Understanding the variation in salivary fluoride levels: A secondary trajectory‐based analysis of trial data

Abstract

This report presents the findings of a secondary analysis of the data from an RCT designed to compare salivary fluoride levels during 3 weeks of using 5000 or 1450‐ppm fluoride toothpaste (n = 24/group). Following this trial phase, salivary fluoride monitoring continued for 2 weeks during which all 48 participants used 1450‐ppm fluoride toothpaste in a wash‐out phase. We investigated the temporal variability and longitudinal growth‐based trajectories of salivary fluoride levels across study arms. During the trial phase, variability in the 5000‐ppm arm exceeded that of the 1450‐ppm arm. The best‐fitting trajectories identified three clusters of participants following: low, medium or high salivary fluoride trajectories. Except for the ‘high’ cluster during the wash‐out period, which featured only participants from the 5000‐ppm trial arm, all clusters had participants from both trial arms. In conclusion, exploration of the variation and trajectories of salivary fluoride after 1450‐ and 5000‐ppm fluoride toothpaste‐use suggests non‐constant responses to the initiation of and wash‐out from the use of high‐dosage fluoride toothpaste. This calls for the use of repeated period cross‐over trial designs to investigate treatment effect heterogeneity. Reasons for and ways to minimize sources of individual variation in salivary fluoride levels are discussed.

INTRODUCTION

It is well known that the principal mechanism of action of fluoride in caries control involves an increased fluoride ion activity in the oral fluids suffusing the dental hard tissues [1, 2, 3]. Much research effort has therefore been devoted to determining how to optimize the caries controlling effects of fluoride. Gaining a better understanding of how to maintain an elevated background salivary fluoride level over prolonged periods of time is one avenue to follow.

In a recent randomized clinical trial (RCT) of the effect of daily brushing over a period of 3 weeks with either 5000 (intervention) or 1450 (control)‐ppm fluoride toothpaste, we observed large inter‐individual variation in the background salivary fluoride levels [4]. This finding is in line with earlier research which has found comparable variation in background fluoride levels in saliva (Table 1) [5, 6, 7, 8, 9]. While many studies have thus noted this individual variation, few have tried to quantify or explain it. Bruun et al. [10] studied the effect on the salivary fluoride level up to 2 h after brushing with toothpastes containing fluoride in the range from 500 to 1500 ppm. Intriguingly, they found that individual differences, rather than the toothpaste fluoride contents, were the main source of variation in the salivary fluoride levels at all time points, except very initially (30 s).

TABLE 1.

Design details of studies investigating background fluoride (F) levels in saliva after fluoride exposure.

Study	Intervention	Time span (latest F‐exposure—saliva sampling)
Duckworth and Morgan [5]	Brushing with 1000‐, 1500‐ and 2500‐ppm F toothpaste	≥18 h
Richards et al. [6]	Brushing with 1500‐ppm F toothpaste	Overnight fasting
Ekstrand et al. [7]	Brushing with 1450‐ and 5000‐ppm F toothpaste	Alternating a
Pessan et al. [8]	Rinsing with 1000‐ and 5000‐ppm F toothpaste slurry	≈12 h
Staun Larsen et al. [9]	Brushing with 1450‐ and 5000‐ppm F toothpaste	6 h

^a Time points varied from early morning to night, and time between exposure and sampling was not standardized.

Research has pointed to several factors that might create individual variation in the fluoride retention in the oral fluids following tooth brushing. According to the model proposed by Duckworth and Morgan [5], the fate of fluoride in the oral cavity after tooth brushing can be described as a function of the fluoride input through the toothpaste used, the amount of fluoride immediately taken up by oral tissue reservoirs, the amount spat or rinsed out after brushing and the amount gradually swallowed. As the salivary fluoride concentration decreases over time, fluoride is gradually released from the oral tissue reservoirs into saliva thereby ‘buffering’ the salivary clearance of fluoride. With this model in mind, it is clear that factors such as the amount of dental biofilm [11, 12, 13, 14] and calculus [15, 16], the amount of calcium fluoride/calcium fluoride‐like deposits present on sound dental surfaces [17], the presence and number of caries lesions [17, 18, 19] and the area of the oral mucosa [9, 20] may impact on the amount of fluoride entering the oral reservoirs. Moreover, both the unstimulated [21] and the stimulated salivary flow rate [22, 23, 24] will influence the rate of oral clearance. Certain behavioural factors may also affect the background salivary fluoride concentration. These include the individuals’ general activities, for example the number of meals and the occurrence of in‐between snacking. The types of beverages and foods consumed may affect both the salivary flow rate and the dissolution rate of the calcium fluoride reservoirs. In a trial context, one also needs to consider the individuals’ adherence to the instructions. These might affect the salivary fluoride concentration owing to variation in the amount of toothpaste used per brushing [25], the time used for brushing [25] or the post‐brushing rinsing vigour [26, 27], just to mention a few possibly influential factors.

For toothpaste fluoride concentrations of 5000 ppm, recommended for caries active individuals [28, 29], no one has, to the best of our knowledge, investigated the individual variation in the background salivary fluoride levels in a longer follow‐up scenario, that is many hours after brushing. Variation in salivary fluoride levels after exposure to high‐dosage fluoride toothpaste could potentially differ from the variation observed after regular toothpaste because the salivary fluoride concentration is sustained above 100 ppm for a long(er) period with the high‐dosage paste [30]. This permits higher levels of calcium fluoride to precipitate [31, 32], which might affect the inter‐individual variation.

The aim of the present study was to investigate the temporal variation and longitudinal trajectories of the background salivary fluoride levels during and after use of 5000‐ and 1450‐ppm fluoride toothpaste. This is achieved through a secondary analysis of the data from our recent RCT [4] and the view is to discuss possible reasons for and ways to overcome/minimize this variation in future studies.

MATERIAL AND METHODS

The present report utilizes data from a previously reported randomized controlled trial of the background salivary fluoride levels following brushing with different fluoride toothpastes [4] (Figure 1) to carry out a secondary analysis of the temporal variation and longitudinal trajectories of the background salivary fluoride levels. The main scope of this descriptive secondary analysis is to be hypothesis generating, rather than hypothesis testing. Hence, secondary hypothesis testing on trial data that have been collected to address a single hypothesis will increase the risk of making type I errors, thereby defying the design of the original trial. In the present analysis, the statistical tests made are therefore used descriptively as indicators of factors that might be of concern if the aim is to reduce the variation in future trials of salivary fluoride levels.

FIGURE 1.

An overview of the design of the clinical trial which forms the basis for this secondary analysis. Source: This figure was first published in Staun Larsen et al. [4]

Trial design

In the parent trial (Figure 1), 48 participants were randomly allocated to brush twice daily for 3 weeks with either 1450‐ppm (Fresh Gel; Colgate Palmolive) or 5000‐ppm (Duraphat 5000 ppm Fluoride Toothpaste; Colgate‐Palmolive) fluoride toothpaste (Figure 1). During the 3‐week trial phase, the background salivary fluoride level, which was defined as the salivary fluoride level at least 4 h after brushing (mean time elapse was 7.2 h, standard deviation [SD] = 1.4), was assessed at five time‐points (Figure 1). After the 3‐week period, all participants went through a 2‐week wash‐out phase during which all participants brushed twice‐daily with 1450‐ppm fluoride toothpaste and had their background salivary fluoride levels assessed five times [4] (Figure 1). Participants were allocated 1:1 to the experimental arms (n = 24/arm) using randomization with minimization based on a pre‐study assessment of their salivary fluoride level. The sample size for the parent trial was calculated assuming a difference between groups in the salivary fluoride concentration at day 21 (end of trial phase) of 0.15 µg fluoride/mL, at α = 0.05 and 1‐β = 0.90 [4].

Participants were identified among students from Aarhus University who volunteered based on posters put up in campus canteens. Participants were required to be aged ≥ 18 years, to retain ≥ 20 teeth, to be non‐pregnant and non‐nursing, and to present an unstimulated salivary flow rate of ≥1 mL/5 min. This was determined during a pre‐study examination during which the candidate was asked to drool into a tube for 5 min. During the pre‐study examination, information was also elicited on the time since last brushing, the type of toothbrush used (manual or electric) and the toothpaste brand used. Of the 50 potential participants evaluated for eligibility, two chose to drop out before the beginning of the study owing to Covid‐19–related restrictions. Thereby, 48 participants were randomly allocated to one of the two trial arms, and they all completed the study as intended.

The experimental toothpastes were prepared and labelled by authorized pharmaceutical staff at the Hospital Pharmacy Central Denmark Region. Both experimental toothpastes were transferred to identical tubes similar to the tubes containing the high‐dosage fluoride toothpaste. All toothpaste tubes thus contained 51 g of toothpaste. The pharmacy kept a sealed list explaining which toothpaste tubes were of the same contents, as well as a sealed randomization list allocating toothpastes to participants.

Each participant was handed their allocated toothpaste for the trial phase at the beginning of the experimental period, and they received the toothpaste for the wash‐out period 3 weeks later. Participants were required to return the used experimental toothpaste tubes, and these were weighed with a view to calculate the individual toothpaste use. All participants were instructed to brush for 2 min mornings and evenings with 1 g of the allocated toothpaste and asked to refrain from using any other fluoride‐containing dental products during the study. Brushing took place, unsupervised, in the participants’ homes.

Participants were scheduled to come to the study centre in the afternoon of the designated saliva sampling days (Figure 1) and had been asked to refrain from eating and drinking for 1 h before sampling. Sitting slightly stooped in a quiet room, participants were asked to drool for 5 min into a tube (Falcon; Sarstedt). The time (in hours) since last toothbrushing was recorded at all sampling sessions. Samples were stored in a freezer (−18°C) until analysed for fluoride content. Covid‐19–related restrictions meant that 59 saliva samples (12%) had to be obtained at the participants’ homes. The concerned participants were instructed to use the same procedures as in the study centre, to put the sample in their −18°C freezer and arrangements were made to have the frozen samples delivered to the study centre. At each visit, participants were interviewed about their adherence to instructions, and any protocol irregularities or disturbances were noted, according to the mandatory regulations on clinical trials on medicinal products. A single participant from the 5000‐ppm fluoride toothpaste group deviated from the protocol for 4 consecutive days, during which a 1450‐ppm toothpaste was used. All other participants reported to have complied with protocol. The interview also elicited information on time since last brushing. Table 2 details the baseline characteristics of the participants.

TABLE 2.

Key baseline characteristics of participants in the two experimental groups.

	Experimental regimen
Characteristic	1450‐ppm F toothpaste n = 24	5000‐ppm F toothpaste n = 24
Age in years (mean, SD)	23.5 (2.2)	22.4 (1.5)
Sex (% female)	67%	79%
Pre‐study salivary flow rate mL/5 min (mean, SD)	2.1 (1.7)	2.3 (1.6)
Toothbrush (% using manual toothbrush)	54%	54%
Fluoride concentration in own toothpaste (ppm)	1450 (n = 24)	1250 (n = 1)
		1450 (n = 22)
		1490 (n = 1)
Time (h) since brushing at pre‐study examination (mean, SD)	5.8 (1.4)	6.2 (1.5)
Time (h) since brushing at baseline examination (mean, SD)	7.7 (2.3)	7.7 (2.5)

Abbreviation: F, fluoride.

Source: This table was first published in Staun Larsen et al. [4].

All salivary fluoride analyses were performed with a combination fluoride electrode (Orion 9609BNWP, ionplus Sure‐Flow Fluoride; Thermo Fisher) connected to a potentiometer (PHM210, STANDARD pH METER; MeterLab). Five hundred microlitres of sample were buffered with 50 µL TISAB III (Orion 940,911; Thermo Fisher) according to the manufacturer's instruction. Calibration included 10 standard fluoride solutions covering the range from 0.01‐ to 10‐ppm fluoride. The frozen saliva samples were analysed for fluoride content in batches which comprised samples collected on different days of the experiment to eliminate a distinct pattern considering the different sampling days [4]. Salivary flow rate (mL/min) was calculated after each saliva sampling, and the amount of toothpaste (g) used per brushing was calculated by dividing the consumed amount by the prescribed number of brushings in each study phase. Additional parent study circumstances are listed in Table 3, and further details on study design and experimental procedures can be found in Staun Larsen et al. [4].

TABLE 3.

Key elements in the design of the parent trial: controlled and uncontrolled factors of possible importance for the background salivary fluoride levels recorded.

Factors controlled or measured	Explanation
Unstimulated salivary flow rate; recorded at pre‐study visit	Participants were instructed to drool into a tube for 5 min.
Sampling time point	Sampling was conducted in the afternoon; most convenient time point for the participants owing to study obligations.
Saliva sampling	In a quiet room, participants were instructed to drool into a tube for 5 min, sitting slightly stooped, feet apart. Some individuals found it difficult to drool passively into the collection tube and were reinstructed repeatedly.
Toothpaste amount	Participants were shown pictures of 1 g of toothpaste, on both a manual and an electric brush, and instructed to brush with that amount.
Other fluoride‐containing dental products	Participants were instructed to abstain from using any other fluoride‐containing dental products than the experimental toothpaste.
Brushing time	Participants were instructed to brush for 2 min per brushing; morning and evening.
Toothpaste type/brand	Both experimental toothpastes contained NaF as the fluoride source, silica as abrasive, the detergent was sodium lauryl sulphate, and the toothpastes appeared with an almost identical taste, even though flavour additives were not completely overlapping. The pH values observed for the two toothpastes were both 7.5.

Factors not controlled or measured	Explanation
Diet	No measures were taken to control participants’ diet or dietary practices during the 5‐week period of the study. However, participants were instructed to abstain from eating and drinking for 1 h before saliva sampling.
Brushing performance	Participants conducted unsupervised brushing in their homes.
Daily routines	For both ethical and practical reasons no measures were taken to control participants’ daily routines during the trial period.
Caries activity	Caries activity was not recorded. However, the university students participating in the study are presumed to demonstrate low caries activity.
Plaque and calculus	No professional oral hygiene was performed prior to the study.

The trial protocol was independently reviewed and approved by the Central Denmark Region Committees on Health Research Ethics (no. 1‐10‐72‐37‐20) and the Danish Medicines Agency. The trial was registered with the European Union Drug Regulating Authorities Clinical Trials Database (https://www.clinicaltrialsregister.eu, Eudra‐CT no. 2020‐000213‐33). The study was also registered with the General Data Protection Regulation list at Aarhus University, Denmark and always complied with the European Union General Data Protection Regulation legislation. All participants gave informed written consent to participate in the trial [4].

Data analysis

Data were analysed according to the intention‐to‐treat principle. To investigate differences in the variability between the study arms, equality of variability at each time point was tested with a non‐parametric Fligner‐Killeen's test, where null hypothesis is that there is no difference in variability across trial arms. To identify patterns of longitudinal trajectories of background salivary fluoride levels over time, we employed Growth‐Based Trajectory Modelling (GBTM) using the flexmix package [33] in R, version 4.4.1 [34]. GBTM is a statistical technique that allows for the identification of meaningful subgroups or clusters of individuals who follow similar developmental trajectories over time. The further view is to understand what factors may account for the distinctiveness of the cluster trajectories identified [35]. To determine the optimal number of trajectory groups and the best underlying trajectory model, we fitted models with 1–4 trajectory groups separately for the trial phase and wash‐out phase. Given the right‐skewed nature of fluoride values and multiple analytic options available for analysing such data, we tested GBTM with two modelling approaches: a linear (Gaussian) model with log‐transformed values and a Gamma model with untransformed values. Each model included linear and quadratic terms for time (days), but we also included time‐since‐tooth‐brushing (h) and salivary flow rate (mL/min) as possible time‐varying predictors for cluster membership.

The best‐fitting GBMT model was selected based on the Bayesian Information Criterion and Weighted Mean Absolute Error. Individual longitudinal trajectory group assignment was determined using posterior probabilities, with each participant assigned to the group for which they had the highest probability of membership. Then, cluster membership by experimental group status (trial arm) was tabulated. We also investigated whether there were differences in time since tooth brushing (h), salivary flow rate (mL/min) and toothpaste consumption (g/brushing) across the identified clusters.

RESULTS

Investigating differences in the variability between the study arms showed that the variability in the background salivary fluoride levels in the 5000‐ppm groups exceeded that of the 1450‐ppm group, particularly towards the end of the trial phase (day 10, 14 and 21), characterized by clearly wider interquartile ranges and statistically significant Fligner‐Killeen's test of equality of variability (Figure 2). At baseline and during the 2‐week washout phase, the differences in variability of background salivary fluoride levels between the trial arms were modest and statistically not significant (Figure 2).

FIGURE 2.

The background salivary fluoride (F) concentrations over time during (A) the trial phase (Days 0–21) and (B) the wash‐out phase (Days 21–35) for the two experimental groups, intervention (n = 24) and control (n = 24). Data presented are the median values with the 25th and 75th percentiles marked as error bars. The control group used a 1450‐ppm fluoride toothpaste during both phases, while the intervention group used a 5000‐ppm fluoride toothpaste during the trial phase and a 1450‐ppm fluoride toothpaste throughout the wash‐out phase. Angled values immediately above the horizontal x‐axes represent p‐values from Fligner‐Killeen's tests, where the null hypothesis is that there is no difference in variability across trial arms. A lower p‐value indicates higher variability in one of the trial arms

In both study phases, the best‐fitting trajectory models included three clusters, were fitted using gamma regression model, and did not include time‐since‐tooth‐brushing and salivary flowrate as time‐varying predictors (Figure 3). The average posterior probabilities of group membership were high for all clusters (over 90%).

FIGURE 3.

Individual (panels A, C) and group‐based (panels B, D) trajectories in salivary fluoride (F) levels in the trial and the wash‐out phase of the study. Group‐based trajectories were fitted using gamma regression modelling separately for each phase. The six clusters are illustrated by different colours; the membership of original group is illustrated by line type (panels A, C). Controls (solid lines, n = 24) used a 1450‐ppm fluoride toothpaste during both phases, while the intervention group (hatched lines, n = 24) used a 5000‐ppm fluoride toothpaste during the trial phase and a 1450‐ppm fluoride toothpaste throughout the wash‐out phase.

The trial phase clusters included a cluster of participants (High_Trial) whose background salivary fluoride levels increased during the trial phase and were considerably higher than in the other two clusters (Middle_Trial, Low_trial; Figure 3). In the Middle_Trial cluster, the average background salivary fluoride level also increased slightly during the trial phase, whereas in the Low_Trial cluster, the levels remained stable and lower than in the two other clusters.

The wash‐out phase clusters included a High_Wash‐out cluster in which members had consistently higher but decreasing background salivary fluoride levels during the wash‐out phase (Figure 3). A slight decline in background salivary fluoride levels was also seen in the Middle_Wash‐out cluster, whereas the Low_Wash‐out cluster showed consistently lower salivary fluoride levels during the wash‐out phase.

Figure 4 shows the allocation of individuals from the two original experimental groups between clusters across the two experimental phases: trial and wash‐out. Except for the cluster High_Wash‐out featuring only participants from the experimental 5000‐ppm group, all clusters had participants from both experimental groups.

FIGURE 4.

Cluster membership, as determined by Growth‐Based Trajectory Modelling for each of the two experimental phases: trial and wash‐out. Three clusters (dubbed High, Middle and Low) were identified at both experimental phases. Hatched areas in each cluster represent individuals from the Intervention group while plain areas represent participants in the Control group. The numbers to the left of the Trial phase column, and to the right of the Wash‐out phase column indicate the count of individuals from each trial arm in each cluster. Arrows between phases indicate how the individuals changed cluster membership between the Trial phase and the Wash‐out phase. Except for the cluster High_Wash‐out featuring only participants from the 5000‐ppm fluoride toothpaste group (intervention), all clusters had participants from both experimental groups.

There were no major differences in hours‐since‐tooth‐brushing and salivary flow rate across clusters (Figures 5 and 6; Table 4). In the trial phase, toothpaste consumption was also similar in all clusters (Table 4). In the washout phase, however, the individuals in the cluster with higher background salivary fluoride levels (High_Wash‐out) had consumed more toothpaste than the individuals in the two other clusters with lower salivary fluoride levels (Table 4).

FIGURE 5.

Individual (panels A, C) and group (panels B, D) trajectories in the time‐since‐tooth‐brushing across clusters for longitudinal trajectories of the salivary fluoride levels in trial and wash‐out phases shown in Figures 3 and 4.

FIGURE 6.

Individual (panels A, C) and group (panels B, D) trajectories in salivary flow rate across clusters for longitudinal trajectories of the salivary fluoride levels in trial and wash‐out phases shown in Figures 3 and 4.

TABLE 4.

Average time since tooth brushing, salivary flow rate and amount of toothpaste consumed during the trial and wash‐out phase by cluster membership status as determined by longitudinal growth‐based trajectory modelling.

	Time since brushing (h)	Salivary flow rate (mL/min)	Toothpaste consumption a (g/brushing)
Trial phase
Cluster	Mean (SD)	Mean (SD)	Mean (SD)
Low (23%)	7.15 (0.71)	0.55 (0.33)	0.60 (0.16)
Middle (54%)	7.22 (0.80)	0.62 (0.37)	0.64 (0.16)
High (23%)	7.31 (0.66)	0.63 (0.20)	0.68 (0.14)
p‐value b	0.88	0.80	0.44
Wash‐out phase
Cluster
Low (56%)	7.30 (0.84)	0.68 (0.37)	0.63 (0.20)
Middle (38%)	7.06 (0.38)	0.69 (0.39)	0.50 (0.16)
High (6%)	6.90 (0.64)	0.74 (0.19)	0.83 (0.23)
p‐value b	0.41	0.96	0.009

^a Calculated by dividing the total amount of toothpaste spent during period by the prescribed number of brushings (2/day).

^b p‐value for one‐way ANOVA test of cluster differences.

DISCUSSION

The findings presented here have shown that the inter‐individual variability in the background salivary fluoride levels increased during the 3‐week period of twice‐daily use of 5000‐ppm fluoride toothpaste and exceeded the variability in the group using 1450‐ppm fluoride toothpaste during the same period. To the best of our knowledge, no prior study has studied the inter‐individual variation in salivary fluoride sampled several hours after brushing. However, it is worth noting that our findings corroborate those of Bruun et al. [10] in the sense that they described an effect of the toothpaste fluoride concentration on the variation in salivary fluoride levels that persisted for at least 30 min.

The longitudinal trajectories of salivary fluoride levels were investigated independently of experimental group assignment. When data from each of the two experimental phases were allowed to cluster into up to four groups, the participants ended up clustering into three groups for the best‐fitting trajectory model for each of the two experimental phases. Except for the cluster High_Wash‐out, which featured only participants from the 5000‐ppm group, all other clusters had participants from both experimental arms (Figure 4). It is thus remarkable that several participants who were instructed to brush with 5000‐ppm fluoride toothpaste presented background salivary fluoride levels comparable to or even lower than seen for individuals brushing with 1450‐ppm fluoride toothpaste, despite the fluoride exposure differing by a factor of almost 3.5.

Although many factors were standardized in the design of this RCT, it is likely that some changes to and additional standardization of the study protocol would have limited the inter‐individual variation. Moreover, more comprehensive measurement of uncontrolled factors could have been gathered to explain the variability across individuals. Table 3 shows some key controlled and uncontrolled or unmeasured factors in the current study. Just to mention a few, tooth brushing could have been supervised at the day of sampling to ensure use of the correct amount of toothpaste and the correct brushing time, and sampling could have taken place in the morning before the first tooth brushing and before any intake of food and drinks to account for the circadian rhythm in salivary flow [36] and the effect of intake of food and beverages. Caries activity could have been recorded, as active lesions are known to constitute an important calcium fluoride reservoir within the oral cavity [17]. Minimization procedures could have been based on a combination of more factors, such as salivary fluoride level, salivary flow rate and number of active caries lesions. Certainly, the findings of our study indicate that minimization based solely on a single estimate of the salivary fluoride level is insufficient to reduce the inter‐individual variation in salivary fluoride levels during an experiment such as the present.

During the study, we observed that some participants found it difficult to drool passively into the collection tube and they were therefore reinstructed. This might be another source of variation since the mere act of spitting may influence salivary flow. A change in salivary flow per se would probably not change the inherent fluoride concentration in the secretion itself [37], but could of course cause a change in the fluoride concentration in whole saliva by affecting the salivary volume. It might therefore be a good thing to introduce more exercises in passive drooling prior to commencing experimental periods, to make sure that all participants are able to drool fully passively.

Better standardization of the amount of toothpaste used, time spent on each brushing and the post‐brushing rinsing habits might also contribute to a reduction of the inter‐individual variation in the background salivary fluoride levels during the experimental periods. Hence, research by Zero et al. [25] has indicated that both the amount of toothpaste and the time spent on brushing may be important for oral fluoride retention, and Duckworth et al. [26] demonstrated that mouth rinsing habits after tooth brushing were important for the oral retention of fluoride. Here, we noted that the participants did, on average, not fully comply with the instruction to use 1 g of toothpaste per brushing but used on average only 64% of this prescribed amount during the trial phase, and 59% during the wash‐out phase. More importantly, considerable inter‐individual variation in toothpaste consumption was noted, as these data indicate a 95% prediction interval (the interval covering 95% of the observations) in the range from 0.33 to 0.95 g/brushing in the trial phase and from 0.18 to 1.00 g/brushing in the wash‐out phase. These findings suggest that the overall underuse of toothpaste might well be due to a combination of a lower than prescribed amount of toothpaste per brushing and a lower frequency of brushing, that is non‐compliance with the twice‐daily brushing instruction. If the latter is a part of the explanation, it would not suffice to introduce toothpaste dispensers to secure use of the prescribed amount of toothpaste. It is worth noting that the three participants, who formed the High_Wash‐out cluster (Figure 3), belonged to the 5000‐ppm fluoride toothpaste arm during the trial phase and had a significantly higher toothpaste consumption than most other participants. This might be an indication of the importance of tooth brushing protocol violations when explaining the inter‐individual variation in salivary fluoride levels.

Owing to the duration of the experimental period of this study, no restrictions were placed on the individual's daily routines and dietary habits, except for the immediate hour prior to the saliva sampling [38]. Imposing more possible restrictions were considered to be too burdensome for the participants and would probably just have increased the risk of protocol violations. It would thus seem that the considerable variability in salivary fluoride levels is a reality that serves to reduce the signal‐to‐noise ratio in clinical trials involving salivary fluoride level measurements. This might, in turn, call for larger sample sizes. In this respect, it is noteworthy that the assumptions underpinning our sample size determination for the parent trial did not hold. Hence, the assumed between‐group difference in the salivary fluoride level of 0.15 µg/mL, which we derived from the study of Pessan et al. [8], was substantially higher than our observation of a much smaller between‐group difference of 0.021 µg/mL. However, as discussed previously [4], many study design differences may also have contributed to this.

As it remains unknown whether those participants whose background salivary fluoride concentrations significantly increased by the 5000‐ppm treatment would respond similarly to this treatment regimen if the experiment were to be repeated, it would be interesting to explore further the possibility of such ‘personal profiles’. Tooth brushing habits are probably part of a personal profile, just as salivary flow rates might be. Exploration of possible individual or subgroup treatment effect heterogeneity could be done by conducting repeated cross‐over trials [39, 40], where participants are treated with 5000‐ and 1450‐ppm fluoride toothpaste during several periods, and their background salivary fluoride trajectories compared over periods.

A more technical contribution to the observed pattern of inter‐individual variation in the background salivary fluoride levels could be that the fluoride measurements, for the most part, concern fluoride levels very close to the limit of detection for the fluoride electrode, which is around 0.02 ppm fluoride. Hence, the background salivary fluoride levels in the 1450‐ppm fluoride toothpaste trial arm during the entire trial duration and in the 5000‐ppm fluoride toothpaste arm at the beginning and at the end of the experiment were on average close to this limit of detection. This working condition has been reported to produce approximate rather than absolute data [41] and might have impacted the variability in the background salivary fluoride levels.

The present exploration of inter‐individual variability in the salivary fluoride levels may be important from two perspectives: In an experimental scenario, one would be interested in little inter‐individual variability as this will allow the demonstration of treatment effects with less trial participants and could thus reduce costs of carrying out a trial. From a clinical perspective, it may lead to a broader understanding of why the caries‐controlling effect of high‐dosage fluoride toothpaste may not have shown the results one might predict considering the substantially higher fluoride concentration than seen in regular toothpaste [42].

In conclusion, exploration of the variation and trajectories of salivary fluoride levels after use of 1450‐ and 5000‐ppm fluoride toothpaste suggests non‐constant and varying responses to the exposure to and wash‐out from high‐fluoride toothpaste and calls for repeated period cross‐over trials to investigate this treatment effect heterogeneity in more detail. Considering the multi‐faceted factors influencing salivary fluoride steady‐state levels many hours after last toothpaste exposure, future long‐term controlled studies would probably only to some extent benefit from a strict methodological protocol to minimize variation in experiments of long duration.

AUTHOR CONTRIBUTIONS

Conceptualization: Line Staun Larsen, Vibeke Baelum and Eero Juhani Raittio. Methodology: Line Staun Larsen, Vibeke Baelum and Eero Juhani Raittio. Formal analysis: Eero Juhani Raittio. Writing—original draft preparation: Line Staun Larsen. Writing—review and editing: Line Staun Larsen, Vibeke Baelum and Eero Juhani Raittio. Visualization: Line Staun Larsen and Eero Juhani Raittio.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.

Staun Larsen L, Baelum V, Raittio EJ. Understanding the variation in salivary fluoride levels: A secondary trajectory‐based analysis of trial data. Eur J Oral Sci. 2025;133:e70039. 10.1111/eos.70039