Characteristics of the trial designs and settings

Eight studies were of parallel design, with the remaining two studies using the cluster‐randomised design (Honkala 2014; Lenkkeri 2012). Of the eight parallel studies, three were multicentre (Bader 2013; Sintes 1995; Sintes 2002), three were conducted at single centres (Oscarson 2006; Petersson 1991; Zhan 2012), and two were unclear in this regard (Milgrom 2009; Taipale 2013). The studies were conducted in the USA (Bader 2013; Zhan 2012), Finland (Lenkkeri 2012; Taipale 2013), Sweden (Oscarson 2006; Petersson 1991), Costa Rica (Sintes 1995; Sintes 2002), Estonia (Honkala 2014), and the Republic of the Marshall Islands (Milgrom 2009). Four studies were carried out in a dental clinical setting (Bader 2013; Oscarson 2006; Petersson 1991; Zhan 2012), two studies in a school setting (Sintes 1995; Sintes 2002), and two studies were carried out in a combination of schools (where the intervention was given) and dental clinics (where the clinical examinations took place) (Honkala 2014; Lenkkeri 2012). The remaining two studies were carried out in a community setting (Milgrom 2009) and a healthcare centre setting (Taipale 2013).

Eight studies performed sample size calculations (Bader 2013; Honkala 2014; Lenkkeri 2012; Milgrom 2009; Petersson 1991; Sintes 1995; Taipale 2013; Zhan 2012). However, in two of these studies, the calculation was based on an outcome which was not of interest in this review (Taipale 2013; Zhan 2012). One study only carried out a post‐investigation sample size analysis (Oscarson 2006), and the remaining study did not mention sample size calculations (Sintes 2002).

Two studies only stated that they had received non‐industry funding (Bader 2013; Oscarson 2006). Three studies stated that they received non‐industry funding but that industry supplied the interventions (Milgrom 2009; Taipale 2013; Zhan 2012). Four studies were clearly industry funded, in other words industry provided economical support (Honkala 2014; Petersson 1991; Sintes 1995; Sintes 2002). The remaining study only stated that industry provided the interventions (Lenkkeri 2012).

Characteristics of the participants

There were 7969 participants randomised to interventions (including only the intervention groups relevant to this review), of which 5903 were included in the studies' analyses. One study investigated the effects of xylitol in adults (Bader 2013), whilst the remaining studies only included children. Five of these studies included children ranging from 8 to 13 years of age (Honkala 2014; Lenkkeri 2012; Petersson 1991; Sintes 1995; Sintes 2002), with the remaining four studies including younger children ranging from 1 month to 3 years of age (Milgrom 2009; Oscarson 2006; Taipale 2013; Zhan 2012). Approximately two thirds of the participants were females in the study on adults (Bader 2013), whilst the other studies all had roughly equal proportions of males and females.

Characteristics of the interventions and comparisons

Four studies involved the use of xylitol products (defined as lozenges, sucking tablets and candies) which were to be sucked (Bader 2013; Honkala 2014; Lenkkeri 2012; Oscarson 2006). A further study also involved xylitol tablets, but they were administered using a slow‐release pacifier/dummy, or crushed on a spoon if the child would not accept a pacifier (Taipale 2013). Three studies investigated xylitol‐containing fluoride toothpastes (Petersson 1991; Sintes 1995; Sintes 2002). One study tested a xylitol syrup (Milgrom 2009), whilst the remaining study tested xylitol wipes (Zhan 2012).

The dosage of xylitol ranged from 200 to 600 mg per day to 8 g per day. The total daily dosage was unclear in the three toothpaste studies, as it was reported as a percentage of xylitol, at 3% (Petersson 1991), or 10% (Sintes 1995; Sintes 2002). Of the four studies including younger children (baseline mean age ranging from 1 month to 2 years), two used very low daily doses of 200 to 600 mg (Taipale 2013), and 1 g (Oscarson 2006), whilst two used higher daily doses of 4.2 g (Zhan 2012), and 8 g (Milgrom 2009). In two studies with older children (baseline mean age ranging from 8 to 10 years), the daily dose was 7.5 g (Honkala 2014), and 4.7 g (Lenkkeri 2012), and in the adult study, the dose was 5 g per day (Bader 2013). The duration of the intervention ranged from 1 to 3 years. In three studies, the final follow‐up occurred after the participants had ceased to receive the intervention: 1.5 years of intervention with follow‐up at 2 years (Oscarson 2006), and 2 years of intervention with follow‐up at 4 years (Lenkkeri 2012; Taipale 2013).

Xylitol products, with their sweet flavour, cause extra saliva to be produced, especially with lozenges/candies/sucking tablets that are sucked over a period of time. Therefore it is difficult to distinguish how much of any effect is due to the xylitol or the extra saliva that is produced. Thus it would be desirable for studies to have both a control arm using a placebo product and a no treatment control arm. Sorbitol is frequently used as a placebo in xylitol studies as it is neither considered to cause or prevent caries We did not consider any products which are thought to prevent caries (e.g. erythritol) or those which are known to cause caries (e.g. sucrose) as appropriate placebos. Two studies used no treatment in the control arm (Lenkkeri 2012; Oscarson 2006). Two studies used sorbitol which we treated as a placebo control arm (Honkala 2014; Taipale 2013). Two studies stated that they used a placebo, one of which used sucralose as the sweetening agent in the lozenges (Bader 2013), whilst the other used wipes containing no xylitol (Zhan 2012). The three toothpaste studies used entirely appropriate placebos, in that all participants either used a toothpaste with or without xylitol (Petersson 1991; Sintes 1995; Sintes 2002). The remaining study used a lower dose of xylitol for the control group, which was demanded by the internal review committee of the Secretary of Health, but the authors cited evidence that the lower dose (2.67 g per day) would not have an effect on caries incidence (Milgrom 2009).

Characteristics of the outcomes

All ten studies assessed the effects of xylitol products on caries, which was our primary outcome. Five studies reported continuous data in the form of caries increment: three reporting decayed filled surfaces (DFS) (Bader 2013; Sintes 1995; Sintes 2002), and two reporting decayed missing filled surfaces on deciduous teeth (dmfs) (Oscarson 2006), and permanent teeth (DMFS) (Lenkkeri 2012). The latter two studies also reported caries as a dichotomous outcome, in other words whether or not there was an increment (change in caries). Two studies only reported caries as a dichotomous outcome (dmfs increment: yes/no) (Taipale 2013; Zhan 2012). The remaining three studies reported the mean number of decayed primary teeth (Milgrom 2009), the mean number of DFS (Petersson 1991), and the mean number of dmfs (Honkala 2014). One of those studies did not report measures of variance (e.g. standard deviation), which would preclude its inclusion in any meta‐analysis (Petersson 1991). Another study combined deciduous and permanent teeth, which we considered inappropriate (Honkala 2014). If the authors had presented the results as the mean incremental change, the results would show a large reduction in caries.

Thresholds for diagnosis of caries varied between the studies, with different scoring systems used. Two studies defined caries to include non‐cavitated enamel lesions (Taipale 2013; Zhan 2012). Four studies defined caries as a visible breakdown in the enamel wall (Bader 2013; Oscarson 2006; Sintes 1995; Sintes 2002). One study reported enamel and dentine caries separately (Honkala 2014), and one study reported only dentine caries (Lenkkeri 2012). One study defined caries as cavitated lesions, but did not specify if this included enamel or dentine (Milgrom 2009), and one reported combined "initial and gross caries," but it is unclear how this relates to either enamel or dentine lesions (Petersson 1991).

Three studies did not mention adverse/side effects (Oscarson 2006; Petersson 1991; Taipale 2013). Of the seven studies that did mention adverse effects, two did not present the data in a usable format (Lenkkeri 2012; Milgrom 2009), one reported raw data on a publicly accessible website (Bader 2013), and the remaining four just stated that there were none observed or reported (Honkala 2014; Sintes 1995; Sintes 2002; Zhan 2012). Adverse effects include sores in the mouth, cramps, bloating, constipation, flatulence, and loose stool or diarrhoea.

No other secondary outcomes of this review were reported.

Blinding of participants and personnel (performance bias)

Two studies used no treatment for the control group and therefore it was not possible to blind the participants or personnel (Lenkkeri 2012; Oscarson 2006). Whilst this could not impact upon the outcome assessment (which was blinded in all cases and caries outcomes were not assessed by the participants), we could not rule out that there might have been an effect on behaviour of the participants or their carers or both, which could potentially affect the outcome. We assessed these two studies as having an unclear risk of bias. The remaining eight studies were all adequately blinded through the use of active controls which were unidentifiably different, and we assessed them as having a low risk of bias.

Blinding of outcome assessment (detection bias)

All 10 studies involved blinded outcome assessment and we rated them as having a low risk of bias.

Overall risk of bias

• One study was at low overall risk of bias (Bader 2013).

• Two studies were at unclear overall risk of bias (Milgrom 2009; Zhan 2012).

• Seven studies were at high overall risk of bias (Honkala 2014; Lenkkeri 2012; Oscarson 2006; Petersson 1991; Sintes 1995; Sintes 2002; Taipale 2013).

We present the results of the risk of bias assessments graphically in Figure 2.

Figure 2

---

---

Risk of bias summary: review authors' judgements about each risk of bias item for each included study.

Xylitol lozenges versus control lozenges

One study, at low risk of bias and analysing 669 participants, compared xylitol (5 g per day) lozenges with control lozenges over 33 months (Bader 2013). There was no difference in caries increment for DFS (mean difference (MD) ‐0.64, 95% confidence interval (CI) ‐1.58 to 0.30, P value = 0.18) (Analysis 1.1). This translates to a non‐significant prevented fraction (PF) of 8% in favour of the xylitol group (Table 1).

Patterns of adverse effects (sores in the mouth, cramps, bloating, constipation, flatulence, and loose stool or diarrhoea) were similar for both groups.

Xylitol candy versus control (sorbitol) candy

One study, at high risk of bias and analysing 252 children, compared xylitol (7.5 g per day) candy with control (sorbitol) candy over 36 months (Honkala 2014). We were unable to use the data in our analyses as the authors combined the primary/deciduous and permanent teeth and reported the mean number of decayed missing filled surfaces. We would require them to report the mean increment and standard deviation for either primary or permanent teeth in order to use the data in a meta‐analysis.

The authors reported that there were no adverse effects for either group.

Xylitol lozenges versus no treatment

One study, at high risk of bias and analysing 200 children, compared xylitol (4.7 g per day) lozenges with no treatment over 24 months, with 48 month follow‐up (Lenkkeri 2012). There was no difference in caries increment for DMFS(MD 0.28, 95% CI ‐0.99 to 1.55, P value = 0.67) (Analysis 2.1). This translates to a non‐significant PF of 10% in favour of the no treatment group (Table 1). There was also no difference in the number of children with a caries increment, in other words those with caries occurring between baseline and final follow‐up (risk ratio (RR) 1.02, 95% CI 0.83 to 1.26, P value = 0.85) (Analysis 2.2).

There were no usable data presented for adverse effects.

Xylitol syrup versus control (low‐dose xylitol) syrup

One study, with an unclear risk of bias and analysing 94 infants, compared xylitol (8 g per day) syrup with low‐dose xylitol (2.67 g per day) syrup over 12 months (Milgrom 2009). The higher dose of xylitol syrup resulted in a statistically significant reduction in the mean number of decayed primary teeth (MD ‐1.10, 95% CI ‐2.03 to ‐0.18, P value = 0.02) (Analysis 3.1). Using PF, this translates to a 58% reduction in caries (Table 1).

Adverse effects were not reported in a usable format but the reported rates of loose stools and diarrhoea were very similar. There were no serious adverse effects experienced during the study.

Xylitol sucking tablets versus no treatment

One study, at high risk of bias and analysing 118 infants, compared xylitol (1 g per day) sucking tablets with no treatment over 18 months, with a 24 month follow‐up (Oscarson 2006). There was no difference in caries increment for dmfs (MD ‐0.42, 95% CI ‐1.12 to 0.28, P value = 0.24) (Analysis 4.1), although when this was converted into PF it was marginally statistically significant, and equated to a 53% reduction in caries in favour of the xylitol group (Table 1). There was also no difference in the number of infants with a caries increment (RR 0.72, 95% CI 0.35 to 1.45, P value = 0.35) (Analysis 4.2).

No other outcomes were considered in this study.

Xylitol toothpaste versus control toothpaste

Three studies, all at high risk of bias, compared fluoride toothpastes containing xylitol with fluoride‐only toothpastes over 30 to 36 months (Petersson 1991; Sintes 1995; Sintes 2002). One of the studies, analysing 248 children, compared low‐fluoride plus 3% xylitol (daily dosage unclear) with low‐fluoride, and normal‐level fluoride plus 3% xylitol with normal‐level fluoride (Petersson 1991). The authors did not report data in a usable format, but found no difference in the number of DFS between any group. The study did not consider any other outcomes.

We were able to pool the data from the other two studies in a meta‐analysis, which revealed that fluoride toothpaste containing 10% xylitol (daily dosage unclear) resulted in a 13% reduction in caries increment for DFS (PF ‐0.13, 95% CI ‐0.18 to ‐0.08, P value < 0.00001, 4216 children analysed) (Analysis 5.1; Analysis 5.2).

Both studies reported that there were no adverse effects in either group.

Xylitol tablet versus control (sorbitol) tablet

One study, at high risk of bias and analysing 62 infants, compared xylitol (200 to 600 mg per day) tablets administered via a slow‐release pacifier/dummy or crushed up on a spoon with control (sorbitol) tablets administered in the same way over 24 months, with a 48 month follow‐up (Taipale 2013). There was no difference in the number of infants with a caries increment for dmfs (RR 3.08, 95% CI 0.69 to 13.65, P value = 0.14) (Analysis 6.1).

No other outcomes were considered in this study.

Xylitol wipes versus control wipes

One study, at unclear risk of bias and analysing 44 infants, compared xylitol (4.2 g per day) wipes with control wipes over 12 months (Zhan 2012). There was no difference in the number of infants with a caries increment for dmfs (RR 0.14, 95% CI 0.02 to 1.07, P value = 0.06) (Analysis 7.1).

The authors reported that there were no adverse effects for either group.