CAESAR.

Methods	Study design: intention to treat Method of randomisation: Randomisation was designed with equal probability (1:1 ratio) of assignment to either immediate endovascular repair or surveillance by means of a computed‐generated random number list, stratified by centre using a permuted block design and carried out online through the Internet.         Concealment of allocation: unblinded
Participants	Country: Italy Number: 360 Sex: men (n = 345) and women (n = 15) Age: 50 to 79 years Inclusion criteria: people with small (4.1 cm to 5.4 cm) asymptomatic AAAs, without high surgical risk, and who would have benefited from immediate repair. Patients were excluded if they had severe co‐morbidities or a suprarenal/thoracic aorta ≥ 4.0 cm, needed urgent repair, or were unable or unwilling to give informed consent or follow the protocol.
Interventions	Treatment: surgery, n = 182 of whom 175 had immediate endovascular surgery; 6 declined treatment and 1 underwent open repair according to person's choice. Surveillance, n = 178 of whom 172 had aneurysm repair when they met the criteria below (6 patients had endovascular repair against protocol: 5 per patient choice and 1 with a surgeon not participating in the study). Participants assigned to immediate endovascular repair underwent aneurysm repair a median of 22 days after randomisation, while participants assigned to surveillance were seen every 6 months and repair allowed if the aneurysm grew to 5.5 cm diameter in size, rapidly increased in diameter (> 1 cm/year), or became symptomatic.
Outcomes	Primary: mortality from any cause Secondary: 1) aneurysm‐related deaths (defined as death caused directly or indirectly by aneurysm rupture or aneurysm repair), 2) aneurysm rupture, 3) perioperative (30 days or inpatient) or late adverse events (defined according to SVS/AAVS reporting standards), 4) conversion to open repair, 5) loss of treatment options (anatomical suitability for endovascular repair), and 6) aneurysm growth rate
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Randomisation was designed with equal probability (1:1 ratio) of assignment to either immediate endovascular repair or surveillance by means of a computed‐generated random number list, stratified by centre using a permuted block design and carried out online through the Internet.
Allocation concealment (selection bias)	Unclear risk	Concealment of allocation: unblinded
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Cannot blind participants
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Unlikely given the study outcome (mortality) and low lost‐to‐follow‐up rate. Vital status was assessed using the same methodology for both participants in the immediate‐repair group and participants in the routine ultrasound surveillance group‐–in case misclassification occurred this would have been non‐differential and its impact on the study results would be limited.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Unlikely given the study outcome (mortality) and low lost‐to‐follow‐up rate
Selective reporting (reporting bias)	Low risk	Authors published findings on the main study outcome of this review.
Other bias	High risk	Conflicts of interest: The sponsor of the study, Cook Medical, withdrew.