Should the Guidelines for Management of Central Venous Catheters in Patients with Candidemia Be Changed Now?

For >20 years, efforts have been made to define the correct approach to the management of central venous catheters (CVCs) in patients with candidemia who have an indwelling CVC at the time of detection of their candidemia. The limitations of previous studies have been reviewed elsewhere [1]. Additionally, thoughtful commentaries have been written [2]. The underlying concept supporting catheter removal is based on the knowledge that catheters have been identified as predisposing factors in the development of candidemia and are known to frequently have biofilms of Candida species attached to them in patients with candidemia. The precise mechanism of the association of the CVC with the occurrence of candidemia is not known, but the possibilities have been carefully reviewed [2]. There is a nontrivial cost of removal and replacement of catheters, a certain risk associated with the insertion of replacement catheters, and, in certain cases, a negative impact on the quality of life of the patients (especially those needing long-term intravenous catheterization) who need a surgical procedure for the insertion of replacement catheters. Thus, any recommendation to routinely remove these catheters must be based on reasoned extrapolation of clinical data. There are 2 recent guideline statements from the Infectious Diseases Society of America containing recommendations regarding CVCs in patients with candidemia. In the article with guidelines for Candida management [3], the recommendation reads, “Central venous catheters should be removed when candidemia is documented, if at all possible” [p 514] and, regarding neutropenic patients, “Intravenous catheter removal should be considered (B-III)” [p 515]. In the article with guidelines for the management of intravascular catheter-related bloodstream infection (CRBSI) [4], the statement reads, “Catheters should be removed in cases of CRBSI due to Candida species (A-II)” [p 35]. The assignment of the II grade indicates that a definitive, properly randomized, controlled trial does not exist. The III grade indicates that the recommendation is based mainly on opinions and experience and not on properly randomized, controlled trials. The assignment of the II or III score in both guideline articles and the wording of the statements imply that the recommendations are “qualified.” There is no overlap in the authorship of these 2 guideline articles, and there are 24 authors between the 2 studies who have come to congruency with their recommendations. Nucci et al [7] have performed a commendable study of a large database containing data on 842 adults studied in 2 recent phase III candidemia trials [5, 6] designed primarily to determine the efficacy of antifungal drugs in the treatment of candidemia. Data regarding early and late removal of catheters were collected in these 2 studies. As Nucci et al [7] point out, neither study was a prospective, blinded, randomized study designed to determine the effect of catheter removal or replacement. Such a study is needed to precisely address the effect on outcomes related directly to catheter removal. It is doubtful that such a study will ever be performed. Therefore, the retrospective subgroup analysis conducted by Nucci et al [7] on the effect of early (within 24 or 48 h after treatment initiation) removal of CVC on selected outcomes in this patient population has the potential to influence future guideline development and clinical decision making. In the authors' univariate analysis strategy, CVC removal did not improve time to mycological eradication or recurrent candidemia but was associated with better treatment success and improved survival. Recognizing the limitations of their non randomized data set, the authors conducted a multivariate analysis to try to account for potential clinical confounders. The benefits of CVC removal appear to have been lost in the multivariate analysis. Does this negative result mean that CVC removal is without value? Several considerations need to be taken into account in deriving conclusions from this retrospective analysis. The impact of unrecognized factors that may have influenced the loss of benefit seen in the multivariate analysis cannot be determined from the data presented. The circumstances influencing the decision by the individual investigators to remove or replace catheters are not described and may have varied considerably over the many sites and the many different countries involved, depending on institutional practices. Although 1 of the 2 studies called for either early or late removal prospectively [5], the catheter manipulation was not done using a randomized strategy. It is possible that fewer catheter manipulations were done in patients who were more ill and who had poor outcomes. Although some clinical factors consistent with this suggestion were included in the multivariate analysis, their ability to completely capture the complexity of the clinical situations may be questioned. If unidentified clinical factors contributed to the decision to remove the CVC, skewing of the data may have occurred. A propensity analysis [8, 9] may have identified factors influencing the decisions to remove or retain the catheters. A multivariate analysis can be performed with a variety of strategies. In this multivariate analysis, a large number of potential confounding factors was chosen a priori, and an objective rule was then used for selection. Statistical power for evidence of a lack of benefit of CVC removal after covariate adjustment will be low if adjustment does not reduce the margin of error in estimating the independent effect of CVC removal, in addition to reducing the confounding bias in the unadjusted estimate. This loss of power can be seen most importantly in Nucci et al's Table 5, in which the effect of early catheter removal on survival is shown. For CVC removal within 24 h, the 95% confidence interval for the odds ratio for 28-day survival is 0.79–1.67, with a point estimate of 1.15. Thus, this wide 95% confidence interval does not rule out a 67% improvement in the odds ratio for survival—an effect that clearly would be clinically meaningful. Similarly, with CVC removal after 48 h, for which the 95% confidence interval is 0.85–1.75 and the point estimate is 1.23, a 75% improvement in the odds ratio for survival cannot be excluded. Many clinicians would consider those levels of improvements in survival highly meaningful. The fact that the point estimates favored CVC removal and that clinically meaningful results could not be excluded on the basis of the confidence interval directly illustrates a lack of power for evidence against CVC removal. Of additional interest in explaining why the benefit from catheter removal on treatment success and survival did not persist in the multivariate analysis would be knowing what influence the potential confounding variables had on each other. This effect might have been identified with an interactions analysis, as exemplified in a recent study [10]. At the next update of the guidelines, this important study will certainly be reviewed in detail because of the strong statement made in the discussion, “We conclude that early CVC removal in nonneutropenic adults with candidemia does not influence patient outcomes and that recommendation to remove all CVCs in nonneutropenic patients may not be justified” [7]. Undoubtedly, a value judgement will be made about the strength of this conclusion, following further critical statistical analysis of this study and in the broad context of the totality of the available data. Whether wording of the guideline statements will be changed will be up to the large expert panels. In the meantime, Nucci et al [7] have suggested that CVC management should be individualized for each patient. That opinion is highly commendable. In fact, carefully individualizing any medical intervention to each patient is the hallmark of good medical practice. Obviously, “guidelines” are just what their name implies, and they are guides, not “rules” to be applied to every patient situation without careful risk-benefit analysis. For example, probably all clinicians would agree that removing a CVC that is no longer medically required would be prudent medical care, regardless of its relationship to candidemia. It is not possible to have a new CVC complication if a CVC is not present. However, the theoretical risks of not removing a CVC in a patient with candidemia are clear, and the expert guidelines remain the current best synthesis of the available data. We thank Peter D. Christenson and Loren G. Miller for helpful suggestions. Potential conflicts of interest. E.B. is a consultant to Sigma tau Research, Merck Research Laboratories, Schering-Plough, 3D Communications, Worldwide Clinical Trials, ARCA biopharma, Consumers Healthcare Products Association, Catabasis, Dyax, NCI Consulting, NovaDigm Therapeutics, Kowa Research, Nektar Pharmaxis, Medtronics, Intermune, and MedImmune and has equity in Calistoga Pharmaceuticals and Catabasis Pharmaceuticals. J.E.E. is a consultant to Cerexa; is cofounder of and has equity in NovaDigm Therapeutics; and has received recent grant support from Astellas, Pfizer, and Merck.