Heijl 1997.

Methods	3‐year follow‐up split‐mouth study including 33 patients. 3 drop outs at 16 months (tooth extractions in 2 cases and accident for 1 patient).
Participants	Patients in good general health and motivated for good oral hygiene. Teeth with PPD greater or equal to 6 mm and IBD greater or equal to 4 mm. 1‐, 2‐ and 3‐wall defects were included. All patients had received systematic periodontal treatments (repeated debridement in some cases supplemented with antimicrobial and surgical treatment over long periods of time). Age ranging between 33 and 68; 7 males and 26 females recruited at 3 specialist clinics.
Interventions	Emdogain versus flap surgery and placebo.
Outcomes	FMPS and for experimental teeth only: BOP, PAL, PPD, IBD on standardised intraoral radiographs at baseline, 8, 16 months and 3 years. Tooth loss, postoperative infections and adverse events. Additional intrasurgical measurements were taken. 1‐year data used.
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "The sites were distinguished by their tooth number (18 through 48) and the randomization code specified the treatment assignment for the site with the lowest as well as highest tooth number. The randomization process targeted one of the sites for test treatment and the other site for control treatment. Patient numbers were assigned in chronological order as patients were enrolled in the trial". Author's reply: "Randomization codes were computer generated in blocks".
Allocation concealment (selection bias)	Low risk	Quote: "At the time of periodontal surgery, and only after the first surgical site was fully prepared, the envelope containing the randomisation code was opened to expose treatment assignments".
Blinding (performance bias and detection bias) All outcomes	Low risk	Quotes: "Readings of all radiographs were performed by a separate, blinded examiner and in a randomised fashion". "All re‐examination measurements were made by the same blinded investigator who made the initial measurements". Comment: Assessors were blinded both for the clinical and radiographic outcomes.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Outcome data are presented in Table 3: Mean values for pocket depth, clinical attachment level and radiographic bone level. Comment: No missing outcome data. Drop outs are explained adequately.
Selective reporting (reporting bias)	Low risk	All the primary pre‐specified clinical outcomes are properly presented in Table 3: Mean values for pocket depth, clinical attachment level and radiographic bone level. Adversed events are reported in Safety (AEs) section. 4 teeth were extracted, 2 for each group. REC and aesthetics were not evaluated as treatment outcomes.
Other bias	Unclear risk	The trial was supported by the manufacturer.