Brothwell 1999.

Methods	FLUOROSIS STUDY Country of study: Canada Geographic location: Wellington and Dufferin (neighbouring counties), South‐Western Ontario Year of study: 1996‐1997 (academic year) Year of change in fluoridation status: NA Study design: cross‐sectional
Participants	Inclusion criteria: children resident in Wellington‐Dufferin‐Guelph Health Unit area; parental consent; children aged 7‐8 years Exclusion criteria: children with non‐erupted or insufficiently erupted central incisors; children absent on day of examination Other sources of fluoride: amount of toothpaste usually used ("48.9% use > pea sized amount, 365/747"); fluoride supplements ("14.5% take supplements, 107/740"); age started brushing; use of mouthwash ("4% routinely use fluoridated mouthwash, 30/752"); breast/bottle fed; whether toothpaste used when brushing Social class: household income; highest level of education received. "It is likely that respondents under‐represented the disadvantaged segment of the population. How the low response rate in this subgroup affects the estimates of prevalence is unknown; however, it is unlikely to be a major source of bias." Ethnicity: not stated Residential history: "The questionnaire assessed … years at current residence", 39% lifelong residents (293/752); 64.8% (487/752 resided at tested source from before the age of 3 (fluorosis‐sensitive period – multivariate analysis restricted to these 487 participants) Other confounding factors: breast‐feeding duration
Interventions	Group 1: ≥ 0.7 ppm (natural fluoridation) Group 2: < 0.7 ppm (natural fluoridation)
Outcomes	Dental fluorosis (TSIF score > 1) Age at assessment: 7‐8 years
Funding	Not stated
Notes	Data extracted from Brothwell 1999 differs from that presented in CRD review
Risk of bias
Bias	Authors' judgement	Support for judgement
Sampling	Unclear risk	Children were selected via schools, however insufficient detail was reported regarding sampling
Confounding	High risk	Bivariate analysis showed that fluoridated mouthwash use and professional fluoride treatments were significantly associated with fluorosis prevalence, however, the data were not reported/presented in a manner which demonstrated adjustment for imbalance at baseline occurred, or was measured well and controlled for
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Testing of water samples for fluoridation level was conducted after screening examination (at the University of Toronto); examinations conducted by a single dental hygienist (in school clinics). It does not appear that, despite the lack of any attempt to blind being reported, that blinding would have had any effect on reducing bias
Incomplete outcome data (attrition bias) All outcomes	High risk	Significant missing data (e.g. 34 participants from the water sample)
Selective reporting (reporting bias)	High risk	Comment: there is much that is either not reported in a sufficient manner to be able to glean the necessary information from (i.e. TSIF scores against fluoridation levels of water samples), or has significant missing data (e.g. 34 participants from the water sample) and so is difficult to draw the conclusions required for this review. No evidence of protocol in advance of obtaining data/undertaking analysis
Other bias	Low risk	Reporting dental fluorosis as TSIF score > 1 rather than ≥ 1 puts the results at risk of misclassification bias