Flickinger 2001.

Methods	Multicentre randomised, prospective, double‐blind, parallel group study. Unclear whether an independent observer was present. Follow‐up median 26 months (1 to 36 months)
Participants	Intervention: one isocentre Diagnostic criteria of patients included in the trial    Classical TN, 12 no prior surgical procedures Age mean (range) 68 (37 to 86) Gender 24 female, 19 male Severity (mean, SD)  No details Duration of condition years median (range)  Median 7 (1 to 31) years Number of patients 43 Intervention: two isocentres Diagnostic criteria of patients included in the trial Classical TN, 12 no prior surgical procedures Age mean (range) 69 (38 to 90) Gender 26 female, 18 male Severity (mean, SD)  No details Duration of condition years median (range) 9 (0.6 to 55) years Number of patients  44
Interventions	Intervention: one isocentre Type of intervention 75 Gy 50% at the centre. Volume 5.4 ± 0.44 mm3 Number patients 43 Intervention: two isocentres Type of intervention 75 Gy 50% at the centre, separated by 3 to 5mm. Volume 8.7 ± 1.1 mm3 Number of patients 44
Outcomes	Primary outcome Pain relief Secondary outcome Time to relapse, other procedures Adverse events
Notes	Treatment groups comparable at baseline; however, the pain severity is not mentioned
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	“Computer‐generated randomizations was blocked and stratified by institution. Enrollment was stopped after 43 of 45 intended patients were enrolled.”
Allocation concealment (selection bias)	Low risk	“Patients were unaware of the randomised treatment assignment to limit any placebo effect. We administered one isocenter radiosurgery in two equal portions, resetting the same treatment coordinates for the second portion to keep patients unaware of whether one or two different isocentres were treated. For two‐isocenter radiosurgery, the change in position between isocenters was imperceptible to patients.”
Blinding (performance bias and detection bias) All outcomes	Low risk	Reset treatment co‐ordinates after one treatment so seemed everyone was getting two. “Physicians and staff unaware of the randomisation assignment conducted the follow‐up evaluations.”  Double‐ blind, patient and assessor.
Incomplete outcome data (attrition bias) All outcomes	Low risk	“One patient randomized to two isocenters was completely lost to follow‐up the day after radiosurgery and was excluded, leaving 87 patients for this analysis” One dropout.
Selective reporting (reporting bias)	Unclear risk	“Follow‐up was 26 months (range 1 ‐ 36).  The product limit method of Kaplan‐Meier was used to calculate the actuarial rates of complete pain relief".  Did not show KM data, data collected at one time point.
Other bias	High risk	“Assignment conducted the follow‐up evaluations. The median follow‐up was 26 months (range 1 ‐ 36).”  Follow‐up short and big range