Association Between Osteoarthritis and Water Fluoride Among Tongyu Residents, China 2019: A Case-control of Population-based Study

Background: Fluoride is an environmental chemical that has adverse effects on health, probably increasing osteoarthritis (OA) risk. However, whether uoride is a serious risk factor for OA is still inconsistent among current evidence. The aim of this study was to determine the relationships between chronic uoride exposure and OA risk among the residents living in Tongyu County, China 2019. Method: 186 cases and 186 controls aged 26 to 86 years were recruited after x-ray examination with Kellgren and Lawrence criteria by two independent evaluators. Urinary uoride (UF) in a spot sample was measured by uoride ion-selective electrode, and water uoride data were used as community level of exposure. Associations between uoride exposure and OA risk were examined by the unconditional logistic regression. Covariates included age, gender, body mass index, duration of living, daily water drunk, income, sport and lter use with stratied analysis. Results: mg/L was higher than the mean of community water uoride concentration signicantly higher in cases (p full sample a 1mg/L increase in UF was associated with a 27% higher risk of OA (95% CI: 1.06 - 1.52, p 0.008), and 4 th quartile participants were associated with higher risk when compared to 1 st quartile (OR: 2.46, 95% CI: 1.34 – 4.57, p 0.003). In compared to 1 st quartile, 4 th quartile participants were 4 times more likely to have OA (95% CI: 1.86 – 8.82, p 0.000) in the non-obese group, 7.7 more likely to have OA (95% CI: 2.58 – 25.05, p = 0.000) among adults ≤ 60-year and 12 times more to have OA (95% CI: 2.15 – 99.65, p = 0.008) in non-obese adult women ≤ 60-year group. Conclusion: Water uoride exposure may increase OA risk and could have more impact on a specic group such as non-obese and adult ≤ 60-year population.


Introduction
Fluoride is an element that can be found in the air, rocks, soil, and water and it is both bene cial and harmful to our body [1][2]. It is a leading cause of uorosis which is a global public health problem. The global prevalence of uorosis is still unclear, but the World Health Organization estimated that excessive uoride concentration in drinking water has caused tens of millions of cases of skeletal and dental uorosis around the world over one year period [3]. Fluorosis is worldwide in distribution and endemic in approximatively 50 countries including Africa, Asia, Europe, North and South America [4]. Fluorosis is a crippling disease that results from the deposition of uoride in the hard and soft tissues of the body; it is caused by excess intake of uoride through drinking water/food products/industrial pollutants over a long period [5][6][7][8]. Skeletal uorosis (SF), known as musculoskeletal disease is endemic in China, India, Mexico, and Argentina, and is generally linked to drinking well water or tea with high uoride [9]. SF manifests by chronic joint pain, backache, stiffness and rigidity of the spine, calci cation of ligaments, physical limitations, inadequate labor capacity, and disability according to the stage of evolution [10][11][12][13]. These symptoms are very similar to osteoarthritis (OA), making diagnosis more di cult. As Roschger's team concluded in 1995, SF can be a very sneaky disease highlighting the di culty of its diagnostic [14]. Many observations made in the endemic areas suggested that uoride can damage articular cartilage and even worsen OA's symptoms, and the most recent and of particular importance was done in Turkey 2001 [15][16]. OA is the most chronic and prevalent ageing joint disease which does not have an effective treatment proven to delay disease progression. Individual with OA experiences pain, stiffness, swelling, and disability [17][18]. According to the ndings from the Global Burden of Disease Study 2017, approximately 61.2 million individuals were suffering from OA in China [19].
Fluoride has a well-known adverse effects on health, but the relationship between uoride exposure and OA occurrence is not yet established. However, several studies have provided clues to a probable link between uoride and OA. Thus, some works including radiological analysis studies and total arthroplasty studies for the treatment of hip and knee severe OA due to uorosis reported that extensive degenerative changes in articular cartilage could occur in a patient with uorosis [15,[20][21][22]]. An ecological study in uorosis-a icted area (China) found that, the incidence of OA in the uorosis area was remarkably higher than in either the adjacent non-endemic area or the nation as a whole [23]. Additionally, another ecological study in which 56 endemic uorosis patients were matched to age and sex 40 non-endemic control patients suggested that endemic uorosis may increase the severity of knee OA and cause OA before SF is obvious. According to study, the radiological severity of knee OA and osteophytes sign were signi cantly higher in endemic uorosis group than in controls group [16]. Meanwhile, US health authorities still assume that uoride does not cause arthritis symptoms before the traditional bone changes (osteosclerosis) of uorosis are evident on x-ray [24]. Moreover, the US National Research Council review concluded that only uoride at high therapeutic doses can cause nodules in articular cartilage not at environmental doses [25]. This because of the con icting results showed in two reports of uoride at therapeutic doses in rheumatoid patients. In one report [26], uoride exacerbated rheumatoid arthritis symptoms, but, it was well tolerated in another case [27]. Despite all the observations and reports on the possible association between uoride and OA, it appears that the evidence is not consistent enough and still at hypothesis generating level.
The belief that uoride is associated with OA has not been fully explored and the few works found in literature are only ecologic and case-report studies. Therefore, to address this weakness, we conducted a population-based case-control study in 2019 with purpose of evaluating whether uoride chronic exposure is associated with the risk of OA in population in order to strengthen epidemiological evidence.

Study population and design
A population-based case-control study with two-step recruitment was carried out in Tongyu County (Baicheng city), Jilin province, as one of uoride endemic areas in China. In the rst part of the recruitment, four towns in Tongyu County were investigated from November 2019 to January 2020 using cluster sampling. We used strati ed random sampling to recruit 26 -86 years old permanent residents at least 10 years (640 participants). All participants were born and raised in the local area. Subjects with incomplete data, rheumatoid arthritis, prior joint injury and trauma (22 participants) were excluded. In the second part, 372 participants from the remaining 618 were selected according to the x-ray examination by two independent evaluators from two different regions. We de ned a subject as having radiographic OA when the Kellgren and Lawrence grade was ≥ 2 in at least 1 knee or 1 elbow. The rst 186 OA cases and 186 controls diagnosed at the same time by both radiologists were retained (see additional le, gure 1). Sample size was calculated from an online Open Source Epidemiologic Statistics for Public health [28], with 59.03% as OA prevalence from the uorosis-a icted area in China [23] with the desired con dence interval (CI) of 95%.
Collection of data, biological and radiological samples Participants were investigated through face to face interview using a structured questionnaire administered uniformly by trained study physicians. The questionnaire included: demographic information, anthropometric measures, drinking water information, health status, diet and behaviors; calcium and vitamin D supplementation information.
About 50mL spot urine samples (non-standard collection and rst-catch urine in the morning) were collected from each participants in precleaned, labelled polythene tubes. All samples were kept in a cooled ice box and then sent immediately to laboratory. Samples were stored at -20 •C until analysis.
For all consenting participants, a single standing, anteroposterior radiograph of both knee and elbow was performed in the Baicheng central hospital. We selected knee and elbow OA subtype in this study because both joints sites are the most frequently reported in the endemic area [16,29]. Also knee OA is the most frequent in normal situation and again with severe complications [30][31].

Fluoride exposure analysis
The 2019 national surveillance analysis of endemic uorosis showed that uoride concentration for Tongyu communities' water samples ranged from 0.94 to 2.30 mg/L (1.49 mg/L as mean) while China's national standard limit is 1.2mg/L (GB5749-2006). The population of Tongyu is mainly supplied by 5 water sources namely: 2 public tap water, 1 public shaft water, personal wells and bottled water. An individual's delity to a single water source is very rare and residents arbitrarily change their water source up to 3 in the same year or after a period of time. Hence we considered UF as the consistent and reliable parameter for measuring uoride exposure at the individual level. UF has been shown to be an accurate evaluation of uoride ingestion on a population basis [32].
Urine uoride content was measured at the Key Lab of Etiology and Epidemiology, Center for Endemic Disease Control, Harbin Medical University by uoride ion-selective electrode. The operating process was performed according to the China standard for determination of uoride in urine [33]. We added 5 mL ionic strength adjustment buffer to each urine sample (5 mL) for controlling the pH of the solution at 5.0 -5.5 to optimize the determination conditions. Samples were retested twice and the mean value of each sample was used for analysis.
Community water uoride concentration 2019 data was obtained from the local Center for Endemic Disease Control of Baicheng city, Jilin province China and served as an ecologic measure of exposure.

X-ray examination
The X-Ray lms were evaluated by an experienced radiologist. Each knee and elbow was evaluated for the presence of lateral or medial osteophytes, joint space narrowing, sclerosis and cysts. Both joint were also graded for overall evidence of radiographic OA according to the Kellgren and Lawrence criteria (grade 0 to 4, where 0 = none; 1 = possible osteophytes only; 2 = de nite osteophytes and possible joint space narrowing; 3 = moderate osteophytes and or de nite joint space narrowing; and 4 = large osteophytes, severe joint space narrowing, and or bony sclerosis) [34]. We de ned cases as radiographic OA if they had a Kellgren/Lawrence grade of ≥ 2. The radiologist had no knowledge about the participants ages and their names.
To ensure the reliability of the diagnosis, all lms were brought to the radiological unit of the 2 nd a liated hospital outpatient's department of Harbin Medical University. They were then read by an experienced, academically based bone and joint radiologist using the same Kellgren and Lawrence criteria. Only the lms with OA diagnosis from both radiologists were accepted and considered for this study.

Potential confounders and effect modi ers
Several studies have found age, gender and obesity to be clearly associated with the occurrence of OA as person level risk factor including a recent systematic and meta-analysis study which included 88 studies [30,[35][36][37]. They reported that OA risk increases in female group, obese group and with age. Some dietary factor such as calcium and vitamin D have been suspected to be associated with OA [37]. Calcium and vitamin D supplementation information were collected by the questionnaire.

Ethics statement
This study was approved by the Ethical Review Board of Harbin Medical University (HMUIRB20120021).
The study has obtained the necessary approvals from the authorities of Baicheng city. Written informed consent was obtained from each participant of the study population.

Statistical analysis
All statistical analyses were performed using R software version 4.0. 3 [38]. Two-tailed p < 0.05 was used for all tests as Signi cance Level.
The Body Mass Index (BMI) was calculated according to height and weight by the formula: BMI (kg/m2) = weight/ height^2. Population characteristics of cases and controls were compared applying Student's t-test for continuous variables and Chi-squared test for categorical variable. We presented the results as mean with standard deviation (SD), number and percent as appropriate. Odds ratio (OR) along with their 95% CI were derived from all logistic regression models and presented. Based on the characteristics of the study population and the literature, age, gender, BMI, duration of living, daily water drunk, income, sport and lter use were selected as confounding and or third factors.
Unconditional logistic regression was used to assess the association between uoride exposure and the risk of OA. To have a better understanding of the quantitative relationship between uoride exposure and OA risk, we used UF as continuous variable, dichotomous variable (with 2.38 mg/L median concentration as cut-off value) and UF as ordinal variable (UF quartile: 1 st Q group from 0.35 to 1.61 mg/L, 2 nd Q group from 1.61 to 2.38 mg/L, 3 rd Q group from 2.38 to 3.30 mg/L and 4 th Q group from 3.30 to 7.01mg/L) in simple logistic regression analysis. Subsequently, we kept UF as an ordinal and continuous variable in multiple logistic regression analysis.
To limit confounding/effect modi ers in uence and explore the independent effect of uoride exposure, we performed strati ed analysis by gender, age, and BMI, and UF was used as an ordinal variable. Multiple logistic regression analysis was done in full sample and strati ed group according to gender, age and BMI. Age was divided into two group, adult ≤ 60-year old and adult over 60-year. BMI variable was divided into two groups no-obese and obese using Chinese diagnosis criteria of obesity (noobese BMI < 27 kg/m2 and obese BMI ≥ 27 kg/m2). To evaluate the magnitude of uoride effect in these speci c groups, we performed further strati ed analysis in non-obese adult women ≤ 60-year (age ≤ 60 years & BMI<27 kg/m2 & female sex). Due to collinearity issue, we did not include age and duration of living in the model at the same time.

Results
Demographic characteristics A total of 372 participants from which 186 cases and 186 controls were enrolled in this study. Mean UF concentration 2.54 ± 1.22 mg/L was higher than the mean of community water uoride concentration 1.49 ± 0.32 mg/L. The mean UF concentration 2.73 ± 1.18 mg/L was signi cantly higher in cases compared to 2.35 ± 1.24 mg/L in controls group (p < 0.02). Besides, the mean age was signi cantly higher in cases compared to that in controls (63.22 ± 7.71 years versus 58.77 ± 10.23 years). The population characteristics in cases and controls were summarized and compared in Table 1. Sample descriptive statistics for uoride exposure and OA subtype rate are shown in Table 2 and 3 respectively. Assessing association between uoride exposure and OA outcome In simple logistic regression analysis shown in Table 4, from model 1(UF as continuous variable), higher UF concentrations were associated with higher odds of getting OA (OR: 1.30, 95% CI: 1.09 -1.55, p = 0.003). In other words, a 1mg/L increase in UF level was associated with 30% higher risk of getting OA diagnosis in this sample. Model 2 (UF as dichotomous variable) shows that high level group (HLG) subjects were associated with 68% higher risk of getting OA as compared to low level group (LLG) subjects (OR: 1.68, 95% CI: 1.11 -2.53, p = 0.013). For model 3 (UF as ordinal variable), the risk of getting OA increased with increasing uoride concentration by category (≤ 1.61, 1.61 -2.38, 2.38 -3.30, >3.30 mg/L). Among the 2 nd , 3 rd , and 4 th quartiles, the odds of getting OA for the 3 rd and 4 th Q group was nearly 2 times and more than 2.5 time the odds as compared to the 1 st Q group reference respectively (OR: 1.87, 95% CI: 1.05 -3.38, p = 0.034 and OR: 2.55, 95% CI: 1.42 -4.63, p = 0.001).
In the multiple logistic regression analysis (

Evaluation of gender, age and BMI-speci c association
In strati ed analysis adjusted for covariates shown in Table 7, sex did not modify the association between uoride exposure and OA. The 4 th Q participants were associated with higher risk of OA in female as well as in male group when compared to 1 st Q participants after adjustment (OR: 2.18, 95% CI: 1.01 -4.79, p = 0.048; OR: 4.76, 95% CI: 1.59 -15.30, p = 0.006 respectively). Before and after adjustment for covariates, body weight modi ed association between uoride exposure and OA such that 4 th Q participants were associated with higher risk of getting OA when compared to 1 st Q participants among non-obese group (OR: 4.21, 95% CI: 2.07 -8.84, p = 0.000; OR: 3.99, 95% CI: 1.86 -8.82, p = 0.000 respectively). Age modi ed association between uoride exposure and OA in such a manner that, before adjustment 4 th Q participants were 6 times more likely to have OA and 7.7 times after adjustment as compared to 1 st Q participants in adult ≤ 60-year group (OR: 6.00, 95% CI: 2.34 -16.41, p = 0.000; OR: 7.69, 95% CI: 2.58 -25.05, p = 0.000 respectively). We did not nd any signi cant association between uoride exposure and OA in obese and adult over 60-year group.
To explore the magnitude of uoride effect in these speci c sub-groups, we performed further strati ed analysis in non-obese adult women ≤ 60-year (

Discussion
To the best of our knowledge, this is the rst case-control study exploring the relationship between chronic water uoride exposure and OA outcome in human. To date, only two ecological studies [16,23] really attempted to link uoride exposure to the induction of OA, showing how much data is lacking to discuss this topic. We examined the association between uoride and OA risk among residents living in Tongyu County (Baicheng city, Jilin province China), an endemic area where water uoride ranged from 0.94 to 2.30 mg/L, and adjusted for factors that can in uence uoride exposure/metabolism as well as OA outcome. Instead of water uoride exposure rate at individual level, we were able to assess exposure at community level and used UF as biomarker due to the long-term exposure with unchanged residence place. UF has been demonstrated as a precise assessment of uoride ingestion on a population basis [32].
We found in logistic regression analysis that a 1mg/L increase in UF level was associated with a 27% higher risk of getting OA disease after adjustment. With UF as an ordinal variable, OA risk increased with increasing uoride concentration by category. The participants in the 4th quartile were associated with a higher risk (OR = 2.55) of getting OA disease as compared to the 1st quartile group reference. The association remained signi cant after adjustment for confounding (OR = 2.46). These results suggest that uoride exposure from water source could be a serious independent predictor of OA, particularly as UF concentration increases in an individual, the risk of developing OA increases. Given the scarcity of data on the topic, additional and cohorts studies are needed to carefully explore this eventuality.  [39], a case of 74 years old female who was on uoride therapy for osteoporosis for 30 months had developed a layer of calci ed cartilage containing 3.9 mg/kg by ash weight in her femoral head. This underlines the possibility that excess uoride can cause damage to the joints. Likewise, in a study evaluating patient's groups with a greater number of subjects, Duell and Chesnut [26] found that the use of uoride at therapeutic doses in rheumatoid patients exacerbated symptoms of rheumatoid arthritis. Another explanation may be the inhibition of osteoblast cell activity. Fluoride stimulates bone cell proliferation by direct inhibition of osteoblastic acid phosphatase activity [40] and by enhancing the mitogenic signals of growth factors [41][42]. The activity of osteoblast cell produces a huge increase in bone formation at the organ level, producing exostoses, calci cation of tendons and ligaments, and osteosclerosis [42]. In the most recent time study from Savas et al. [16], it was demonstrated that patients with skeletal uorosis had a greater severity of knee OA symptoms and osteophyte formation than age and sex-matched control group patients. In our study, OA proportion increased as UF level increased by category. Everything suggests that people exposed to uoride have an additional risk of developing OA, even if the entire mechanism is not yet clear. The development of OA relies on an interaction between several factors and so this process may be considered as the nal product of an interplay between systemic and local factors, genetics and imbalance in the physiological process [43][44], which may give the possibility to uoride to play a certain role knowing that its target tissue is bone and cartilage [25,39].
Any increase of OA risk in the obese group, female group and with ageing could be of particular concern because they have been established with an elevated risk [35][36][37]. We also explored group speci c association and found that a non-obese adult woman ≤ 60-year with UF > 3.30 mg/L is associated with 12.55 times greater odds of getting OA disease as compared to one with UF ≤ 1.61 mg/L. This suggests that, being an adult woman ≤ 60-year and non-obese at the same time is a factor that could allow chronic water uoride exposure to considerably increase OA risk. Given the lack of studies on the topic, further investigations are needed to draw any conclusion, although association has been estimated in more detail in each group.
After dividing age into two groups of adult ≤ 60-year of age and adult over 60-year, we found an increased risk of OA in adult ≤ 60-year group. Before adjustment, 4th quartile participants were 6 times more likely to have OA when compared to 1st quartile participants. The association remained even more stronger and signi cant after adjustment (OR: 7.7). This surprising association could suggest that, an adult under 60-years old is at increased risk (sevenfold) of developing OA once exposed to uoride as compared to an adult over 60-years old. Bone uoride concentration tend to increase with age due to the continuous accumulation over time [45][46]. The potential reason for this is the preferential removal of crystallites with little or no uoride in the elderly [25]. One would have expected an increased risk of OA in adults over 60-years age group as a result. May be the wear and tear of joints structure due to ageing is too evolved to be affected by the additional effect of uoride, perhaps there is an intervention of other unknown factors.
In the non-obese group analysis, the 4th quartile participants were 4.21 times more likely to have OA as compared to the 1st quartile participants. Association remained signi cant after adjustment with 3.99 greater odds of getting OA in 4th quartile participants. Meanwhile, no signi cant association was observed in the obese group. This result could mean that, in the uoride exposure context, only the nonobese group is at increased risk of OA. It seems that obesity is a stage of impairment where the uoride effect does not have too much in uence. However, additional studies are needed for more exploration as data are sparse. Fluoride adverse effect depends on the magnitude and the length of exposure, and how it behaves in the body, whereas the mechanisms underlying its metabolism and biological effects are not clearly understood yet [47]. Any environmental, biochemical, physiological, and pathological condition which interferes with the absorption or excretion of uoride, will in uence its destiny in the body and may ultimately increase the risk of musculoskeletal disorders [48].

Limitations
One limitation of our study is the measurement of exposure rate at the individual level. Even if we measure the concentration of UF for every participant, this may not be directly related to the community water. Of course, the source of exposure is from community water, but there might probably be some unidenti ed additional sources of exposure contributing to the exposure rate at the individual level as demonstrated by the difference between mean UF (2.54 ± 1.22 mg/L) and mean community water uoride (1.49 ± 0.32 mg/L). At this point, we were unable to measure the other sources of exposure such as: toothpaste, consumed foods and products locally made, and tea consumption. Different level of exposure should be taken into account in futures investigations. We observed a close relationship between uoride exposure and OA outcome with some group speci c associations (non-obese and adult ≤ 60-year group). However, this may not mean any causal inference. Therefore, these results should be interpreted with caution. Also, the cross-sectional nature of our study does not allow us to easily state a direction of the association between uoride exposure and OA outcome. Given this limitation and the scarcity of evidence on the topic, our ndings should be viewed along with others and as hypothesis testing. Ultimately, we recommend more prospective studies with a large sample size as possible to deeply explore the in uence of uoride exposure on OA outcome.
The strength of our study is that we used radiographic OA rather than self-report. We also provided data on OA rate at the individual level in uoride endemic areas as well as exposure rate at the individual level (UF concentration) with regards to the community level of exposure. Considering an association between uoride and OA, our study design is better than the previous ones to address this issue. Finally, it is commonly known that age, gender, and body weight are personal level risk factors for OA [35][36][37]. We believe that, by performing strati cation analysis on these variables, we had overcome the issue of confounding and effect modi cation, at least to a certain extent.

Conclusion
Fluoride may be a serious independent risk factor for OA disease despite taking into account other factors such as age, gender, and body weight. Our ndings suggest an additional effect on the risk of OA, particularly in a non-obese adult woman ≤ 60-year where OA risk is twelvefold. However, our results should be interpreted with caution and we recommend other large-scale cohort studies. Nonetheless, we would like to raise awareness of the healthcare professionals on the possible existence of uorotic osteoarthritis, especially in endemic areas. Availability of data and materials The dataset used and analyzed in this study is available from the authors in reasonable request.

Ethics approval and consent to participate
The study was approved by the Ethical Review Board of Harbin Medical University (HMUIRB20120021). Written informed consent was obtained from each participant of the study population.

Consent for publication
Not applicable.
47. Meng X, Yao Y, Ma Y, Zhong N, Alphonse S, Pei J. Effect of uoride in drinking water on the level of 5methylcytosine in human and blood. Environ Toxicol Pharmacol.