Vallés 2008.

Methods	Single‐centre RCT (Spain) Study period: 1 Jan 2005 to 3 June 2006 Setting: adult medical‐surgical ICU in a university hospital
Participants	Patients requiring a CVC Number of participants: 420 Number of catheters; 998 (631 analysed) Age: mean age from 60 to 61 years (SD 16‐17) (reported separately in three groups) Sex: not reported.
Interventions	3‐arm comparison of the following skin antiseptic regimens applied prior to CVC insertion and every 72 h thereafter. Intervention A: 10% aqueous povidone‐iodine solution. Intervention B: 0.5% alcoholic chlorhexidine‐gluconate solution. Intervention C: 2% aqueous chlorhexidine‐gluconate solution.
Outcomes	Catheter colonisation "Catheter related sepsis" (catheter‐related BSI) "Catheter related bacteraemia" Mortality was not specified as an outcome in the methods, but mortality figures were reported in the "Patient characteristics" table. Outcomes assessed at various points during in‐patient stay.
Notes	Funding source: not stated.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	'Methods', 'Study design': The random sequence was generated by "[b]y use of a blinded block randomisation schedule"
Allocation concealment (selection bias)	Unclear risk	Not adequately reported
Blinding of participants and personnel (performance bias) All outcomes	High risk	Although not stated by the authors, blinding to patients and caregivers appeared highly unlikely, as the antiseptic solutions used differed in appearance.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	'Methods', 'Bacteriologic methods': "The microbiologists who performed the catheter‐tip cultures had no knowledge of the antiseptic group to which the catheter had been assigned."
Incomplete outcome data (attrition bias) All outcomes	High risk	Authors excluded from analysis 367/998 (36.7%) of the catheters initially randomised for various reasons (Figure 1 of the paper). They excluded 279 catheters post enrolment because they did not meet the inclusion criteria. However, among these excluded catheters, the reason given for 179 of them was that they were "not cultured". It was unclear what the underlying reasons were for failure to obtain culture in these catheters, and whether the excluded data here were missing at random. Trialists excluded 88 further catheters because they were inserted beyond 72 h after discharge from ICU. These 88 catheters were evenly distributed among the 3 assigned groups (61 between the 2 chlorhexidine groups and 27 in the povidone‐iodine group). However, following the construction of the best‐ and worst‐case scenarios using the dropouts, the direction of the effect estimates swung from significantly favouring the chlorhexidine group (best‐case scenario for chlorhexidine group) to significantly favouring the povidone‐iodine group (worst‐case scenario for chlorhexidine group). It was unclear whether the authors followed the intention‐to‐treat principle by analysing all available data according to the originally assigned groups, as there was no mention of participants who crossed over to the other group. We accorded the study high risk in this domain due to the high absolute dropout rate including the 179 catheters that were not adequately accounted for, as mentioned above, and the vulnerability of the result estimates to best‐ and worst‐case scenarios.
Selective reporting (reporting bias)	Low risk	Authors reported all 3 outcomes stated in the 'Methods', namely, catheter colonisation, catheter‐related BSI ("catheter‐related sepsis") and catheter‐related bacteraemia in sufficient detail in the 'Results'. In addition, they also reported the important outcome of mortality in the "Patient characteristics" table. although this was not a pre‐specified outcome in the methods..
Other bias	High risk	The study employed a block randomisation schedule with high likelihood that blinding of participants and personnel were not achieved. This posed a risk to the integrity of the random sequence, which would be vulnerable to disruption following educated guesses by those involved in the study on the likely assigned group of the future participants. There was a serious unit of analysis issue in which the number of catheters analysed exceeded the number of participants by over 50%, and the major outcomes were reported using catheters as the units.