Tammiala‐Salonen 1999.

Methods	Single centre, placebo‐controlled parallel RCT
Participants	37 BMS patients Mean age 58.6 years (range 39 to 71) Sex: 37 F (F 100%) Inclusion/exclusion and diagnostic criteria appropriate
Interventions	Intervention category: antidepressants and antipsychotics Group 1: (n = 18) trazodone 200 mg daily Group 2: (n = 19) placebo Duration: 8 weeks
Outcomes	VAS at 0, 2, 4 and 8 weeks using 100 mm VAS scale (endpoints were 0 = no pain, 100 = worst pain possible) Short McGill Pain Questionnaire (MPQ) at baseline and 8 weeks ‐ "intensity and character of the pain were further defined by the use of the Finnish version of the McGill Pain Questionnaire" Beck Depression Inventory (BDI) Global assessment at 8 weeks
Source of funding	Finnish Dental Society
Notes	Groups differed at baseline with regard to pain intensity Number of dropouts was higher in those who were depressed SES: not reported Conflict of interests: not reported Data analysis: per protocol
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quotation: "Randomization was performed in blocks of 6 by the manufacturer of the drug (Orion)" Comment: third party randomisation
Allocation concealment (selection bias)	Low risk	Quotations: "The randomization code was not opened during the trial"; "The examiner could not guess the treatment of the subjects" Comment: probably achieved
Blinding (performance bias and detection bias) Blinding of participants	Low risk	Quotations: "Identical capsules of trazodone and of passive placebo were packed in the same way"; "nothing suggested that the blinding had not remained intact for the patients"; "Seven subjects in the trazodone group and 2 in the placebo group failed to finish the trial because of side effects, mainly because of dizziness" Comment: probably achieved
Blinding (performance bias and detection bias) Blinding of outcome assessors	Low risk	Quotation: "The examiner could not guess the treatment of the subjects" Comment: blinded outcome assessment; patient‐reported outcomes only
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Comment: reasons for dropouts were reported appropriately. Unclear as to how these data were handled – no mention of numbers analysed for the outcomes
Selective reporting (reporting bias)	Unclear risk	Comment: prespecified outcomes were reported but not completely – although VAS measurement on speaking, eating and suffering – no data provided – only mention that there was no significant differences between the groups. McGill pain score raw data not provided, reported to be no significant difference in standard deviations for VAS and Beck data only shown graphically with no raw data provided
Other bias	Low risk	Comment: no other risks of bias were noted

ALA = alpha lipoic acid; BMS = burning mouth syndrome; F = female; ITT = intention‐to‐treat; M = male; OHI = oral hygiene instruction; ppm = parts per million; QoL = quality of life; RCT = randomised controlled trial; SD = standard deviation; SES = socioeconomic status; VAS = visual analogue scale.