We found a high proportion of primary and secondary literature and some prominent tertiary sources that attributed the efficacy of the combination of CHG and alcohol to CHG alone. The rates of incorrect attribution among the articles that we assessed ranged from 29% for catheters to 43% for surgery. The rates of incorrect and ambiguous attribution combined ranged from 42% for blood cultures to 65% for catheters. For surgery, there were more articles with incorrect (43%) than with correct attribution (36%). These conclusions were found at all levels of evidence gathering and knowledge translation, including primary clinical trials, systematic reviews, clinical practice recommendations and evidence-based guidelines.
The omission of alcohols in the process of evidence assessment can be seen, for example, in the draft CDC catheter guidelines 
which initially recommended CHG alone for central venous catheter insertion and maintenance. This was subsequently changed to CHG-alcohol in the final guidelines 
. We are unaware of the sequence of events, but assume that the change may have come through external submissions during the public comment phase. This change effectively rectified the section on skin antisepsis in the final guidelines. Another area of impact is a potentially mistaken rejection of alternatives or competitor products on the basis that they do not contain CHG, even if they have not been sufficiently tested in clinical trials. This appears to be affecting PVI plus alcohol in surgery, by way of negative implication 
In our analyses, we found good evidence favouring CHG-alcohol combinations over aqueous PVI, the most commonly tested alternative, in all three areas of skin antisepsis. This is a comparison of two active agents against one. However, this superiority does not hold against PVI plus alcohol or other competitors combined with alcohol, either due to equivalent performance in meta-analyses (for blood cultures) or a lack of relevant studies (for catheters and surgery). For blood cultures, alcohols alone may be effective, according to another analysis 
. For surgery, the question of CHG-alcohol versus iodine-alcohol is unresolved. For both blood cultures and surgery, we found no evidence that CHG alone is effective. For vascular catheters, the situation is more complex. There is evidence that CHG alone is superior to PVI alone for preventing colonisation, but its effect did not reach significance for CR-BSI. In contrast, CHG-alcohol was superior to PVI alone for both outcomes, colonisation and CR-BSI.
Each of the three applications has different biological and functional requirements. Blood culture collection requires immediate activity at the venipuncture site, but no prolonged action. Alcohols, with their strong immediate activity that typically exceeds those of CHG and PVI by about a factor of 10 
, fulfill this requirement well. Surgery requires significant immediate activity before incision and some persistent activity during the operation for several hours. Thus, surgical skin preparation is expected to benefit from the immediate action of alcohols plus persistent or enhanced activity from added CHG or PVI 
. Vascular catheter sites also require good immediate activity before insertion, but since catheters often stay in place for a week or more, good persistent action is also required. This requirement is fulfilled well by CHG 
Another fact deserves consideration. Most catheter studies used the antiseptics both before insertion and during maintenance (). Thus, when viewed strictly, it is not known exactly from the study results whether the antiseptics were more effective at the point of insertion or during maintenance, and which component would be better suited to which phase.
Our review has several limitations. First, it is partially hypothesis-driven, as indicated at the end of the Introduction section. This is unusual for systematic reviews, but nevertheless we used an explicit and rigorous systematic review methodology, including adherence to the PRISMA Statement 
. Second, our assessment of authors’ attribution is partially based on subjective judgement. This judgement was straightforward in virtually all articles classified as “incorrect”. We tried to err on the side of caution and assigned all articles in which this was less clear to the “intermediate” category. Third, since our search and assessment strategy focussed on any chlorhexidine-containing versus any other antiseptics, our review is not comprehensive in terms of capturing relevant comparisons between different non-CHG-based antiseptics in the three areas. Fourth, we faced limitations from heterogeneity in study design, differences in antiseptic compositions, and a lack of other relevant information while performing our analyses. Attribution was assessed for all RCTs, non-RCTs and systematic reviews, because this was not affected by study design. However, only RCTs were included in meta-analyses, in which only studies with the same basic antiseptic components were pooled together. Nevertheless, due to inherent differences between antiseptic products, we had to retain some variability of antiseptic concentrations and alcohol types in our analyses.
We were unable to trace the exact origins of the CHG misattribution. However – even though this is speculative – some observations in the literature provide a few potential reasons. First, some authors may regard alcohol simply as a solvent for CHG. This is reflected by the commonly-used term “chlorhexidine in alcohol” and references to alcohol as a “base solution” for CHG. Second, alcohol may not be universally regarded as an effective antiseptic. For example, wording in the CLSI blood culture guideline 
suggests that it is viewed as a cleansing agent at the venipuncture site. Third, some authors may be using the term “chlorhexidine” to actually mean the CHG-alcohol combination. This is suggested by some text passages in the draft CDC catheter guidelines 
. In any case, this would constitute incorrect usage of the term.
Our findings have broader implications. An important scientific principle – the fact that it is generally not possible to attribute effects to only one factor when several factors have been tested together – has frequently been overlooked. The individual published analyses may have been done correctly at a technical level of evidence assessment 
, but the conclusions appear incorrect. What are possible causes, and what are further implications?
First, it may be a matter of subjective views held by authors. If, for example, alcohol is regarded as a mere solvent for CHG, then authors are unable to draw appropriate conclusions. This means that the assessment of evidence remains susceptible to subjective influences, and this will continue to require attention in this area as well as in other subject areas. Second, this highlights the principle of biological plausibility. In the CHG example, plausibility could have been checked by what is known from microbiological studies of antiseptics 
; this would have indicated that alcohol is a key component. While biological plausibility is part of the Bradford-Hill Criteria in epidemiological studies 
, there is currently no explicit requirement to address this in clinical trials and systematic reviews 
. However, we think this should become a requirement.
Our findings also have potential implications for patient safety. When following recommendations to use “chlorhexidine”, caregivers may inappropriately use CHG on its own, in aqueous solution, as this is readily available. The clinical impact from blood cultures and vascular catheterisation may be small, because contaminated blood cultures do not directly harm patients and CHG alone appears to exert some protective effect in vascular catheterisation. However, tangible negative consequences may arise in surgery, because marked differences in surgical infection rates have been observed between different antiseptic regimens 
. Conversely, if caregivers are unaware of the presence and significance of alcohols, they might accidentally use alcohol compounds for antisepsis on mucous membranes, where they are contraindicated.
In summary, there is good evidence that CHG-alcohol is superior to aqueous PVI – an important competitor – in all three areas of skin antisepsis. However, this does not apply to competitors combined with alcohols. The perceived efficacy of CHG in skin antisepsis is often in fact based on evidence for the efficacy of the CHG-alcohol combination. In conjunction, the role of alcohol has frequently been overlooked in evidence assessments. This has broader implications for knowledge translation as well as potential implications for patient safety.