Bello 2011.

Study characteristics
Methods	Study design: RCT parallel group Conducted in: Department of Oral and Maxillofacial Surgery, National Hospital, Abuja, Nigeria Number of centres: 1 Recruitment period: not stated
Participants	Inclusion criteria: patients referred for extraction of 1 or 2 impacted mandibular third molars Exclusion criteria: patients with acute pericoronal infection, systemic diseases, or bleeding disorder, patients receiving steroid therapy or contraceptives, and smokers were excluded. Patients whose extraction procedure took more than 35 minutes were also excluded. Number randomised: unclear Number evaluated: 82
Interventions	Partial versus complete wound closure Group  A (n = 40): partial wound closure was achieved using 4 interrupted sutures leaving a window communicating with the oral cavity Group  B (n = 42): complete wound closure was achieved using 5 interrupted sutures that sealed off communication with the oral cavity All procedures done under local anaesthetic by the same surgeon. All participants received pre‐emptive antibiotics (amoxicillin and metronidazole) for 5 days and diclofenac for pain and inflammation for 3 days.
Outcomes	Pain (VAS 0 to 10) reported daily for 7 days. Maximal interincisal distance (as % of baseline value) and swelling (difference from baseline) were evaluated on days 2, 5, and 7. Numbers of postoperative complications (dry socket, infection, and secondary haemorrhage) were also noted.
Notes	Sample size calculation: not reported E‐mail sent 12 March 2012 requesting further information about the methods used. Reply received 21 March 2012 with unpublished data.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "randomly allocated" Comment: allocated by drawing lots to either partial or total (e‐mail correspondence)
Allocation concealment (selection bias)	Low risk	A paper was drawn by the assistant and shown to the surgeon prior to suturing (e‐mail correspondence).
Blinding (performance bias and detection bias) patient	Unclear risk	Not mentioned
Blinding (performance bias and detection bias) assessor	Unclear risk	Assessment done using a "clean proforma", but it is possible that the assessor knew which group the participant was in.
Incomplete outcome data (attrition bias) All outcomes	Low risk	90 patients were screened for inclusion but unclear how many were randomised. Patients were excluded when the procedure took more than 35 minutes (n = 3). 8 patients did not return for follow‐up assessment, 5 and 3 in each group.
Selective reporting (reporting bias)	Low risk	Planned outcomes of pain, trismus, swelling, and complications reported.
Other bias	Low risk	No other sources of bias identified. This study excluded patients whose surgeries exceeded 35 minutes; the reason for this was not explained.