Franssen 1986.

Methods	Design: RCT Number of centres: NA Treatment duration: 12 weeks and 48 week extension Flare design: Yes Wash‐out period: Yes (2 weeks or until flare) Time point of assessments: BL, 2, 4, 6, 8, 10, 12 weeks
Participants	Inclusion criteria: 1. Male; 2. Age 18 to 55; 3. Definite AS confirmed by two or more of the following symptoms: back pain, morning stiffness and progressive limitation of movement; 4. Flare‐up after wash‐out (defined as a worsening of the patient's condition in which pain and stiffness was an essential component requiring treatment). Exclusion criteria: 1. Active peptic ulcer or other significant internal or neurological disease; 2. Hypersensitivity or serious side effects while taking NSAIDs; 3. End‐stage and incapacitating disease (ARA functional class IV). Classification: NA Diflunisal (1000 mg): Number of participants: 19 Number of completers: 14 Phenylbutazone (400 mg): Number of participants: 19 Number of completers: 17 All participants: Age (mean): 37 (range 20 to 54) Male (%): 100 Symptom duration: NA Disease duration: NA HLA‐B27 positive (%): NA
Interventions	Diflunisal (1000 mg) vs Phenylbutazone (400 mg) Co‐medication: Patients received only one active compound. Any programme of physical therapy remained unchanged during the double‐blind period. No other drug intake was allowed during the study except 500 mg acetaminophen as a supplementary analgesic.
Outcomes	Extracted outcomes: Twelve weeks results (included in the meta‐analysis) Pain on Likert scale Withdrawals due to adverse events Duration of morning stiffness ESR Number of adverse events per organ system Forty‐eight weeks results (not included in the meta‐analysis) 1. Pain on Likert scale (BL, post‐treatment after 48 weeks (± SD)) (scale 0 to 4, higher is worse) Diflunisal: 2.8 ± 0.6 (N = 19), 1.6 ± 0.9 (N = 14) (P < 0.01 vs BL) Phenylbutazone: 2.6 ± 0.7 (N = 19), 1.5 ± 1.2 (N = 17) (P < 0.01 vs BL) 2. Withdrawals due to adverse events (post‐treatment after 48 weeks) Diflunisal: n = 3 (total N = 19) Phenylbutazone: n = 3 (total N = 19) 3. Duration of morning stiffness (BL, post‐treatment after 48 weeks (± SD)) (in hours, higher is worse) Diflunisal: 2.1 ± 2.0 (N = 19), 1.4 ± 1.3 (N = 14) (P = not significant) Phenylbutazone: 2.7 ± 2.5 (N = 19), 2.0 ± 2.7 (N = 17) (P = not significant) 4. ESR (BL, post‐treatment after 48 weeks (± SD) (in mm/hr, higher is worse) Diflunisal: 31 ± 20 (N = 19), 20 ± 22 (N = 14) (P < 0.05 vs BL) Phenylbutazone: 19 ± 12 (N = 19), 17 ± 13 (N = 17) (P = not significant)
Notes	The outcomes chest expansion and Schober's test were also presented, but these data could not be used due to presentation in graphs from which the data could not be extracted. Funding source: Supported with a grant from Merck Sharp and Dohme/Chibret, Haarlem, The Netherlands.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"all patients were allocated at random to receive either…" Probably done, but no further information provided on sequence generation.
Allocation concealment (selection bias)	Unclear risk	"all patients were allocated at random to receive either…" Probably done, but no further information provided on allocation concealment.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	"Patients received only one active compound together with the corresponding placebo of the alternative drug."
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	"double blind"; "Clinical assessment was made weekly by the same investigator for each patient." Probably done, but no further information provided on blinding outcome assessor.
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	A completers‐analysis is done, so possible attrition bias.
Selective reporting (reporting bias)	Low risk	All pre‐specified outcomes according to the methods section, are reported in the results section.
Other bias	Low risk	"The average patient compliance in the double‐blind period was 95±4% for DIF and 96±5% for PBZ..."; high compliance, no other bias detected.