Femoral vs Jugular Venous Catheterization and Risk of Nosocomial Events in Adults Requiring Acute Renal Replacement Therapy:  A Randomized Controlled Trial FREE

Femoral, jugular, and subclavian venous catheterizations are routinely performed during critically ill patient care. These invasive procedures contribute to additional morbidity, mortality, and costs derived from the interactions between mechanical, infectious, and thrombotic complications.^1- 2 Femoral venous catheterization, which is rapid to perform, is considered an emergency procedure to gain vascular access, but which should be avoided to limit nosocomial complications.^1,3- 7 The subclavian site, although often a first choice,⁸ is less suitable for larger catheters, such as noncuffed temporary dialysis catheters. The US Centers for Disease Control and Prevention states: “Place catheters used for hemodialysis in a jugular or femoral vein rather than a subclavian vein to avoid venous stenosis if catheter access is needed.”⁸

Some studies reported a higher incidence of complications associated with femoral vs jugular catheterizations,^9- 12 while other studies reported lower or similar incidences with femoral catheterizations^13- 16; however, all of these studies were observational. Consequently, the choice between femoral and jugular sites remains a subject of debate and is dictated by preference rather than evidence-based decision in the setting of severely ill, bed-bound patients. Most studies on acute dialysis catheters are performed in ambulatory patients with end-stage renal disease starting chronic dialysis.¹⁷ In this context, the National Foundation of Kidney Disease Outcome and Quality Initiative guideline¹⁸ suggests noncuffed, nontunneled dialysis catheter not to exceed 3 weeks for jugular and 5 days for femoral accesses, because the risk of catheter-related bloodstream infection increases after these points. Patients with chronic dialysis, however, differ in many respects from critically ill patients, raising the question of whether these recommendations can be extrapolated to critically ill patients.

Because a higher load of microorganisms cultured from the catheter tip on removal are predictive of catheter-related bloodstream infection,^19- 20 we compared the rates of catheter colonization on removal (primary end point) between internal jugular and femoral catheterization among patients in the intensive care unit (ICU) requiring renal replacement therapy (RRT). We hypothesized that using the jugular site would decrease the rate of catheter colonization on removal and, therefore, catheter-related bloodstream infection (secondary end point) compared with the femoral site catheterization. This assumption is based on the fact that catheter-tip colonization on removal and catheter-related bloodstream infection are both highly correlated clinically and statistically (r = 0.69, P = .001).²¹ Other end points included insertion complication rates and thrombotic events.

The Cathedia Study was a concealed, randomized, multicenter, prospective, evaluator-blinded, parallel-group trial comparing femoral and internal jugular access for RRT in critically ill patients within a network of 9 tertiary care university medical centers and 3 general hospitals in France between May 2004 and May 2007. The study was approved by the institutional review board at the Côte de Nacre University Hospital, Caen, France. Informed written consent was obtained from all participants or their proxies in cases of impaired decision-making capacity at the time of enrollment.

Eligible patients were critically ill adults (≥18 years) who were expected to require support with RRT. Only patients who were undergoing their first venous catheterization for acute RRT and without contraindications to attempt both jugular and femoral access were considered. Consequently, patients with coagulopathy (definition left at the operator's discretion), morbid obesity with a body mass index (BMI, calculated as weight in kilograms divided by height in meters squared) of more than 45, local skin infection, profound volume overload that precluded Trendelenburg positioning, chronic renal failure with arteriovenous fistula, thoracic life-threatening condition, and patients with only 1 site available (femoral or jugular) were not included. Eligible operators were trained physicians with at least 50 successful catheter insertions in both sites. We chose this level of operator skill experience because it limited the risk of mechanical complications.²

Patients fulfilling the inclusion criteria were randomized to 1 of 2 treatment groups, stratified by center and type of RRT, before catheterization. Allocation concealment was obtained by a centralized 24-hour Internet or telephone service (EOL; MedSharing, Bondy, France), involving a dynamic semideterminist computed-generated algorithm, to ensure complete separation of the randomization process from those physicians providing care.

Insertions took place in the ICU. Surgical hand-scrubbing or hand-rubbing was performed by the operator before putting on sterile gloves.²² Operators wore sterile surgical long-sleeved gowns, caps, and mask. Skin disinfection and catheter care used the same alcohol-based povidone-iodine disinfectant, as previously described.²³ Large sterile drapes were placed over the insertion site, which was disinfected again by the operator. Physicians inserted all catheters by using the Seldinger technique. Vein puncture was achieved by using anatomical landmarks on the skin's surface. Ultrasound guidance and right-side positioning were recommended for jugular insertions, but these were left to the investigators' discretion. In cases of unsuccessful insertions, physicians were instructed to switch to the contralateral side, if possible, and then switch from one site to the other site. Catheter maintenance was performed according to local protocol and was similar between groups in each center. Catheters were not used for routine blood sampling or to administer drugs. None of the studied catheters was tunnelized or antibiotic-impregnated. One center used antiseptics-impregnated catheters (ARROWg + ard Blue; Arrow International Inc, Reading, Pennsylvania). Unfractionated heparin or low-molecular-weight heparin was used for anticoagulation, when indicated. During the interdialytic period, catheters were filled with heparinized saline or saline fluids only, according to each center protocol. No alcoholic or antibiotic locks were used.

Decisions to remove catheters were made independently by the physicians in charge of each patient when the catheters were no longer needed (renal function recovery or death) or new access was required (suspicion of catheter-related infection, catheter dysfunction, or thrombosis). Patients were followed up until death or ICU discharge.

Catheter Insertion and Mechanical Complications. Time required for catheter insertion, number of skin punctures, number of failures, and occurrence of mechanical complications were recorded from insertion to removal.

Catheter-Related Infections. According to the simplified Brun-Buisson quantitative technique,²⁰ catheter colonization was defined as cultures with at least 10³ colony-forming units (CFUs) per milliliter from the catheter tip. The microbiologist technicians in charge of the bacterial counting required to define catheter-tip colonization were blinded to the study group.

A blood sample was drawn from a peripheral vein at the time of the removal if there was any suspicion of catheter-related infection. As previously described,² catheter-related bloodstream infection was defined as catheter-tip colonization plus at least 1 peripheral blood culture yielding the same species with the same antimicrobial susceptibility as the catheter tip within 48 hours of catheter removal, with no other apparent source of sepsis. Two different peripheral blood cultures were used to define catheter-related bloodstream infection with potential skin contaminants.

Catheter-Related Thrombosis. The presence of thrombotic complications (partial or complete) of the vein was assessed systematically by ultrasonography performed within 4 days of catheter removal in 2 participating centers only, because a smaller sample size was desired to demonstrate the absolute 25% difference in deep venous thrombosis expected between the upper and lower extremity sites.²⁴ These 2 centers were able to perform an onsite permanent ultrasonography.

Estimation of the study sample size was based on the expected difference in the time to colonization on removal of jugular vs femoral catheters (primary end point). With 650 patients, the study had 90% power to detect a treatment difference at a 2-sided 5% significance level, if the true hazard ratio (HR) is 0.77. This is based on the assumption that the true median times to colonization on removal for the femoral and jugular sites are 14.0 and 18.2 days or more, respectively.²³ Assuming that 15% of the catheters would not be inserted, removed outside the ICU, or not sent to culture, we planned to include 750 patients.

The statistical unit was the patient because only the first inserted catheter was analyzed, in keeping with the independence of observations criteria. Patients were analyzed according to a modified intention-to-treat strategy, in which catheters with missing data for the primary end point were excluded. In addition, we complemented this analysis with a per-protocol strategy, in which patients with crossover (insertion failure to the allocated group) were analyzed in the group the catheter was inserted rather than randomized.

Rates of catheter insertion failure and mechanical complications were compared between study groups by Fisher exact tests. Comparisons of the incidences of colonization on removal, catheter-related bloodstream infections, and thrombosis per 1000 catheter-days between groups were analyzed by Poisson regression, in which the incidence was used as the numerator and the number of catheter-days for patients was used as the denominator to form a ratio. Exact Poisson 95% confidence intervals (CIs) were computed.

The proportions of catheters that were free of colonization on removal (primary end point) were compared between groups using time-to-event methods with log-rank tests. After checking for the proportionality assumption, we fitted a Cox proportional hazards regression model to estimate HRs, with values of less than 1 favoring the jugular over the femoral group. Subgroups analysis for the primary end point according to the catheter insertion time (≤5 days vs >5 days), BMI (cutoffs according to the population tercile), sex, and type of RRT were prespecified by the study group. The a priori hypothesis was that the effect of jugular site catheterization in preventing colonization was stronger in patients with longer catheterization times^11,18 and higher BMIs.²⁵ Interactions were assessed in a Cox proportional hazards regression model that included the group, the subgroup variable, and the interaction between group and subgroup variables. In case of significant interaction with a continuous factor and because the use of tercile is very dependent of the sample studied, we explored the relationship between the continuous variable and the occurrence of the primary end point by plotting a logistic model curve, separately for jugular and femoral catheters. We also computed multivariate Cox proportional hazard regression models including the group and adjusting for potential confounding variables significantly associated with primary end point. Overall and each subgroup adjusted HRs were similar to unadjusted HRs and thus were not reported.

All tests were 2-sided; P < .05 denoted statistical significance. In cases of multiple testing regarding the primary end point, Bonferroni adjustment was applied, resulting in a significance level of P < .05 divided by 2 (P < .025) or 3 (P < .017), depending on the number of subgroups. We used SAS version 9.1 (SAS Institute Inc, Cary, North Carolina) for statistical analysis.

Figure 1 shows the patient inclusion process for the Cathedia Study. Characteristics of the patients and catheters, including duration of catheterization, were similar in both groups (Table 1).

Jugular catheters were more difficult to insert than femoral catheters and required longer insertion times, had more failures on 1 side, and more crossovers (Table 1). Rates of arterial puncture between jugular and femoral groups did not differ significantly (19/366 patients [5.1%] vs 13/370 patients [3.6%], respectively), although the rate of hematoma formation was significantly higher for the jugular group (13/366 patients [3.6%]) compared with the femoral group (4/370 patients [1.1%]; P = .03).

In the jugular group, 2 patients had severe respiratory distress due to compressive hematoma and required intubation. Another patient in the jugular group, admitted for cardiogenic shock following cardiac arrest, experienced catheter insertion in the carotid artery and required vascular surgery for removal. One patient in the femoral group, admitted for septic shock, had arterial occlusion with acute leg ischemia and required amputation. Surgical examination revealed an arteriovenous fistula related to catheter insertion in the femoral vein, which occluded the femoral artery. None of these 4 patients had ultrasound guidance for catheter insertion.

The results of catheter-related infections are shown in Table 2. The rate of randomized participants with missing data and not included in this analysis did not differ significantly between jugular and femoral groups (62/375 patients [16.5%] vs 51/375 patients [13.6%]; P = .31). Figure 2 shows a Kaplan-Meier curve of overall catheters free of colonization on removal. Colonization occurred in 84 of 324 patients (25.9%) with femoral catheters (incidence of 40.8 per 1000 catheter-days) and in 78 of 313 patients (24.9%) with jugular catheters (incidence of 35.7 per 1000 catheter-days). The corresponding HRs were 0.85 (95% CI, 0.62-1.16; P = .31) in the modified intention-to-treat analysis and 0.83 (95% CI, 0.61-1.13; P = .24) in the per-protocol analysis.

In cases of colonization (Table 2), significantly more Gram-positive bacteria were found in the jugular site (P = .04) and significantly more Gram-negative bacteria (P = .03) were found in the femoral site. The mean (standard deviation) load of microorganisms cultured from the catheter tip did not differ significantly between jugular and femoral catheter-tip culture (3.59 [0.93] vs 3.77 [0.90] for log₁₀ CFU/mL; P = .20 by t test).

Catheter-related bloodstream infection occurred in 3 of 324 patients (0.9%) with femoral catheters (incidence per 1000 catheter-days, 1.5; 95% CI, 0.1-6.4) and in 5 of 313 patients (1.6%) with jugular catheters (incidence per 1000 catheter-days, 2.3; 95% CI, 0.3-7.7). This difference was not significant by Poisson regression (P = .42).

Overall, 2 of 370 patients (0.5%) in the femoral group and 2 of 366 patients (0.5%) in the jugular group had symptomatic deep venous thrombosis. In 2 participating centers, systematic ultrasound evaluation of thrombotic events was performed. The mean duration of catheterization with ultrasound evaluation was 6.2 days (95% CI, 5.3-7.1) and 7.3 days (95% CI, 6.2-8.4) in the femoral and jugular groups, respectively (P = .15). Rates of thrombosis were found in 8 of 76 patients (10.5%) in the femoral site vs 17 of 75 patients (22.7%) in the jugular site. This difference was not significant by Poisson regression (P = .16).

The results of the predefined subgroup analyses are shown in Table 3. A significant interaction (P < .001) was found for the effect of BMI on the relationship between catheter site and colonization-free catheter on removal survival, suggesting risk stratification. Patients with a lower BMI had a higher incidence of colonization in the jugular vs femoral group (45.4 vs 23.7 per 1000 catheter-days, respectively), and those patients with a higher BMI had a significantly lower incidence of colonization in the jugular vs femoral group (24.5 vs 50.9 per 1000 catheter-days, respectively). The corresponding Kaplan-Meier curves are shown in Figure 3. The effect of BMI on the a priori (ie, when the catheter is inserted) probability of catheter colonization on removal is shown in Figure 4.

In this large, multicenter randomized trial, we did not detect any clinically relevant benefit of the jugular site catheterization compared with femoral site catheterization for reducing the risk of nosocomial complications in critically ill adults requiring venous access for acute RRT. This result is inconsistent with the widely accepted convention to avoid femoral catheterization to prevent the risk of catheter-related infection. Interestingly, in the subgroup analysis, there was qualitative heterogeneity of the effect of insertion site on the risk of catheter colonization on removal according to BMI. This probability regularly increased with the BMI for the femoral group, but not for the jugular group.

To our knowledge, our study is one of the largest randomized trials of the prevention of catheter complications in adult patients in the ICU. Consequently, we consider that the absence of an overall statistical effect in our population precludes any clinically important effect size.

Random allocation avoided channeling bias commonly associated with observational studies. Patient factors or therapeutic interventions that could increase or decrease the risk of catheter colonization were prospectively monitored and were similar between groups.

Assessment of the catheter-tip colonization on removal was evaluator-blinded and objective. Catheter colonization, however, is a dynamic process and it is unlikely to occur exactly at the time of removal. A bias would occur if the delay between intravascular colonization and colonization on catheter removal differed between jugular and femoral groups. In this case, it is likely that one catheter group would have been more heavily colonized than the other and consequently more prone to lead to bacteremia, which is not what we observed.

Our study involved severely ill, bed-bound patients who required RRT with poor outcomes and high-illness acuity.²⁶ Consequently, our primary end point results may not be applicable to ambulatory patients with end-stage renal disease admitted to hemodialysis units. In addition, data concerning infectious complications of hemodialysis catheters may not necessarily be applied to catheters used to administer drugs. Conversely, the epidemiological characteristics of colonization were similar between temporary catheters used for hemodialysis and those used for drug administration in the intensive care setting.^27- 29 In addition, patients with acute renal failure represent a subset of critical patients with higher risk of nosocomial bacteremia.³⁰ Consequently, we believe the findings can be extended to the larger population of critically ill patients in general.

The catheter duration observed in our study are consistent with those results reported in other randomized studies that demonstrated significant reductions in catheter colonization^31- 32 and with those results of other studies that investigated temporary dialysis catheters.^28- 29 However, our incidence of colonization is higher than previously reported.^28- 29 Several factors inherent to the study population or study design may explain these discrepancies. For example, Harb et al²⁸ reported a low colonization incidence of 5.4 per 1000 catheter-days for the 79 dialysis catheters inserted in 47 critically ill patients with cancer. A total of 82% were inserted at the femoral site. Although not reported, it is likely that this population had a very low BMI, which would correspond to a low risk of femoral colonization (Figure 4). Souweine et al²⁷ excluded dialysis catheters inserted for less than 48 hours, while we found a high incidence of colonization among catheters inserted for 5 days or less (Table 3), in accordance with 1 prospective study of catheter replacement every 5 days in consecutive patients in the ICU who were treated by hemodiafiltration.³³ In addition, the incidence of catheters colonized from consecutive patients included by this center²⁹ in our study was homogeneous with other centers.

The predominance of Gram-negative bacteria and fungi cultured from femoral catheters was also described for catheter-related bloodstream infection,³⁴ confirming the extraluminal pathogenesis of short-term, catheter-related infection³⁵ and the validity of using colonization as a surrogate for catheter-related bloodstream infection.²¹

Despite our relatively large sample size, we did not reach statistical significance for detecting a higher incidence of colonization between the 2 groups, except among patients stratified according to BMI (with opposite effects). Inadequate dressings were previously associated with increased BMI in a study that included 66% femoral catheters.²⁵ In contrast, BMI does not seem to influence dressing integrity in jugular catheters.

The cumulative hazards of colonization on removal in femoral and jugular catheters are quite linear with an instantaneous risk relatively constant over time (Figure 2). This result suggests that catheter colonization and subsequent infection is a random event and that a threshold duration at which the probability of catheter-related bloodstream infection sharply increases does not exist within the range of catheter durations examined in our study. The implication for clinical practice is to avoid systematic catheter change after a predetermined length of time in ICUs, in accordance with current Centers for Disease Control and Prevention guideline.⁸ However, the National Foundation of Kidney Disease Outcome and Quality Initiative recommendation¹⁸ to change femoral acute dialysis catheters every 5 days to prevent infectious complications does not appear appropriate in critically ill adults.

The rate of mechanical complications in the jugular group is similar to previously reported rates of complication.^2,36 In contrast, the low rate of mechanical complications in the femoral group is noteworthy and could be explained by several factors. First, the operators were similarly experienced for both sites. Second, the femoral route was selected by chance, not by choice. Deshpande et al¹⁵ found that the reasons for selecting a site for central catheterization differed between jugular and femoral regarding operator preference (higher for jugular) and patients risk factor (bleeding, emergency, and agitation, all higher for femoral). Third, patients were excluded from our study if the operators had no choice (eg, coagulopathy making jugular site hazardous). In those situations, the risk of complications might be higher in femoral as well.

Three randomized studies have already compared the risk of thrombosis according to site of insertion.^24,37- 38 A significantly higher risk in the femoral group was found by Trottier et al²⁴ (25% for femoral vs 0% for jugular and subclavian sites) and by Merrer et al³⁷ (21% for femoral vs 2% for subclavian site). It is difficult to compare the results of these 2 studies with our study because within upper extremity sites, rates of thrombosis may be higher in jugular site than in subclavian site (42% vs 10%, respectively).³⁹ In addition, all of our patients received routine anticoagulation for RRT, which might have contributed to lower rates of thrombosis in both groups.³⁹ Karakitsos et al³⁸ compared the “low-approach” vs “standard approach” for femoral catheterization in patients in the ICU requiring RRT. Their rate of deep venous thrombosis using the standard approach (4/40 patients [10.0%]) is very similar to our estimate in the femoral group (8/76 patients [10.5%]), increasing the external validity of our finding.

Subgroup analysis should always be conducted and interpreted with extreme caution. However, several factors inherent to our study need to be acknowledged. First, a formal interaction test identified a highly statistically significant heterogeneity in our overall population.⁴⁰ Second, the occurrence of a qualitative interaction with opposite effects is unlikely to occur by chance.⁴⁰ Third, although subgroup analysis generally implies lower statistical power, the tests comparing femoral vs jugular groups reached statistical significance in one subgroup and was significant in the other subgroup. Finally, the differential effect is anatomically plausible, at least for the higher BMI subgroup.²⁵ However, the unexpected protective effect of femoral vs jugular in the lower BMI subgroup probably needs to be interpreted as hypothesis-generating rather than hypothesis-supporting until replicated in a different population. A second possible limitation is the absence of systematic ultrasound guidance during catheter insertion in the jugular group. The mechanical complication rate reported for the jugular group might be overestimated because ultrasound guidance decreases this risk significantly.^2,36 Conversely, the use of ultrasound guidance is not currently routine practice. Third, catheter survival was rather short, with only 9% of the catheters inserted for more that 15 days. The conclusions can therefore not be extrapolated to dialysis catheters with longer stay. Similarly, the dialysis catheters were inserted early in the ICU stay and the conclusions can therefore not be extrapolated to catheters that are inserted at a later stage of ICU stay.

In conclusion, the decision for the best site of insertion to prevent complications might be more complex than previously suggested.¹² Our results support the current guideline for preventing catheter complications regarding the optimal site for catheter insertion in the ICU.⁸ If a subclavian approach is not available^37,41 and the a priori individual risk of complications between the jugular and femoral sites is equal, the jugular site should be strongly considered for patients with higher BMI. We suggest that first-choice careful femoral catheterization by an experienced operator with full sterile precautions and appropriate postinsertion site care in nonobese, bed-bound, severely ill patients is acceptable and could reduce catheter-related morbidity compared with jugular catheterization.