therefore, due to global water quality concerns, the spatial distribution of fluoride in water resources, especially in arid areas, is essential to design an appropriate water management strategy. A general monitoring of fluoride in groundwater resources of Iran have indicated that Hormozgan province is one of the regions with elevated fluoride level (higher than 1.5 mg/L) in Iran (Mesdaghinia et al. 2010). Also some evidences of prevalence of dental fluorosis were reported in this province (Davari et al. 2004). However, until now, there is no comprehensive study on the monitoring and health risk assessment of fluoride in groundwater resources of this province.
Health risk assessment is a proper method to assess the potential effects of pollutants on human health. In the deterministic health risk assessment, the risk is usually based on the average value of various parameters. However, when the data standard deviation is large, it may underestimate or overestimate the actual risk. Consequently, the use of probabilistic approaches such as Monte Carlo simulation provides a more accurate risk assessment (Fernández-Macias et al. 2020).
In the current study, fluoride concentration and its distribution in drinking water resources of Hormozgan province were assessed. Over the past two decades, the province has faced significant freshwater scarcity and precipitation reduction challenges. Besides, Monte Carlo simulations and sensitivity analysis were applied to estimate the potential health risk of fluoride and quantify the risk-related uncertainty, respectively.

Study area
Hormozgan province is one of the southern provinces of Iran with an area of 70,000 km 2 and a population of 1.7 million. Hormozgan province is located between the geographical coordinates of 25°24 0 to 28°57 0 N latitudes and 53°41 0 to 59°15 0 E longitudes (Fig. 1). The province includes 11 counties of Jask, Bashagard, Sirik, Rudan, Minab, Bandar Abbas, Hajiabad, Khamir, Bastak, Bandar Lengeh and Parsian. According to the De Martone climate classification, the climate of the study area is very hot and humid (de Martonne 1926;Bazrafshan and Dehghanpir 2020). The average rainfall, temperature, and humidity are 215.8 mm, 27°C, and 19-100%, respectively. This province has a complex geological structure and structurally can be divided into Zagros, Makran, and Central Iran zones (Stöcklin 1968; Asadpour 2017).

Sample collection and analysis
A total of 56 drinking water samples, including tap water, well water, and springs, were collected from 11 cities of Hormozgan province in 2019 (Fig. 1). Pre-rinsed lowdensity polyethylene bottles were used for sampling. Firstly, they were soaked in nitric acid solution (20%) for 24 h and then washed with tap water and distilled water. Before sampling, the bottles were washed three times with sample water. The samples were then stored at 4°C until chemical analysis. The collected water samples were transferred to the chemistry laboratory for fluoride analysis using the SPADNS colorimetric method. After the reaction of fluoride with zirconium, the produced color in the solution was measured by a DR 5000 TM UV-Vis Spectrophotometer(LPG408.99.00012) at a maximal absorption wavelength of 570 nm (Das et al. 2017).

Exposure and health risk assessment
United States Environmental Protection Agency (USEPA) Health risk assessment is a method to assess the potential harmful effects on human health from exposure to specific chemical agents for a specific period (Means 1989). This method has been widely used and validated in various cases of health risk estimation in different environments (Bodrud-Doza et al. 2020). The optimal fluoride concentration is assigned by the WHO for temperate regions and is unsuitable for tropical, arid, and semi-arid regions with higher air temperatures because it is created as a function of the average temperature of 16°C (Muller et al. 1998).
The optimal level of fluoride concentration in drinking water is calculated based on the average annual maximal temperature (Tm) and varies from 0.5 to 1.5 mg/L depending on the temperature and climate of the region. Because the average annual maximal temperature of Hormozgan province is high, so it is necessary to calculate the optimal fluoride concentration.
The regional optimal concentration of fluoride in drinking water was calculated by Eq. (1) (Muller et al. 1998;Badeenezhad et al. 2021a, b): where D is the optimal dose of fluoride in drinking water (mg/L) and Tm is the average annual maximal daily temperature for the last 5 years (°F). Due to the existence of different regions and climatic conditions in Iran, the temperature range across the country may vary from -20°C to ? 50°C. The average calculated Tm in Hormozgan province is 31.66°C (Zazouli et al. 2017). People may expose to fluoride through oral consumption (drinking water), skin absorption, and inhalation. However, the routes of inhalation and skin contact have not been included in this study due to lack of toxicological data, such as inhalation reference dose for fluoride, water-to-air transfer efficiency, and low skin exposure. The intake dose of fluoride through water consumption was estimated for three age groups (C:children; T: teens; A: adults) according to the USEPA method, according to Eq. 2 (EPA 1997;Badeenezhad et al. 2019): The explanations of the parameters are presented in Table 1. Also, HQ, which represents the non-carcinogenicity risk through different exposure pathways, was calculated using estimated daily intake (EDI) and oral reference dose (RfD) (Eq. 3) (Emergency and Response 1989;Badeenezhad et al. 2021a, b): A HQ value of [ 1 implies a significant risk level where can affects the teeth and bones and or can lead to potentially skeletal problems.

Monte Carlo simulations and sensitivity analyses
Monte Carlo simulations investigate uncertainties in a probabilistic approach constructing possible resulted models by substituting a wide range of values for the probability distribution in each parameter (Malakootian et al. 2020). During the simulation paths, the values of the selected parameters are randomly sampled based on the input probability distribution. The input probability distribution can take many forms, including uniform, normal, and triangular.
The output of the simulation process is a function of the model output distribution. One of the main advantages of Monte Carlo simulation is the ability to reflect the overall uncertainty of input variables in risk assessment models. Sensitivity analysis, which is another part of Monte Carlo simulation, allows the researcher to identify the parameter that has the greatest impact on risk assessment (Chen et al. 2019). In this study, R6.32 software and mc2d Package were used for simulation and sensitivity analysis with 10,000 replications.
All data are expressed as triplicate with standard error. The Fitdistrplus package was used to find the best fluoride concentration distribution in drinking water. Analysis of variance was used by One-way ANOVA to determine significant differences between water supply sources in different countries. Also, the inverse distance weighting (IDW) technique was utilized in the ESRI ArcGIS software (version 10.8) to compute the fluoride concentration values for non-sample areas.

Fluoride concentration
Figures of 2 and 3 are presented the distribution GIS map and Box plot of fluoride concentration in drinking water in Hormozgan. The mean concentration of fluoride in drinking water ranged between 0.01 and 0.9 mg/L with mean value of 0.454 ± 0.205 mg/L. The highest fluoride concentration were observed in the central and northern regions of the province, namely in Hassan Langi region in Bandar Abbas County, and the Sargaz Ahmadi region in Hajiabad County (0.9 mg/L). Also, the lowest concentration was obtained in Rudan-Posht Banan with a concentration of 0.01 mg/L. Based on One-way ANOVA analysis, there was a significant difference between fluoride concentrations in different sampling points in Hormozgan province (P value \ 0.05). According to the post hoc test (LSD), except for Sirik County, the mean fluoride concentration was significant at least between two different counties, but the lowest and highest fluoride concentration was obtained in Khamir and Bandar Abbas county, respectively. The range of fluoride concentration in study area is similar to the some regions in Iran including Mashhad,   Since lithology plays a major role in fluoride levels in groundwater hence, its concentration is mainly controlled by weathering and dissolution of fluoride minerals. The fluoride contamination of groundwater affected chiefly by several factors, including the availability and solubility of fluoride minerals, pH, temperature, anion exchange capacity of the aquifer materials, type of geological materials, residence time, porosity, structure, depth, groundwater age, and carbonates and bicarbonates concentration in water (Sajil Kumar 2021). In a study by Zheng et al., the authors demonstrated that fluoride concentration in drinking water in Chongqing, China, ranged from 0.100 to 0.503 mg/L, with an average of 0.238 ± 0.045 mg/L. They also concluded that geological conditions largely influence fluoride concentration and spatial distributions in drinking water of Chongqing urban regions (Zheng et al. 2020). Mountains cover most of the Hormozgan province. The mountains are a continuation of the Zagros Mountains, which stretch northeast to southeast. Moreover, with the decline of altitude, the mountain range comes down to the hills of limestone, gypsum, and sand and eventually connects to the lowlands of the Persian Gulf and the Gulf of Oman. This coastal lowland region has developed around the Strait of Hormuz and has created favorable agriculture and cultivation of summer crops. Therefore, the variation in fluoride concentration in this province's drinking water may be due to these different soil properties and different geological and geochemical characteristics (Abbaslou et al. 2013(Abbaslou et al. , 2014. Using the probability distribution input data, various parameters in the model are estimated, and then based on this probability distribution, the probable risk assessment is performed ). The density plots along with the empirical distribution function related to fluoride concentration is shown in Fig. 4. The Fitdistrplus package in R software was used to adapt the fluoride data probability distributions. Figure 4 shows the degree of proximity of experimental data to a variety of known probability distribution functions. Different distribution functions are labeled with their specific mark. In this condition, the more observation point (blue dot) was closer to any mark (sign), the more probable that the data show that theorical distribution. As can be seen, the Cullen and Frey diagram predicts that the data distribution was close to the normal distribution function. Moreover, the theoretical distributions and diagrams of the cumulative distribution function on the actual data are plotted in Fig. 5. It is indicated that Weibull, and Gamma functions represented the best distributions to describe fluoride concentration data in drinking water.
The summary of statistical results of each used distribution function is presented in Table 2. One of the evaluation criteria of statistical models is the Akaike Information Criterion (AIC), which is calculated in terms of Likelihood Function (Cavanaugh and Neath 2019). Usually, the model with the lowest AIC and BIC is selected as the best model. The AIC coefficient is ordinarily positive, but what is important is the difference between the two AIC values (or better AICc), which indicates the suitability of the two models (Carvajal-Rodríguez 2018). According to Goodness-of-fit statistics and Kolmogorov-Smirnov and Cramer-von Mises statistical tests, the best distribution for the data were the Weibull distribution, but the AIC and BIC coefficients prefer the normal distribution. Under these conditions, the p value for the K-S test for the normal and Weibull distributions was 0.936 and 0.967, respectively. Therefore, both distributions were suitable for these data. Since the normal distribution is simply distribution with a certain shape, it will be the basis of subsequent calculations.
Hierarchical clustering is one of the most widely used clustering methods. Hierarchical clustering can be used to investigate similarities or differences in the space of variables such as sampling location, physicochemical and microbial parameters of water. In this clustering, the distance between the two observations firstly is calculated. After determining the distance between the two observations, due to the proximity of the observations to each other, the observations form a new cluster. This goes so far that all observations are in one cluster. This algorithms display data in the form of a tree, called a hierarchical tree dendrogram. In this study, clustering based on sampling Fig. 4 Comparison of different statistical distributions for drinking water fluoride concentrations in R software for Hormozgan, Iran location was considered. In Fig. 6, fluoride concentrations at different sampling points were categorized using hierarchical clustering based on Euclidean distance method. The results showed that of the samples in Saran barshkou and Konarejadid(well5) regions, which had the most similarities in terms of fluoride concentration, were merged, and in the later stages, the classification was completed.

Calculation of fluoride optimal concentration
According to Galagan and Vermillion's formula, the calculated regional optimal concentration of fluoride in drinking water of Hormozgan province is 0.66 mg/L, while the maximum standard fluoride concentration recommended by EPA is 1.5 mg/L. Accordingly, the fluoride concentration in all sampling points was lower than the calculated optimal concentration (Fig. 7). In the study of Zozoli et al., the optimal concentration of fluoride in all Iranian provinces were in the range of 0.64-1.04 mg/L, but in Alborz, Khuzestan, and Hormozgan provinces, the fluoride concentration was less than acceptable ranges (Zazouli et al. 2017). In another study in one of the cities of Hormozgan, there was a direct relationship between fluoride concentration and DMFT index and with increasing fluoride concentration, DMFT index also increased (Dindarloo et al. 2016).
Since fluoride can also enter human body through food and beverages, especially tea and dermal absorption, the lack of fluoride concentration in drinking water will probably be compensated. However, failure to provide the fluoride needed by the body can lead to problems such as tooth decay (Malinowska et al. 2008;Keshavarz et al. 2015).  The values of the parameters required to select the best distribution function are presented in bold using the statistical test

Human health risk estimation
EDI and HQ related to fluoride in drinking water in different cities of Hormozgan province for three age groups (children, teenagers, and adults) are presented in Table 3. The highest EDI rate belonged to Bandar Abbas in the children group (0.041) and the lowest EDI rate belonged to Bastak in the adult group (0.007). Also, the average HQ for children, teenagers, and adults in different cities of Hormozgan province were 0.48, 0.26, and 0.21, respectively, which indicated that children were the most sensitive group to fluoride exposure through drinking water. The HQ for different cities of Hormozgan province in different age groups is given in Table 3. Because the HQ obtained in all groups was \ 1 for the cities of Hormozgan province, so there was no non-carcinogenic risk of fluoride in drinking water for the residents of this province. The results also showed that the health risks of fluoride in the children group were 1.8 and 2.2 times higher than in the teenager and adult groups. In the study of Ghaderpoori et al. in the Mashhad drinking water network, the calculate HQ of fluoride for women, men, and children were \ 1

Uncertainty analysis
In the present study, Monte Carlo simulation were used to determine the quantity and uncertainty in the fluoride risk assessment process according to the input parameters.
The histogram corresponding to HQ in different groups is shown in Fig. 8. The HQ's results showed that the confidence interval in the children group was 89.4%. Therefore, there is a risk of dental fluorosis in them, but in teenagers and adults, no particular problem has been observed due to increased fluoride concentration. In studies conducted in Iran, China and India, children were identified as the most sensitive group to fluoride exposure, which was determined by the higher ratio of water consumption with respect to body weight (Ghahramani et al. 2020, Ji, Wu et al. 2020, Karunanidhi et al. 2020. Sensitivity analysis was used to determine the most effective factor in HQ for different groups. Tornado charts showed the median estimates of the spearman's rank correlation between the input variables and the non-carcinogenic effects for three groups that consumption of drinking water. As shown in Fig. 8, the concentration of fluoride in drinking water was the most effective factor in the non-carcinogenic effects of drinking water in Hormozgan province. The rate of water consumption has a positive effect on indicating higher level of health risk with increasing water consumption. However, body weigh showed negative effect since the amount of HQ decreased with increasing body weight. The results of this analysis were consistent with studies performed on drinking water fluoride in Bangladesh (Rahman et al. 2020).

Conclusion
This study investigated the distribution of fluoride in drinking water and its health risk assessment in Hormozgan province located in southern Iran. The average fluoride concentration in this province was less than the amount recommended by the EPA and the calculated optimal dose. The calculated regional optimal concentration of fluoride in drinking water of Hormozgan province is 0.66 mg/L. The highest EDI rate belonged to Bandar Abbas in the children group (0.041) and the lowest EDI rate belonged to Bastak in the adult group (0.007). Overall, the results of health risk assessment in this study showed that exposure to fluoride through water consumption in all groups was within acceptable limits. The calculated HQ was \ 1, but the health risk in the probabilistic method in children was  [ HQ adults . However, because the consumption of foods, tea, and the use of fluoride-containing toothpaste was not considered in this study, the amount of fluoride intake may be higher and may cause problems in the long period. Therefore further studies are needed. Hence, a regular monitoring program of the water source and proper education about the proper diet for the residents of this province should be done.