Exposure to fluoride occurs mainly through drinking water, which may include fluoride from natural sources and fluoride added to prevent tooth decay. Naturally occurring fluoride concentrations can vary substantially, from insignificant to well above the World Health Organization (WHO)-recommended limit of 1.5mg/L (WHO 2006), whereas the concentration in artificially fluoridated water is typically around 0.7mg/L (U.S. Department of Health and Human Services Federal Panel on Community Water Fluoridation 2015). Although low levels of fluoride are beneficial for oral health, the therapeutic range is narrow, and it is well established that individuals living in areas with high naturally occurring fluoride have an increased risk of dental and skeletal fluorosis (the former a result of early life exposure) (NRC 2006). Because of its ability to induce bone formation (Farley et al. 1983), the antifracture effect of fluoride was extensively studied in randomized controlled trials (RCTs) in the early 1990s. However, although high-dose fluoride therapy increased bone mineral density (BMD), it had no effect on the overall vertebral fracture rate, and it increased the risk of nonvertebral fractures (Haguenauer et al. 2000; Riggs et al. 1990).

A number of observational studies have assessed the association of chronic low-to-moderate exposure to fluoride, mainly via drinking water, with bone health. Although some reports suggest that fluoride may increase susceptibility to fractures (Danielson et al. 1992; Jacobsen et al. 1992; Kurttio et al. 1999; Li et al. 2001; Sowers et al. 1991), others have reported no association (Cauley et al. 1995; Feskanich et al. 1998; Hillier et al. 2000; Karagas et al. 1996; Näsman et al. 2013; Sowers et al. 2005) or evidence of a protective association (Lehmann et al. 1998; Phipps et al. 2000; Simonen and Laitinen 1985). Authors of a systematic review and meta-analysis published in 2015 concluded that chronic exposure to fluoride in drinking water was not associated with a significant increase in hip fracture risk (Yin et al. 2015). However, most of the data included in the analysis were from ecological studies with potential biases due to exposure misclassification and insufficient control of confounding.

The aim of the present study was to examine associations of urinary fluoride, and individual-level estimates of fluoride intake through drinking water and diet, with baseline BMD and fracture incidence in a population-based prospective cohort of postmenopausal women living in an area where municipal drinking water natural fluoride concentrations ranged from 0 to 1mg/L.

Discussion

In this comprehensive prospective population-based study of postmenopausal women, we explored indicators of bone health in relation to fluoride exposure in an area with natural public drinking water concentrations ranging from 0 to 1mg/L, well below the 1.5mg/L maximum level recommended by the WHO (WHO 2006). We estimated 50% (95% CI: 4, 217%) and 59% (95% CI: 11%, 230%) higher rates of hip fractures for the highest compared with the lowest tertiles of creatinine-adjusted urinary fluoride concentrations and estimated intake via drinking water and diet, respectively. Associations with all fractures and major osteoporotic fractures were positive but nonsignificant for urine fluoride, and null for dietary fluoride. Restricting analyses to women whose drinking water fluoride concentrations were likely to be constant from 1982 to baseline (2004–2009) strengthened the associations between urinary fluoride and the risk of fractures during follow-up. In addition, the adjusted mean BMD was slightly higher among women with higher urine and dietary fluoride levels at baseline. Altogether, these findings suggest that daily high consumption of tap water and beverages based on tap water [i.e., approximately 10 (±3) servings of 150–200mL/d] with fluoride content of ?1mg/L may increase both BMD and bone fragility.

Fluoride has a potent effect on bone cell function, bone structure, and bone strength. Hydroxyapatite, the most abundant inorganic component of bone and the source of its rigidity, is converted to fluorapatite when fluoride ions replace hydroxyl ions. Fluorapatite is harder and more resistant to acidic mineral dissolution than hydroxyapatite (NRC 2006). Fluoride can also induce bone growth through stimulation of osteoblasts, an effect especially evident at the lumbar spine (Farley et al. 1983; Haguenauer et al. 2000). In the 1990s, the potential antifracture effect of fluoride on bone was extensively explored in RCTs (Haguenauer et al. 2000). However, the results were disappointing because fluoride failed to reduce vertebral fracture rates and, instead, increased the risk of nonvertebral fractures at high doses (>30mg/d) (Haguenauer et al. 2000; Riggs et al. 1990). The increased risk was suggested to be attributed reduced elasticity and strength of the newly formed bone (Fratzl et al. 1994; Riggs et al. 1990).

In the present study, urinary fluoride excretion and estimated dietary fluoride intake were associated with increased BMD, with a stronger association for lumbar spine BMD than femoral neck BMD. The highest tertiles of urine and dietary fluoride were also associated with an increased risk of hip fractures. Although our estimates suggest a very small effect on BMD, they are in line with effects reported for therapeutic doses of fluoride in RCTs (Haguenauer et al. 2000; Riggs et al. 1990). The stronger positive association for lumbar spine than femoral neck density has been suggested to reflect differences in the effects of fluoride on trabecular vs. cortical bone (Riggs et al. 1990). It can be argued that the results of RCTs, based on very high fluoride doses during a limited period (<4y), may not be translatable to long-term low-to-moderate exposure in real-life settings. However, because fluoride accumulates in bone (NRC 2006), low-dose exposures over extended periods also may be sufficient to produce adverse effects. Our finding of stronger associations with urine fluoride when restricted to women likely to have consistent drinking water fluoride levels for at least 20 y before baseline further supports the possibility that long-term exposures may have adverse effects, even when drinking water fluoride concentrations are below recommended limits.

Most observational evidence regarding the effects of fluoride on bone has been based on fracture rates or BMD in relation to ecological assessments of fluoride concentrations in drinking water, and results have been discordant, with some studies reporting no association between drinking water fluoride and bone parameters (Cauley et al. 1995; Feskanich et al. 1998; Hillier et al. 2000; Karagas et al. 1996; Näsman et al. 2013; Sowers et al. 2005) and others suggesting increased fracture risk or changes in BMD (Danielson et al. 1992; Jacobsen et al. 1992; Kurttio et al. 1999; Li et al. 2001; Sowers et al. 1991). We are aware of only two previous studies that used individual-level biomarkers to assess fluoride exposure (Feskanich et al. 1998; Sowers et al. 2005). A nested case–control study within the Nurses’ Health Study cohort (Feskanich et al. 1998) found no association between toenail fluoride concentrations and self-reported forearm or hip fractures in 241 matched case–control pairs. Similarly, a prospective study of 1,300 women residing in three U.S. communities with contrasting fluoride levels in drinking water (1–4mg/L

) found no association between serum fluoride concentrations and baseline BMD or the risk of self-reported fractures over 4-y of follow-up (Sowers et al. 2005). Discrepant findings may be related to the use of different biomarkers, with urine and serum reflecting short-term exposures, whereas toenail clippings reflect exposure over several months. Urine is the most frequently used biomarker and, excluding fluoride concentrations in dentin and bone, is considered the most valid (EFSA Panel on Dietetic Products, Nutrition, and Allergies 2013). Discrepancies may also be related to differences in study design, fracture ascertainment, sample sizes, and study population characteristics.

Strengths of the present study include its population-based prospective design, range of fluoride exposures, and use of two different measures of fluoride exposure. The almost complete ascertainment of cases via register-linkage, and the use of DXA measurements to assess BMD are other strengths. Moreover, bottled water consumption is very low in Sweden, which had the lowest estimated bottled water consumption per capita of 25 EU countries in 2019 (10L/y, compared with 118L/y for the EU as a whole) (Conway 2020). In addition, an online survey of >1,000

Swedish adults indicated that tap water accounted for 78% of nonalcoholic beverage consumption in 2013 (Säve-Söderbergh et al. 2018). Thus, although some degree of exposure misclassification is inevitable, bottled water consumption is unlikely to have a substantial effect on the accuracy of estimated intakes based on municipal drinking water fluoride concentrations. Finally, although we cannot rule out the possibility of residual confounding, we had detailed information on a number of factors relevant to bone health and fluoride exposures and adjusted for them as potential confounders.

Self-reported information on dietary habits is inevitably associated with some degree of measurement error. Nevertheless, tap water, coffee, and tea, which are likely reported with higher precision than other dietary items (Wolk et al. 1997), accounted for 78% of estimated dietary fluoride intake in our study. In addition to estimating dietary intakes, we used urine fluoride excretion to classify exposure. Using a single urine sample is a limitation, given that the circulating half-life of fluoride is short (Buzalaf and Whitford 2011). Fluoride release due to bone degradation also may have increased urinary fluoride excretion in some women. However, we adjusted for plasma Beta-CrossLaps, a biochemical marker of bone resorption, to limit this potential source of confounding. The correlation between urinary and dietary fluoride was moderate (rho=0.37). There may be several reasons underlying this observation, including that urinary fluoride is a biomarker of short-term exposure, whereas the FFQ reflected average consumption during the past year. Finally, our study population consisted solely of middle-aged to elderly women in Sweden, and thus, our findings may not apply to other population subgroups or to women in other countries. Notwithstanding this potential limitation, our results provide insights into the association of fluoride exposure with bone health in postmenopausal women—the group where the burden of hip fractures is the largest.

Conclusion

The risk of hip fractures was increased among Swedish women who had the highest levels of urine fluoride excretion and the highest estimated fluoride intake from beverages and food relative to women with the lowest levels of each exposure. Our findings, which are consistent with the effects of high fluoride exposures observed in RCTs (resulting in a denser but more fragile skeleton), suggest that long-term consumption of tap water with a fluoride concentration of 1mg/L, which is below the 1.5mg/L maximum concentration recommended by the WHO, may adversely affect bone health in postmenopausal women.