Impact of early central venous catheter removal on outcome in patients with candidaemia


Corresponding author and reprint requests: D. Rodriguez, Infectious Diseases Division, Hospital Universitari Vall d'Hebron, Pg. Vall d'Hebron 119-129, 08035 Barcelona, Spain


Removal of central venous catheters (CVCs) from candidaemic patients is considered the reference standard of care, although this practice is not always possible. The impact of prompt catheter removal on outcome was investigated by analysing data from an active population-based surveillance study in Barcelona, Spain. Patients with candidaemia and a CVC were identified between January 2002 and December 2003. Cases with CVC removal within 2 days were classified as having early CVC removal. Outcome, defined as in-hospital mortality 2–30 days after diagnosis of candidaemia, was determined among hospitalised adults using univariate, Kaplan–Meier and multivariate logistic regression analysis. Outpatients, paediatric patients and those who died or were discharged within 2 days were excluded. The study identified 265 patients with candidaemia and a CVC. Median time from diagnosis of candidaemia to catheter removal was 1 day (range 0–29 days). Overall, 172 patients met the criteria for inclusion in the outcome study. Patients with early CVC removal differed significantly from those with delayed CVC removal. According to univariate, Kaplan–Meier and multivariate analysis, the marker most predictive of in-hospital mortality among candidaemic patients with CVCs was severity of illness. These data suggest that timing of CVC removal may best be determined after carefully considering the risks and benefits to individual patients.


Several studies have identified central venous catheters (CVCs) as constituting an increased risk for the development of candidaemia [1–3]. Although removal of CVCs from candidaemic patients is considered the reference standard of care [4,5], this practice is not always possible. Previous studies have investigated the impact of CVC removal on outcome, often defined as death, and the persistence or recurrence of candidaemia [1,4,6–14]. However, few data exist concerning the most appropriate time for CVC removal, or the sub-population of patients that would benefit most from such an intervention. A recent study, limited to patients suffering from cancer, attempted to answer these questions, and revealed an association between early CVC removal and improved outcome [14]. In order to provide a better description of a population with catheter-related candidaemia, and to understand the effect of early CVC removal on these patients, the present study analysed data from an active, population-based surveillance study of candidaemia [13].

Patients and methods

Study population and definitions

Prospective population-based surveillance for candidaemia was conducted in greater Barcelona (population 3.9 million) between 1 January 2002 and 31 December 2003 [13]. Fourteen major hospitals participated, ranging in size from 214 to 1295 beds. All candidaemic patients with a CVC in place for at least 24 h at the time of candidaemia were included in this analysis.

A candidaemia patient was defined as a resident of Barcelona for whom a Candida sp. was isolated from blood. Cases were defined as having primary candidaemia if there was no apparent portal of entry or if the infection was probably catheter-related. Secondary candidaemia cases were defined as patients with infections that occurred after a potential source of infection was identified. Catheter-related candidaemia was recorded when a patient had primary candidaemia and either (i) a catheter tip colonised with the same Candida sp. that was isolated from the bloodstream, or (ii) a quantitative blood culture sample obtained through the CVC that showed a ≥five-fold greater colony count than a concurrent peripheral venous quantitative blood culture. Significant colonisation of the catheter was defined by the isolation of >15 CFU of a Candida sp. using the roll-plate semiquantitative culture method [4,15].

Adult candidaemic patients with an APACHE II score ≥20 were defined as having a high severity of illness (SOI); adult candidaemic patients with an APACHE II score <20 were defined as having a low SOI.

To explore the effect of early catheter removal on outcome, patients with early CVC removal were defined as those who had the CVC removed on day 0 or 1 following the date of the incident candidaemia. Outcome was defined as in-hospital mortality measured at day 2–30 post-candidaemia onset. Patients who died or were discharged before day 2 post-candidaemia onset were excluded from the outcome analysis because it was not possible to classify these cases into either the early or delayed CVC removal categories. Inclusion of patients who died before a decision was made regarding removal could bias results towards showing an association between non-removal and death.

Outpatients, paediatric patients and patients with multiple CVCs were also excluded. Patients who developed candidaemia as outpatients would not have been subject to the outcome measure of in-hospital mortality. Paediatric patients were excluded because no SOI score was available. Patients with multiple CVCs were omitted because many did not have all CVCs removed on the same day, and therefore it was not possible to define the exact date of removal. Patients with multiple CVCs removed on the same day were also excluded, as inclusion of these patients was felt to bias the study population towards the group with CVCs removed.

Patients with early CVC removal were compared to those with delayed CVC removal by comparing the time to death for each of these groups using a Kaplan–Meier analysis. Kaplan–Meier curves were compared using the log-rank test. In addition, in-hospital mortality at 2–30 days post-candidaemia onset was examined by comparing early removal to delayed removal by univariate analysis. Stratified analysis was conducted among subgroups of patients in intensive care units (ICUs), those with preceding surgery, and those with malignancies. Statistical analysis was performed using SPSS software v.12.0 (SPSS Inc., Chicago, IL, USA) and SAS software v.8.2 (SAS Institute, Cary, NC, USA). Categorical variables were compared using chi-square analysis or Fisher's exact test, as appropriate. Median values were compared using the Wilcoxon rank sum test. Variables significant at p <0.2 following univariate analysis, and other variables felt to confound the outcome, were included in a multivariate model that determined factors associated with in-hospital mortality at days 2–30 post-candidaemia.

Microbiological methods

Detection of candidaemia and identification of isolates to the species level were performed at the participating laboratories. Isolates were sent to the Mycology Reference Laboratory, National Center for Microbiology, Madrid, Spain for confirmation of species identity and antifungal susceptibility testing. Catheter tips were cultured by means of the semiquantitative roll-plate method described by Maki et al. [15].


Cases with CVCs

During the study period, 265 cases of candidaemia were detected in patients with CVCs. The majority of cases were episodes of primary candidaemia (n = 251, 95%); 145 (55%) of these had no apparent portal of entry, and 106 (45%) were probably catheter-related. Fourteen (5%) cases were classified as secondary candidaemia (Table 1).

Table 1.   Characteristics of 265 patients with candidaemia and a central venous catheter (CVC), stratified according to type of candidaemia
CharacteristicNumber (%)
non-CVC- related
(n = 145)
(n = 106)
(n = 14)
All cases
(n = 265)
Male gender81 (56)61 (57)9 (64)151 (57)
Neonate11 (8)8 (8)4 (29)23 (9)
 Malignancy58 (40)35 (33)7 (50)100 (38)
 Diabetes28 (19)23 (22)3 (21)54 (20)
Location in intensive care unit57 (39)36 (34)11 (79)104 (40)
Surgery in previous 3 months56 (39)61 (58)9 (64)126 (48)
Previous antifungal agents32 (22)17 (16)5 (36)54 (20)
Previous corticosteroids49 (34)32 (30)3 (21)84 (32)
Parenteral hyperalimentation52 (36)62 (58)10 (71)124 (47)
Candida species identification
 C. albicans71 (49)52 (49)4 (29)127 (48)
 C. glabrata13 (9)2 (2)4 (29)19 (7)
 C. tropicalis19 (13)7 (6)1 (4)27 (10)
 C. parapsilosis26 (18)40 (38)3 (21)69 (26)
 C. krusei7 (5)2 (2)2 (14)11 (4)
 Other9 (6)3 (3)0 (0)12 (5)
Multiple CVCs17 (12)10 (9)2 (14)29 (11)
Non-temporary CVC23 (16)13 (12)0 (0)36 (14)
Early CVC48 (33)65 (61)9 (64)122 (46)
Treatment with antifungal agents118 (81)101 (95)11 (79)230 (87)
Disseminated infection6 (4)4 (4)3 (21)13 (5)
High severity of illness score28 (19)13 (12)5 (36)46 (17)
In-hospital death 0–1 days post-candidaemia21 (14)2 (2)1 (7)24 (9)
In-hospital death 2–30 days post-candidaemia53 (37)22 (21)7 (50)82 (31)

At the time of candidaemia, 104 (40%) patients were admitted to an ICU, 80 (32%) were in medical wards, 50 (19%) were in surgical wards, and 13 (5%) were in paediatric wards; 11 (4%) were outpatients. Temporary non-tunnelled CVCs were the most common type of CVC (167 patients; 63%); 36 (14%) were peripherally inserted CVCs, 24 (9%) were infusion ports, and nine (3%) were tunnelled CVCs. In addition, 29 (11%) patients had multiple CVCs.

Candida albicans was isolated most frequently (48%), followed by Candida parapsilosis (26%), Candida tropicalis (10%), Candida glabrata (7%), Candida krusei (4%), and other Candida spp. (5%) (Table 1).

In-hospital mortality occurring on days 0–1 post-candidaemia onset occurred in 24 (9%) patients, and post-candidaemia death in 2–30 days occurred in 82 (31%) patients (Fig. 1). The median time from candidaemia onset to catheter removal was 1 day (range 0–29 days).

Figure 1.

 Days to death from onset of candidaemia for all cases (n = 265).

Outcome study

Of 265 candidaemia patients with a CVC, 172 were included in the outcome study. Among the excluded patients, 24 were excluded because of death on days 0 or 1 post-candidaemia onset, and one was discharged on these days. Additionally, 48 patients were paediatric patients, and 21 patients had multiple CVCs. Among the remaining patients eligible for the outcome study, 37 (21%) CVCs were temporary CVCs inserted in a jugular vein, 80 (46%) were inserted in a subclavian vein, 21 (12%) were femoral, 17 (10%) were peripherally inserted CVCs, five (3%) were tunnelled CVCs, and 13 (8%) were infusion port catheters.

The timing of CVC removal varied among certain patient groups, and demonstrated that the approach to CVC removal was heterogeneous. ICU patients, patients receiving surgery in the 30-day period before infection, and those with a high SOI were significantly more likely to have their catheter removed early, whereas patients with a malignancy or a non-temporary CVC were significantly less likely to have their catheter removed early (Table 2).

Table 2.   Characteristics associated with early or delayed central venous catheter (CVC) removal among outcome study patients (n = 172). Relative risk (RR) values >1 indicate variables associated with early CVC removal, whereas values <1 indicate variables associated with delayed CVC removal
n (%)
n (%)
RR (95% CI)p value
  • a

    Data concerning SOI category were missing for seven patients.

  • ICU, intensive care unit; SOI, severity of illness category.

ICU32 (37)12 (14)1.7 (1.3–2.2)<0.01
High SOIa20 (24)7 (9)1.6 (1.2–2.1)<0.01
Surgery55 (64)30 (35)1.8 (1.3–2.5)<0.01
Malignancy30 (34)53 (62)0.6 (0.4–0.8)<0.01
Non-temporary CVC1 (1)17 (20)0.1 (0.01–0.7)<0.01

Overall, the time to death among cases with early CVC removal, compared to delayed CVC removal, did not differ significantly according to Kaplan–Meier survival analysis (Fig. 2). Kaplan–Meier curves were also created to examine whether early removal was of benefit for patients in ICUs, those with preceding surgery, or those with a concomitant malignancy; however, no difference in time to death was seen among these groups (data not shown).

Figure 2.

 Kaplan–Meier survival curve of the time to death among candidaemic patients with early removal of central venous catheter (CVC) (n = 87, dashed line), compared with patients with delayed CVC removal (n = 85, solid line).

Following univariate analysis, cases with a high SOI were associated significantly with in-hospital mortality (relative risk (RR) 2.4, 95% CI 1.4–4.4), whereas cases with C. parapsilosis isolates (RR 0.7, 95% CI 0.6–0.9) were associated with decreased risk of in-hospital mortality (Table 3). Since the approach to CVC removal in the present cohort was significantly different for different exposure groups, and these exposures were likely to affect the measured outcome, these factors were controlled in the multivariate model. After controlling for ICU admission, malignancy, preceding surgery and the presence of a non-temporary CVC, a high SOI score was associated independently in multivariate analysis with greater odds of mortality (OR 6.1, 95% CI 2.4–15.6, p <0.01), while isolation of C. parapsilosis was associated with lower odds of mortality (OR 0.4, 95% CI 0.1–0.9, p 0.04) (Table 3). Early CVC removal was not associated with significantly decreased odds of in-hospital mortality.

Table 3.   Characteristics associated with mortality during the 2–30-day period post-candidaemia onset according to univariate and multivariate analysis among adults with central venous catheters (n = 173). Multivariate analysis was conducted after controlling for high severity of illness, location in an intensive care unit, concomitant malignancy, preceding surgery, and the presence of a non-temporary central venous catheter
CharacteristicNumber (%) RR (95% CI)p value
(n = 61)
(n = 111)
  • a

    Significant following multivariate analysis; odds ratio for high severity of illness category = 6.1, 95% CI 2.4–15.6, p <0.01; odds ratio for C. parapsilosis isolate = 0.4 (0.1–0.9), p 0.04.

  • RR, relative risk.

High severity of illness categorya19 (33)8 (7)2.4 (1.4–4.4)<0.01
C. parapsilosisa7 (11)31 (28)0.7 (0.6–0.9)0.01
Early removal32 (52)55 (50)1.0 (0.8–1.3)NS
Antifungal treatment54 (89)105 (95)0.7 (0.4–1.3)NS

Stratified univariate analysis was also performed to evaluate whether early CVC removal was of benefit to certain subgroups of patients. Among ICU patients, those with preceding surgery, those with malignancy and those with CVC-related candidaemia, a high SOI score was associated significantly with a higher risk of in-hospital mortality; early CVC removal was not associated with improved outcome (Table 4). The subgroup of CVC-related candidaemia was also analysed by multivariate analysis after controlling for SOI, but no benefit was observed (data not shown).

Table 4.   Characteristics associated with mortality during the 2–30-day period post-candidaemia onset among selected subgroups of patients
CharacteristicDeathSurvivalRR (95% CI)p value
  1. RR, relative risk; ICU, intensive care unit; SOI, severity of illness score; CVC, central venous catheter.

Among ICU patients (n = 44)
 High SOI12 (80)7 (31)2.3 (1.2–4.2)<0.01
 Previous colonisation13 (68)10 (40)1.6 (1.0–2.8)0.08
Among surgical patients (n = 85)
 High SOI11 (46)5 (9)2.5 (1.2–5.3)<0.01
Among patients with malignancy (n = 83)
 High SOI8 (28)2 (4)3.5 (1.0–12.1)<0.01
 Infection with  Candida parapsilosis3 (9)14 (27)0.7 (0.5–0.9)0.09
Among patients with CVC-related candidaemia (n = 75)
 Previous immunosuppression3 (18)25 (43)0.8 (0.6–1.0)0.09
 High SOI5 (29)4 (7)1.8 (0.9–3.8)0.02


Although CVC removal is an important component of treatment for candidaemia, particularly in patients with an infected catheter, it is not clear which patients benefit most from this practice, or whether the timing of removal is critical to improved outcome in all patients. In the present study, data were analysed for a variety of adult candidaemic patients identified through population-based surveillance. The findings revealed that the overall SOI remained the most important predictor for in-hospital mortality, and it was not possible to identify any subgroups of patients who benefited significantly from early CVC removal.

To date, there have been no randomised trials in candidaemic patient populations that have been conducted specifically to address the effects of catheter removal vs. retention. In some studies, the role of catheter removal as a prognostic factor of mortality has been evaluated using observational data [1,6–14,16–22]. Many of these studies have used varying patient populations and approximations of SOI, which are factors that should be considered when comparing the results with those obtained in the present study. Raad et al. [14] attempted to determine the appropriate time for CVC removal, and the sub-population of patients (patients with catheter-related candidaemia, compared with those with candidaemia from a non-catheter source) that would benefit most from such an intervention. Removal of CVCs <72 h after diagnosis of candidaemia in patients with catheter-related infection was found to result in an improved response to antifungal therapy, defined as resolution of all clinical manifestations of the candidaemia and bloodstream sterilisation [14]. However, this previous study was limited to patients with cancer (49% with haematological malignancies and 51% with solid tumours), did not adjust for SOI, and did not exclude patients who died or were discharged before the defined day of early catheter removal [14].

It is likely that catheter removal, particularly in patients with infected catheters, is beneficial for the treatment of candidaemia [5]. Unexpectedly, the present study revealed that early catheter removal was not beneficial, even in patients with catheter candidaemia, and even after controlling for SOI, which in multivariate analysis proved to be the most predictive marker of in-hospital mortality. Moreover, catheter removal is not always possible without submitting patients to greater risk, especially in clinical settings involving cancer or haemodialysis, because of difficulty with venous access or co-morbidities such as thrombocytopenia. Thus, in clinical practice, there is a proportion of candidaemic cases that is treated with catheter retention, despite long-standing recommendations for immediate removal of all CVCs [1,4,5,8,9,13,14,16–19]. The present findings emphasise that each candidaemic patient must be assessed individually, and that the benefits and risks of early catheter removal must be considered carefully when deciding on the management of such patients.

One of the limitations of using observational data is the inability to control for all confounding variables, including those that are known, e.g., SOI, and those that are unknown, e.g., the tendency to remove CVCs earlier for some subgroups of patients. Although some previous studies have shown a benefit in removing CVCs [1,7–14,16–21], other studies have shown that patients whose CVC was retained had a significantly higher SOI score [1,9,16]. The present study attempted to understand some of these confounding issues by investigating which factors in the cohort were associated with early or delayed removal, and found that different approaches to CVC removal were evident. Patients in ICUs, and those with preceding surgery, were more likely to have early removal of a CVC, whereas patients with malignancy and those with a non-temporary CVC, were not. Interestingly, patients with a high SOI score were more likely to be in the early CVC removal group, possibly accounting for the absence of any mortality benefit for early removal in univariate analysis. However, even after these confounding issues were controlled by including them in the multivariate model, no benefit of early CVC removal was detected. It was also impossible to identify any subgroups of patients for whom early CVC removal was beneficial. Intuitively, a benefit was expected among patients with CVC-related candidaemia. However, even after controlling for SOI, early removal remained non-significant.

Observational studies that attempt, either explicitly or implicitly, to assess the cause and effect relationship between catheter removal and outcome, e.g., mortality or improved response, involve a difficult task. Factors that are associated with outcome, e.g., underlying disease, location in an ICU and SOI, can also be factors involved in a clinician's decision whether to remove the catheter. These differences in background characteristics between the treatment groups (early removal vs. delayed removal) indicate possible differences in baseline risk for the outcome, and can thus lead to biased or even incorrect conclusions. Appropriate methods and sample sizes, adequate to address the heterogeneity with respect to baseline risk, are necessary in order to explore the question of whether catheter removal is beneficial for certain sub-populations of patients.

In conclusion, it was not possible to demonstrate that early removal of CVCs in patients with candidaemia, as proposed by current guidelines [4,5], was associated with improved outcome in all patients. Patients are likely to benefit most from a careful, individualised analysis of the risks and benefits of the timing of catheter removal. Further research should focus on determining the subgroups that benefit most from this practice, and should take into account the complicated biases and confounders in these analyses.


This study was presented, in part, at the 45th Interscience Conference on Antimicrobial Agents and Chemotherapy (Washington, DC, 2005), and was supported by research grants from Pfizer, Inc. and Gilead Sciences. DR, BJP and BA have full access to all the data contained in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. D. Rodriguez received a research grant from Ministerio de Sanidad y Consumo, Instituto de Salud Carlos III, Spanish Network for the Research in Infectious Diseases (REIPI RD 06/0008).