The fluoride concentration of tap water samples was determined directly using a fluoride ion-selective electrode (Orion Research, model 96–09) upon adding TISAB III. The rest of the food/meal samples were analyzed using the HMDS-facilitated diffusion method (52, 53). All samples were analyzed in triplicate. The fluoride concentration (µg/g) of each sample was then calculated as the average of each triplicate. Additionally, semi-skimmed milk (UK equivalent of 2% milk) and whole milk samples were analyzed using a microwave-assisted acid digestion procedure, which served as a pre-treatment to the direct method of fluoride analysis to allow for the complete extraction of fluoride from food where fluoride might exist bound to cations that prevent detection of fluoride ions (54). In short, 1 g of each sample was weighed in a Teflon vessel, covered with 8 ml of 7 mol/L HNO₃, and subjected to microwave digestion for 15 min at 180°C. Once cooled down, samples were neutralized with NaOH (8 mol/L and 1.8 mol/L stock solutions) before direct fluoride analysis.

A hypothesis indicating that the fluoride concentration of a meal would match the sum of the fluoride concentrations of individual food samples used for its preparation (expected value) was developed. Expected values were calculated using Equation 2(2)E=∑i=1nRi*Fiwhere

E = expected fluoride content of meal (µg/g),

R_i = ratio of the i-th food sample used to create a meal,

F_i = fluoride concentration of the i-th food sample used to create a meal (µg/g),

n = number of food samples used to create a meal.

One hundred and three food/meal samples were duplicated and subjected to a standardized static in vitro digestion procedure that mimics oral, gastric, and small intestinal phases of human digestion (55, 56). Before digestion, chewing of solid food/meal samples was mimicked using a household blender (IMURZ, TC-18), converting food into small enough particles that would be safe for swallowing. Five grams of each sample were then placed into 50 ml centrifuge tubes. The oral phase involved the addition of human salivary α-amylase (Sigma-Aldrich, A1031) and simulated salivary fluid (SSF) followed by a short incubation of approximately 2 min accompanied by manual stirring. The gastric phase of the digestion began with the addition of pepsin (Sigma-Aldrich, P7012) and pH adjustment to∼3.0. Incubation was performed at 37°C for 2 h at 150 rpm in an orbital shaker/incubator (Sartorius Stedim Biotech GmbH, CERTOMAT® BS-1). The final phase of the digestion included the addition of bile (Sigma-Aldrich, B3883) and pancreatin (Sigma-Aldrich, P7545), adjustment of the pH value to∼7.0, and incubation at 37°C for 2 h at 150 rpm in the orbital shaker/incubator. Once completed, digestion was stopped by placing the centrifugation tubes in ice, which stopped enzymatic activity. Samples were centrifuged (Eppendorf, 5810 R) at 3,900 rpm for 20 min and the resultant supernatants were stored frozen until further analysis. Fluoride concentrations of supernatants were determined in triplicate using the HMDS-facilitated diffusion method (52, 53).

Alongside food/meal samples, NaF standards of different concentrations (0.1, 1.0, and 10.0 µg/ml) were subjected to the same in vitro digestion procedure described above, as earlier bio-accessibility studies indicated different fluoride bio-accessibility values detected between gastric and whole gastrointestinal digestion (40, 57). To examine this further, standards were analyzed using (a) in vitro digestion consisting only of the oral and gastric phases of the digestion, and (b) the whole gastrointestinal digestion process.

Final fluoride bio-accessibility was calculated using the following equation:(3)Bio-accessibility(%)=F2F1where

F₁ = fluoride content of the sample determined before in vitro digestion (μg/g), and

F₂ = fluoride content determined after in vitro digestion (bio-accessible fraction) (μg/g).

To check the validity of the analytical method, a known concentration of fluoride standard (NaF) was added to approximately 10% (n = 11) of the samples. These samples were then analyzed in triplicate to measure the recovery of added fluoride. To check reliability, fluoride measurement of approximately 20% of samples (n = 24) was repeated. Samples were analyzed in triplicate and double-distilled, deionized water was used as a blank.

To compare laboratory-measured and expected fluoride concentrations of meals, a Bland-Altman plot was created (58) using RStudio 4.2.2. This procedure is a standard approach that allows comparison between two methods of measurement, usually a new method and an already established one, and it helps with the decision of whether the new method is acceptable. The analysis involved plotting the average of two methods against the difference between them, enabling a visual assessment of their agreement. The methods are considered to show good agreement if the points on the scatterplot lie close to the middle line.

Details on fluoride concentration (µg/g) of individual food samples are found in Supplementary Table S1. The overall mean (SD) fluoride concentration of the 36 individual food items was 0.518 (1.811) µg/g. Ten food samples analyzed as representatives of infancy resulted in a mean (SD) fluoride concentration of 0.136 (0.095) µg/g ranging from 0.011 µg/g to 0.249 µg/g for powdered infant formula and vegetable-based baby foods, respectively. Food samples regularly consumed during early childhood (n = 26) resulted in a mean (SD) fluoride concentration of 0.665 (2.123) µg/g ranging from 0.023 µg/g to 10.730 µg/g for fruit and canned fish in tomato sauce, respectively.

Food samples available in the UK and US markets were also analyzed in this study (Table 1) indicating mean (SD) fluoride concentrations of 0.097 (0.093) µg/g and 0.195 (0.089) µg/g for the UK and US market, respectively.

Sixty-seven analyzed meal samples were grouped according to preparation method. The overall mean (SD) laboratory-measured fluoride concentration of meals was 0.140 (0.213) µg/g while the overall mean expected value (SD) equaled 0.152 (0.219) µg/g. When groups were examined, the highest laboratory-measured mean (SD) fluoride concentration was detected for meals created with water: 0.442 (0.360) µg/g. The corresponding expected fluoride concentration was almost identical: 0.439 (0.379) µg/g. A graphical comparison between laboratory-measured and expected fluoride concentrations for each group is found in Figure 1. Further details are available in Supplementary Table S2.

Figure 2 shows a Bland-Altman plot, where the average of the two methods is shown on the x-axis while the difference between the two is plotted on the y-axis of the scatterplot. The analysis was conducted using “laboratory-measured” and “expected” fluoride concentrations (µg/g) for 67 meal samples. The average difference (−0.012 µg F/g) is represented by the middle line on the scatterplot while the upper and lower lines represent the limits of agreement (LoA). The upper and lower LoAs (0.186 and −0.210 µg F/g, respectively) were calculated as average difference ± 1.96*standard deviation of the difference. The plot shows a small mean difference between the measured and expected values and narrow LoAs, indicating a good agreement between the methods.

When small intestinal digestion was excluded from the procedure, the measured fluoride bio-accessibilities were 64.7, 77.6, and 100.1% for the 0.1, 1.0, and 10.0 µg/ml fluoride standards, respectively. The corresponding whole gastrointestinal digestion results were: 56.9, 60.0, and 66.2%, respectively.

Details on fluoride bio-accessibility (%) of individual food samples are found in Supplementary Table S1. The overall mean (SD) fluoride bio-accessibility of the 36 individual food items was 44.7% (37.5%). Ten food samples analyzed as representatives of infancy resulted in a mean (SD) fluoride bio-accessibility of 24.2% (11.8%) ranging from 0.1% to 46.7% for powdered infant formula available on the UK market and fruit-based baby food available on the US market, respectively. Food samples regularly consumed during early childhood (n = 26) resulted in a mean (SD) fluoride bio-accessibility of 52.6% (41.1%) ranging from 1.0% to 152.3% for canned fish in tomato sauce and jam, respectively.

The mean (SD) fluoride bio-accessibilities for meals when consumed with juice, carbonated drinks, tap water, and milk were 79% (21.9%), 64.3% (20.7%), 40.2% (20.9%), and 102.0% (75.8%), respectively. For the rest of the meals, the average fluoride bio-accessibility was 50.8% (55.9%) (Supplementary Table S4). Figure 3A shows higher fluoride bio-accessibilities recorded for those meals prepared with fluoridated tap water compared with those prepared with non-fluoridated water. Furthermore, all meals shown in Figure 3A resulted in fluoride bio-accessibilities below 100%. Figure 3B shows higher fluoride bio-accessibilities for a majority of meal samples created with juice when compared to those created with carbonated beverages with a notable outlier resulting in 127% bio-accessibility when digestive cookies are consumed with juice. High fluoride bio-accessibilities were also recorded for meals created with whole milk (Figure 3C). Overall, four outliers with fluoride bio-accessibilities above 150% were detected – (a) white bread with peanut butter and jam mixed with both types of milk and (b) white bread with peanut butter mixed with both types of milk. When those were removed, the overall mean (SD) fluoride bio-accessibility of meals when consumed with milk was 71.5% (17.1%). Finally, Figure 3D highlights two outliers – (a) white bread with peanut butter and jam and (b) white bread with peanut butter that resulted in fluoride bio-accessibilities of 164.0% and 129.1%, respectively.

The mean (SD) fluoride concentration of individually measured food items was 0.518 (1.811) µg/g. Fluoride concentrations of individual food samples analyzed in this study and consumed during infancy are consistent with results found in the literature. For example, a study conducted in Canada in 1982 (69) revealed the fluoride concentration of vegetable-based baby foods between 0.02 and 1.29 µg/g, fruit-based baby foods 0.04–0.39 µg/g, meat-based baby foods 0.05–3.93 µg/g, and cereals 1.24–4.98 µg/g. Heilman et al. (70) reported the fluoride concentration of vegetable-based baby foods to be between 0.01 and 0.42 µg/g, fruit-based baby foods and deserts 0.01–0.49 µg/g, meat-based baby foods 0.01–8.38 µg/g, and cereals 0.01–0.31 µg/g. As for food samples commonly consumed during early childhood, the highest value of 10.730 µg/g was recorded for canned fish in tomato sauce, which aligns with the value recorded for a similar sample in the UK fluoride database (71). The fluoride concentration of fluoridated tap water from Newcastle upon Tyne (0.98 µg/g) matches fluoride levels set by the water fluoridation scheme that aims to achieve 1 mg F/L of water (72). Although in line with previous findings, minor differences in fluoride concentration between samples analyzed in this study and those presented in the literature can be expected. Brand, production site, or the fluoride concentration of the water used for production are just some factors that can impact the final fluoride content (73–76), which highlights the complexity of fluoride measurements. Finally, food samples available in the UK and US markets were analyzed in this study indicating small but detectable differences in fluoride concentration between the different countries (Table 1). Although such variations among different geographical areas are common (73, 77), they should not be ignored. The global movement of goods, including dietary sources, complicates monitoring systemic fluoride exposure and efforts to link total fluoride exposure to health outcomes. This highlights the need for a universal policy on fluoride labelling of food and drink products, especially those consumed in infancy and early childhood (74).

Semi-skimmed and whole milk samples were analyzed for fluoride concentration using an indirect method of fluoride determination and a combination of microwave-assisted acid digestion with a direct method of fluoride determination (54). A noticeable difference in fluoride concentration was observed between the two methods. For both whole and semi-skimmed milk, the indirect method of fluoride measurement yielded 0.001 µg/g, whereas the microwave-assisted acid digestion followed by direct method of fluoride measurement resulted in concentrations of 0.029 and 0.047 µg/g for both whole and semi-skimmed milk, respectively. An earlier study found the average fluoride concentration of different samples of homogenized milk to be 0.011–0.035 µg/g, with the highest recorded value being 0.063 µg/g (excluding plant-based milk) (78). Similarly, the fluoride concentration of milk available in Brazil ranged from 0.02–0.07 µg/ml for whole milk and 0.02–0.05 µg/ml for skimmed milk samples (79). Agreement between results obtained using a combination of microwave-assisted acid digestion and those found in the literature prompted usage of specifically these values for further bio-accessibility calculations.

The overall mean (SD) laboratory-measured fluoride concentration of meals was 0.140 (0.213) µg/g, ranging from 0.004 to 1.219 µg/g. Such results are comparable to those obtained by Buzalaf et al. (80) and Pagliari Tiano et al. (81), where the fluoride concentration of meals ranged from 0.007 to 0.580 mg and 0.011 to 0.743 µg/g, respectively. Similarly, the expected fluoride concentration of meals averaged 0.152 (0.219) µg/g, ranging between 0.003 and 0.861 µg/g. While for approximately 50% of samples the expected fluoride concentration of meals was higher than the laboratory-obtained values, suggesting the formation of insoluble complexes between fluoride and other cations potentially lowering fluoride detectability (82), the mean (SD) difference between the two groups was low: 0.012 (0.101) µg/g. A strong positive relationship between laboratory-obtained and expected values was detected through correlation and regression analysis, while a Bland-Altman analysis confirmed the close agreement between the two methods (58) suggesting that they are both valid, produce consistent results, and can be used interchangeably in the future.

Finally, upon examining the groups, the highest laboratory-measured mean (SD) fluoride concentration of 0.442 (0.360) µg/g was observed for meals created with water. This closely matched the expected fluoride concentration of 0.439 (0.379) µg/g. The fluoride concentration of the water used for food preparation is known to affect the overall fluoride concentration of the prepared product. For instance, the UK fluoride database showed varying fluoride levels in food and drink samples prepared with waters of different fluoride concentrations (75). Similarly, Heilman et al. (70) noted variability in the fluoride concentration of analyzed infant foods, identifying the fluoride concentration of the water used for processing as the primary factor, a finding corroborated by Wiatrowski et al. (83).

Earlier studies on fluoride bioavailability showed complete fluoride bioavailability when NaF was consumed in a fasting state (31) but reduced bioavailability when it was consumed alongside other food items (32, 33, 84). Although the present study also showed 100% fluoride bio-accessibility for pure NaF of the highest concentration (10 µg/ml), a gradual reduction in fluoride bio-accessibility was recorded from 10 to 0.1 µg/ml NaF standards analyzed in this study. Such results are most likely a consequence of a lower proportion of fluoride ions available at lower concentrations when the effect of other, possibly present, cations could have been more significant. The gastric phase of digestion resulted in slightly higher fluoride bio-accessibility when compared to those values obtained after the whole gastrointestinal process. Similar results were recorded in an earlier study when digestion of seafood samples resulted in a reduction of over 50% between gastric and intestinal phases (57), highlighting the significance of complete gastrointestinal digestion when examining the potential effects of fluoride intake (57).

The overall mean (SD) fluoride bio-accessibility for individual food samples was 44.7% (37.5%) with the highest value of 152.3% recorded for jam. Such a value might be considered an outlier as a consequence of jam's sticky texture and high viscosity that might have limited mixing with gastric fluids and enzymes. A study from 2023 conducted using peanut butter indicated that higher mixing forces are needed for the successful digestion of viscous products, therefore favoring dynamic models (85). On the other hand, a result of 150% might not be impossible. Digestion processes might have released more fluoride ions than acid diffusion used for the initial measurement of fluoride concentration (52). The challenging texture of jam might have restricted the contact between the acid and the food item preventing the effective release of fluoride making it undetectable. As for the rest, a majority of samples commonly consumed during infancy and early childhood resulted in fluoride bio-accessibilities below 100%, which emphasizes the limited availability of consumed fluoride for utilization in the human body.

The results of this study indicate significant variability in fluoride bio-accessibility depending on the type of beverage consumed alongside meals. Meals consumed with juice and milk exhibited higher fluoride bio-accessibility, averaging 79% and 102%, respectively, while those consumed with carbonated drinks and tap water showed lower bio-accessibility, with averages of 64.3% and 40.2%, respectively. These differences can be partly attributed to the varying fluoride concentrations in the beverages and their distinct chemical properties, which may influence fluoride release and absorption during digestion. As expected, meals prepared with fluoridated tap water generally showed higher fluoride bio-accessibility compared to those made with non-fluoridated water, highlighting the impact of water fluoridation on overall fluoride intake. Meals consumed with milk demonstrated exceptionally high fluoride bio-accessibility, with some outliers exceeding 150%. After removing these outliers, the mean fluoride bio-accessibility for milk-consumed meals dropped to 71.5%, still higher than most other beverage categories. This suggests that milk, particularly whole milk, may enhance fluoride bio-accessibility, possibly due to its fat content or other constituents that influence fluoride absorption. It is interesting to see that peanut butter and jam were common constituents of removed outliers. As explained above, such anomalies could have been a result of a sticky texture and high viscosity characteristic for those products which interfered with digestion and release of fluoride ions (85).

As with individual food samples, the majority of analyzed meal samples resulted in fluoride bio-accessibilities below 100%. Such an outcome can be attributed to the chemical properties of fluoride. Fluoride ions, owing to their electronegativity, tend to form complexes with various cations (57). This must be considered as samples analyzed during this study were mainly complex food matrices containing a variety of ingredients, while the simulated digestive fluids used throughout this project contained a wide range of cations that could have potentially reacted with fluoride rendering it undetectable. In acidic conditions, the creation of hydrogen fluoride (HF) is favored whereas, in alkaline environments, fluoride exists in a dissociated form. Thus, when in its dissociated form, fluoride readily binds with these cations which could result in reduced bio-accessibility, especially during the intestinal phase. This was already proven to be the case above while inspecting the influence of the digestion phase on overall fluoride bio-accessibility from NaF standards. These results showed that the intestinal phase of digestion decreased fluoride bio-accessibility by more than 40% while a reduction of more than 50% between gastric and intestinal phases was observed in an earlier study (57). These results highlight the importance of the entire gastrointestinal process when evaluating the potential risks or benefits of fluoride ingestion.

Furthermore, it should be noted that the static in vitro digestion model used in this study, while commonly applied to assess the bio-accessibility of various nutrients (86, 87), has limitations. Unlike dynamic models, which simulate the gradual movement of food through the digestive system, the static model exposes food samples to different digestive phases without changing conditions like pH or enzyme concentration. For example, the gastric pH in this study was adjusted to 1.5–2.5, but in reality, food intake temporarily raises stomach pH (88), potentially reducing fluoride solubilization. One study found that at pH 6, less than 13% of fluoride was soluble when certain seafood types were digested (57).

Overall, these findings highlight the need for careful interpretation of bio-accessibility data, especially given the constraints of the static in vitro digestion model and suggest that further research is needed to refine measurement techniques and better understand the factors influencing fluoride absorption in different dietary contexts.