Whilst it is expected that this systematic review’s search criteria located the most relevant papers, we cannot claim to have yielded a complete, thorough and comprehensive list. Only 8 papers met the inclusion and exclusion criteria and were deemed suitable for this review. Two studies [8
] (deleted “of which”) used the same data set thus (deleted “therefore”) leaving us with 7 unique sets of data from which to draw our conclusions. One clear limitation and potential selection bias stems from our decision to only include studies published in the English literature.
A number of studies that investigated the comparative efficacy of PICC and CVC in critical patients had to be excluded on the grounds that data included medical patients. We excluded such studies to ensure that the conclusions drawn were accurate and a true representation of the surgical patient population.
There were some studies which investigated the use of CVCs and PICCs in surgical intensive care and seemed apparently suitable for our review but upon further analysis were excluded because although they referred to infections as a complication, they did not document catheter colonisation and/or CRBSI as one of their investigative parameters but rather focused on phlebitic, thrombotic and/or other such common complications.
Since the aim of the review was to compare CVC and PICC, a number of studies had to be excluded as they consisted of data relating to arterial catheters. We excluded these studies on the basis that arterial and venous catheters are entirely different entities with differing variable factors including haemodynamics and hence the sequelae of both are different. Studies which contained mixed venous and arterial catheter data but from which we could extract the venous catheter data were included, i.e. Sandoe et al [10
]. and Dimick et al [9
]. Further studies had to be excluded because although they referred to CVCs and PICCs, the findings were reported collectively and the two sets of data could not be demarcated from one another.
Only 1 of the 8 studies used a prospective randomised trial (PRT) study design; the others used a mixture of observational study designs. Whilst observational studies are prevalent in the infection control and critical care practice literature [16
] they do limit, by design, the conclusions of our review. However it should also be noted that even PRTs are susceptible to bias, specifically those relating to the way the studies were conducted and data analysed [17
All the authors followed published guidelines that clearly define positive catheter colonisation as either ≥15 colony forming units (CFU) by semi- quantitative culture [18
] or ≥10 [3
CFU/mL by quantitative technique from culture of the distal end of the catheter [19
]. The consistency in the definition of colonisation by the various studies reduced selection bias.
CRBSI is an ideal investigative parameter in comparing CVC and PICC performance as it represents the most serious form of venous catheter-related complication. However, the incidence of CRBSI is dependent upon the definition used. The Centre for Disease Control and Prevention (CDC) guidelines accepts various definitions for CRBSI [20
]; these are further subcategorised into two broad groups, namely clinical definitions and surveillance definitions.
Clinical definitions of CRBSI include positive signs of bacteraemia with the catheter as the only focus of infection after meticulous exclusion of all other potential sources. In addition, both peripheral blood culture and catheter tip culture must test positive for the same organism.
The less stringent surveillance definitions of CRBSI include the resolution of fever after the removal of a CVC suspected of infection; such definitions greatly inflate the true incidence of CRBSI as the bacteraemia may be secondary to sources other than the catheter such as the postoperative surgical site, pancreatitis, urinary tract infection…etc.
Notwithstanding Miyagaki et al [14
]. who, despite referring to CDC guidelines, failed to specify the CRBSI definition used, all the other studies used sub- definitions (Table
) that fell within the scope of the more stringent category of ‘clinical definitions of CRBSI’. Dimick et al [9
]., Sandoe et al [10
]. and Gunst et al [13
]. used the same definition of CRBSI, Bijma et al [7
]. and Pawar et al [11
]. used a different definition of CRBSI whilst Le Guillou et al [12
]. used yet another. Also, Gunst’s et al [13
]. retrospective design meant their study could not stringently follow the set definition.
It is important to be cognisant of the possibility that these varying definitions of CRBSI translate into differing thresholds for diagnosis of infection. (deleted “and” started a new sentence) Therefore the reported incidence of CRBSI would not only be different between studies but some may not have been a reflection of the true incidence of CRBSI. (deleted “and thus” started a new sentence). This could be pose another potential bias in our data analysis.
Our pooled data suggests that incidence of CRBSI is lower in CVCs (1.01%) than PICCs (3.23%) however Gunst et al [13
]., our only comparative study, found the opposite to be true (CVC 4.9%: PICC 2.7%). It could be argued that the design of some of the CVC studies was such that they didn’t include risk factors for CRBSI and so under reported the true incidence of CRBSI thus conferring onto CVC an unwarranted level of safety with regards to BSI but without the missing data all inferences are inconclusive.