Winner 2013.

Methods	Retrospective review
Participants	1004 patients between January 1991 and December 2005 with stage IV colon adenocarcinoma
Interventions	281 patients (28.0%) underwent surgery
Outcomes	Hospitalisation length of stay: non‐surgical 7 days; surgical 11 days, P value < 0.001 30‐day mortality: non‐surgical 30.8%; surgery 18.5%, P value = 0.003 Absolute difference in medical survival between treatment surgery verses non‐surgery 56 days
Notes	Uses Surveillance, Epidemiology, and End Results ‐ Medicare database, which tracks 26% of patients with cancer in US population by searching for coded information on billing for procedures and services rendered Reduction in use of surgery over time ‐ 32.4% 1991 and 26.2% in 2005 Stenting not included as not coded for
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	Coding/billing information reflected that patients were allocated to surgery by clinician/patient choice and retrospectively reviewed
Allocation concealment (selection bias)	High risk	Coding/billing information reflected that patients were allocated to surgery by clinician/patient choice and retrospectively reviewed
Blinding (performance bias and detection bias) All outcomes	High risk	Coding/billing information reflected that patients were allocated to surgery by clinician/patient choice and retrospectively reviewed
Blinding of participants and personnel (performance bias) All outcomes	High risk	Coding/billing information reflected that patients were allocated to surgery by clinician/patient choice and retrospectively reviewed
Blinding of outcome assessment (detection bias) All outcomes	High risk	Coding/billing information reflected that patients were allocated to surgery by clinician/patient choice and retrospectively reviewed
Selective reporting (reporting bias)	Unclear risk	Not reported by surgical procedure ‐ unclear if this information was available via coding or not