Milgrom 2009.

Methods	Trial design: parallel (3 arms) Location: communities in the Republic of the Marshall Islands (in the northern Pacific Ocean) Number of centres: unclear Recruitment period: April to August 2006
Participants	Inclusion criteria: children aged 9 to 15 months Exclusion criteria: in the lower 10th percentile of US standard weight and height; history of oesophageal or digestive disease; congenital craniofacial malformation; history of adenoidectomy, tympanostomy tubes, or tympanic membrane perforations (due to secondary outcome of reduction of acute otitis media) Baseline caries: not stated Age at baseline (months): Gp A: mean 15.9 (SD 2.6); Gp B: mean 13.7 (SD 2.4); Gp C: mean 15.6 (SD 2.7) Gender: Gp A: 57.6% female; Gp B: 56.3% female; Gp C: 48.3% female Any other details of important prognostic factors: high rate of early childhood caries on the islands; drinking water contains no appreciable fluoride; generally poor diets Number randomised: 100 (Gp A: 35; Gp B: 33; Gp C: 32) Number evaluated: 94 (Gp A: 33; Gp B: 32; Gp C: 29)
Interventions	Comparison: xylitol topical oral syrup (A) versus xylitol topical oral syrup (B) versus xylitol topical oral syrup (C) Gp A (n = 35): three doses of syrup per day: two doses contained 4 g xylitol each, plus one dose with 2 g sorbitol (total dose = 8 g xylitol plus 2 g sorbitol per day) Gp B (n = 33): three doses of syrup per day: each dose contained 2.67 g xylitol (total dose = 8 g xylitol per day) Gp C (n = 32): three doses of syrup per day: one dose contained 2.67 g xylitol, plus two doses contained 2 g sorbitol each (total dose = 2.67 g xylitol plus 4 g sorbitol per day) Duration of treatment: 1 year Sorbitol is considered to be inert (i.e. neither causes nor prevents caries) so we did not consider its inclusion in the syrups to be a problem The study authors state that the internal review committee required that the control group syrup contain a small amount of xylitol, but that evidence does not suggest that a dose of 2.67 g per day would have any effect. The study authors therefore considered this to be a control group Due to the fact that the xylitol dose was the same in Gp A and Gp B, and also because sorbitol is not expected to have any effect, we combined these two groups in the analysis
Outcomes	Caries: number of decayed primary teeth. Assessed at 1 year Acute otitis media (not an outcome of interest in this review) Adverse effects
Notes	Sample size calculation: "We estimated that the rate of decayed cavitated lesions for children at 24 months of age was 60% in the control group and 30% in the xylitol‐treated groups. Based on 80% power to detect a significant difference (2‐sided P=.05) between the xylitol‐treated and control groups, 32 children were required for each study group" Adverse effects: percentage experiencing loose stools or diarrhoea is reported per group. Appears to be at participant level (i.e. not counting multiple events experienced by the same child) but the percentages stated do not equate to whole persons (e.g. 11.7% of 33 participants is 3.861 persons) Funding source: grants from: a) the Health Resources and Services Administration Maternal and Child Health Bureau; and b) the National Institute of Dental and Craniofacial Research. Danisco USA donated the raw materials for making the syrups Declarations/conflicts of interest: none of the authors reported any disclosures
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "subjects were given identification numbers that had been randomly assigned to study groups by a statistician using block randomization and the sample function of commercially available statistical software...Block sizes of 30 and 15 were used for the Laura district, and block sizes of 36 and 18 were used for the Delap district. Except for the statistician, all study team members were blinded until study completion" Comment: this is an appropriate method of random sequence generation
Allocation concealment (selection bias)	Low risk	Quote: "subjects were given identification numbers that had been randomly assigned to study groups by a statistician using block randomization and the sample function of commercially available statistical software...Block sizes of 30 and 15 were used for the Laura district, and block sizes of 36 and 18 were used for the Delap district. Except for the statistician, all study team members were blinded until study completion" Comment: use of different block sizes administered by a statistician implies that a reasonable attempt was made to prevent those admitting participants from knowing upcoming assignments. We feel that this was probably done properly in this study
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Quote: "double‐blind" and "The syrups were matched for color, taste, and viscosity" Comment: participants and personnel would not know to which group a participant was assigned
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Quote: "The examiner was always blinded to study group assignment" Comment: outcome assessment appears to have been adequately blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	Only 6% of randomised participants were not included in the final analysis (Gp A: 6%; Gp B: 3%; Gp C: 9%). Attrition was actually 16% but most of these had at least an interim examination and were included in the intention‐to‐treat analysis. The reasons for drop‐out in each group were the same and in similar proportions (either moved off island or parent stopped giving syrup and withdrew) Comment: we do not believe that any of the above could pose a risk of bias significant enough to have led to a distortion of the true intervention effect
Selective reporting (reporting bias)	Unclear risk	There were no usable data reported for adverse events. This should be considered an important outcome in xylitol trials
Other bias	Low risk	Quote: "A single dental examiner" and "Compared with another examiner (P.M.), the study examiner demonstrated excellent reliability for caries diagnosis (interrater correlation coefficient, 1.00 at the prestudy dental examination and 0.96 at the midstudy examination)" Comment: we consider that the risk of differential diagnostic activity was low