Staphylococcus aureus Bacteremia and Buttonhole Cannulation: Long-Term Safety and Efficacy of Mupirocin Prophylaxis

Abstract

Background and objectives: Buttonhole (constant-site) cannulation (BHC) continues to gain popularity with home and in-center dialysis programs worldwide. However, long-term safety data are lacking. This paper reports the authors' single-center experience with Staphylococcus aureus bacteremia (SAB) and the efficacy of topical mupirocin prophylaxis (MP).

Design, setting, participants, & measurements: This study was a retrospective prepost comparison of SAB rates after establishing MP. Fifty-six consecutive patients on home nocturnal hemodialysis via arteriovenous fistulae, mean age 51.5 ± 10.6 years, 38% women, and vintage 44.5 ± 34.5 months were observed for a total of 93.4 (pre-MP) and 193.5 (post-MP) patient-years.

Results: Ten episodes of SAB were observed, with metastatic complications in four cases, including pneumonia (n = 2), septic arthritis, and a fatal C3 epidural abscess. When analyzed by observation period, the odds ratio (OR) for SAB before versus after the introduction of MP was 6.4 [95% confidence interval (CI) = 1.3 to 32.3; P = 0.02]. Two SAB episodes occurred after the MP started. Both patients had discontinued the MP for 3 weeks (nonadherent) preceding infection; hence, no SAB episodes were observed on treatment. In an as-treated analysis, the OR for SAB in the absence of MP was 35.3 (95% CI = 2.0 to 626.7; P = 0.01).

Conclusions: BHC is associated with a significant risk of SAB with metastatic complications. In this prepost comparison of SAB rates, no infections were observed with MP. While awaiting more definitive studies, this simple intervention should be considered for patients using BHC.

Arteriovenous fistulae (AVF) are associated with better patient survival and lower infection risk when compared with central venous catheters (CVCs) and synthetic grafts; hence, they remain the preferred vascular access for patients receiving hemodialysis (HD) (1–6). The standard “rope-ladder,” or rotating-site technique, is the most commonly used method of cannulation with AVF. With this method, a new pair of venipuncture sites is used for each successive cannulation, thus allowing for skin healing to occur between dialysis sessions. The constant-site, or “buttonhole” cannulation (BHC) technique, has been in use since the 1970s (7,8) and has become increasingly popular among patients who self-cannulate (9), such as those receiving home HD. With the BHC technique, a limited number of puncture sites (typically three) are repeatedly cannulated, resulting in scarring with tract formation.

Although published reports to date suggest that BHC is safe (9,10), long-term safety data are lacking. Vascular access-related bacteremia remains a major cause of morbidity and mortality among patients receiving HD. Although the incidence of bacteremia is relatively low among patients using AVF, several factors associated with BHC may increase the risk of infection with this technique. These include frequency of cannulation, longer dialysis time (as with nocturnal HD), self-cannulation in an unsupervised environment (home HD), and loss of skin integrity.

In this report, we describe our center's experience with systemic infection rates among patients receiving home nocturnal HD via AVF, using BHC, and evaluate the efficacy of topical mupirocin in reducing rates of bacteremia with Staphylococcus aureus.

Materials and Methods

Study Design and Setting

We undertook a retrospective prepost comparison of the frequency of bacteremic episodes related to BHC of native AVF in patients receiving home nocturnal HD at Humber River Regional Hospital, Canada, between January 1, 1998 and June 8, 2009. The primary outcome was the odds ratio (OR) for developing S. aureus bacteremia (SAB), comparing the periods before and after the implementation of a program-wide antimicrobial cream application procedure that began on January 1, 2004 after several index cases with SAB. ORs comparing infection rates by treatment period and by actual treatment received were estimated. Infection rates among patients using the standard rope-ladder cannulation technique on conventional HD (CHD) at our full-care dialysis center were also reported for comparison. Exploratory analyses examined patient characteristics associated with an increased risk of SAB with BHC.

Study Population

Patients were included in this study if they were using a brachial or radial AVF, with self-cannulation, and BHC for home nocturnal HD. All consecutive patients training for home HD with AVF were offered the opportunity to learn the BHC technique and were not selected on the basis of cannulation difficulties such as aneurismal dilation, short usable segments, or cannulation pain.

Training, Cannulation Procedure, and Intervention

Patients learned the BHC technique during nocturnal HD training. Buttonholes were created by any one of five experienced nurses. Sharp needles were used for the first week, followed by “blunt” needles (Medisystems, Lawrence, MA) thereafter. Before cannulation, the fistula and surrounding skin were cleansed with chlorhexidine gluconate (0.5% in 70% alcohol) or providone-iodine solution. Scabs were covered with an alcohol pad for 5 minutes and then removed using a sterile needle. Mupirocin calcium 2% cream (Bactroban, Glaxo-Smith-Klein) was applied to each buttonhole with a sterile cotton swab after hemostasis was achieved and left to air dry. A single application was provided for each HD procedure. No dressings were applied. No patients received nasal antimicrobial prophylaxis or required methicillin-resistant S. aureus (MRSA) decolonization therapy. Sterile cannulation technique and adherence to the mupirocin cream was reviewed at each clinic visit (at 8-week intervals), but no specific techniques for verifying adherence were used.

HD Prescriptions

Home nocturnal HD patients received dialysis on alternate (every other) nights or 6 nights/wk for 6 to 10 hours per session using Fresenius 2008H or 2008K home HD machines. Prescription details are provided elsewhere (11,12). In-center patients received CHD for 3.5 to 4.5 h/session thrice weekly with conventional dialysate composition using Fresenius Optiflux 160-200 NR or Polyflux (polyamide) dialyzers (Gambro Inc., Hechingen, Germany). The rope-ladder cannulation technique was used for all patients receiving CHD.

Data Collection and Outcomes

All blood culture results for the study cohort were extracted electronically from the hospital database and imported into a spreadsheet for analysis. Microbiology results from external laboratories were forwarded to our facility and collected prospectively. Patients presenting to an external facility with suspected infection were evaluated by the attending physician at our home dialysis unit within 24 hours of discharge and received instructions for home self-administration of antibiotics and further investigation if warranted. Microbiology laboratory staff were not aware of the study or intervention.

Follow-up times were calculated starting from the date of first cannulation using BHC while receiving nocturnal home HD and ending at the time of any of the following exit events: transplantation, death, change of dialytic modality, access failure, or termination of the BHC technique for technical or other reasons.

Statistical Methods

Patients contributed variable lengths of follow-up time to either or both treatment periods. To use all available follow-up data, univariate logistic regression was used to calculate ORs for infection, comparing infection risk in the pre- and postintervention time periods. Computations were performed using PROC LOGISTIC in SAS (SAS statistical software, SAS Institute, Inc., Cary, NC). In the event of zero cells, 0.5 was added to each cell and ORs were calculated manually. Correlation within patients was not accounted for in the analyses; therefore, standard errors may have been underestimated, leading to an increased likelihood of rejecting the null hypothesis. Two-sided P values were considered significant at the 0.05 level, with 95% confidence intervals (CIs) reported.

Results

Study Subjects

One hundred and thirty patients treated with home nocturnal HD were identified; 56 (43%) had AVF, 16 (12%) synthetic grafts, and 58 (45%) used CVCs. All patients with AVF used BHC and were included in the study. Baseline characteristics are shown in Table 1. All AVF were created at least 6 months before starting BHC. A total of six patients (10%) died (Table 2). Only one death was attributed to vascular-access-related infection. Other exit events included transplantation (n = 8) and move to a new dialysis program (n = 5). No AVF failures were documented during the entire study period.

Table 1.

Patient characteristics at first BHC use for home nocturnal HD (n = 56)

Mean age, years	51.5 ± 10.6
Female gender, n (%)	21 (38)
Race, n (%)
Caucasian	39 (70)
black	5 (9)
Asian	12 (21)
Cause of ESRD, n (%)
diabetes	12 (21)
GN	30 (54)
nephrosclerosis	6 (11)
polycystic kidney disease	4 (7)
obstructive uropathy	3 (5)
interstitial nephritis	1 (2)
Mean time on nocturnal HD, months ± SD	44.5 ± 34.5
HD frequency 6 sessions/wk, n (%)	31 (55)
HD frequency 3.5 sessions/wk (alternate-night), n (%)	25 (45)
Laboratory parameters (predialysis, mid-week), mean ± SD
hemoglobin (g/L)	11.5 ± 1.4
creatinine (mg/dl)	8.3 ± 3.0
urea (mmol/L)	43 ± 19
albumin (g/dl)	3.8 ± 0.5
calcium (mg/dl)	9.5 ± 0.7
phosphorous (mg/dl)	4.2 ± 1.1
ferritin (μg/L)	286 ± 151
C-reactive protein (mg/L)	5.9 ± 9.5
Peripheral vascular disease, n (%)	6 (11)
Current smoking, n (%)	3 (5)
Receiving immunosuppressive therapy, n (%)	3 (5)

Table 2.

Number of exit events and changes in dialysis frequency during the observation period (n = 56 subjects)

Nonaccess-related death	5
cardiovascular	4
ischemic bowel	1
Vascular access-related death (SAB with epidural abscess, quadriplegia)	1
Transplant	8
Move to new dialysis program	5
Change of modality	0
AVF failure	0
Switch from 3.5 sessions/wk to 6 sessions/wka	8
Switch from 6 sessions/wk to 3.5 sessions/wk	0

^aAlternate-night HD began in March 2004, after implementation of the antimicrobial cream prophylaxis protocol.

SAB

A total of 11 blood cultures from 10 patients were positive for S. aureus during the study period (Table 3), all in association with local features of infection at the AVF. One was thought to represent a recurrent infection (see below). Six patients presented with features of local infection and fever. All six patients' symptoms were resolved with a 2-week course of treatment with cefazolin. Four patients with SAB had evidence of metastatic infection. One patient had a cervical abscess compressing C3 to C7 that resulted in quadriplegia. This patient withdrew from HD and died. Two additional patients presented with local infection and clinical and radiographic features of pneumonia. Ventilation-perfusion scanning yielded a high probability of pulmonary embolism in both patients, suggesting septic pulmonary emboli from the infected AVF. Both patients made an uneventful recovery with a 2-week course of cefazolin with HD. One of these patients subsequently suffered a relapse of bacteremia 21 days after the initial infection, which resolved with a second identical course of antibiotic therapy. Endocarditis was excluded by transesophageal echocardiography. One patient presented with septic arthritis (left knee), as confirmed by joint aspirate, and required an extended course (6 weeks) of treatment with cefazolin. Oxacillin sensitivity was demonstrated for all episodes of SAB in the pre- and postintervention periods. One of the uncomplicated bacteremic episodes and the episode of recurrent bacteremia after pneumonia occurred after cream prophylaxis started. In both cases, the patients admitted to noncompliance with the cream for at least a 3-week duration before their infections.

Table 3.

Summary of all positive blood cultures (n = 22) obtained from 56 study patients during the observation period

Causative Organism	Number of Patients with Positive Culture (number of patients with one recurrent positive blood culture)
Staphylococcus aureus	10 (1)
Coagulase negative Staphylococcus	5 (3)
Enterococcus faecaelis	1
Enterobacter cloacae	1
Hemolytic Streptococcus	1

Main Study Outcomes

A total of 286.9 patient-years of follow-up were obtained. Using the intervention start date of January 1, 2004, this yielded 93.4 and 193.5 patient-years before and after the introduction of the antimicrobial cream, respectively (Table 4). Before the introduction of mupirocin cream, a total of eight SAB episodes were documented. When analyzed by treatment period (regardless of treatment received), the infection rate was 0.32 infections/1000 AVF-days before the use of antimicrobial cream prophylaxis and 0.03 infections/1000 AVF-days after January 1, 2004. The OR for developing SAB before the introduction of mupirocin prophylaxis was 6.4 (95% CI = 1.3 to 32.3; P = 0.02). In the “as-treated” analysis, there were no infections in the intervention group, with an estimated OR of 35.3 (95% CI = 2.0 to 626.7, P = 0.01).

Table 4.

Follow-up times and SAB episodes before and after the introduction of the mupirocin cream prophylaxis start date (January 1, 2004)

	By Treatment Perioda	As-Treatedb
Preintervention
follow-up time for entire cohort (years)	93.4	98.1
mean ± SD follow-up time in years per subject (range)	2.7 ± 1.7 (0.2 to 5.7)	2.7 ± 1.7 (0.2 to 5.7)
infection episodes	8	10
infection rate (events/1000 AVF-days)	0.23	0.28
Postintervention
follow-up time for entire cohort (years)	193.5	188.6
mean ± SD follow-up time in years per subject (range)	4.3 ± 1.9 (0.4 to 5.4)	4.3 ± 1.9 (0.4 to 5.4)
infection episodes	2	0
infection rate (events/1000 AVF-days)	0.03	0
OR (95% CI)c	6.4 (1.3, 32.3)	35.3 (2.0, 626.7)
CHD controls
follow-up time for entire cohort (years)	–	565.9
infection episodes	–	1
infection rate (events/1000 AVF-days)	–	0.005

^aFollow-up time on treatment calculated from January 1, 2004; infections attributed to study period.

^bFollow-up time on treatment calculated from actual intervention start date for subjects initially not adherent to study intervention (n = 2); SAB episodes attributed to intervention status—“as-treated” analysis.

^cOR as calculated by univariate logistic regression.

Bacteremia in Conventional In-Center HD Using the Conventional Rope-Ladder Cannulation Technique

A total of 298 patients with native AVF received CHD at Humber River Regional Hospital between January 1, 2002 and October 30, 2006 for a total of 206,584 AVF-days (Table 4). In this cohort, only one episode of SAB was documented in association with local features of AVF infection, corresponding to a SAB rate of 0.005/1000 AVF-days.

Patient Characteristics Associated with SAB

The mean age of patients with SAB was 49 ± 11.4 years, with only one patient more than 55 years of age. The mean time on dialysis for this group was 42.7 ± 25.7 months. None of the patients with bacteremia smoked, were MRSA nasal carriers (by routine nasal swabs), or received immunosuppressive therapy (Table 5). Eight of 10 patients that developed SAB had an elevated C-reactive protein (>8.0) level at baseline.

Table 5.

Patient characteristics associated with risk of SAB while using BHC with an AVF (n = 56)

Patient Characteristic	Percent Patients with SAB (n = 10)	Percent Patients with no SAB (n = 46)
Age (years)
>55	10	30
<55	90	70
Diabetes
yes	20	22
no	80	78
Race
white	80	74
black	20	11
Asian	0	15
Smoker
yes	0	7
no	100	93
Peripheral vascular disease
yes	20	9
no	80	91
Immunosuppressive therapy
yes	0	7
no	100	93
Hemoglobin (g/L)
>11	60	67
<11	40	33
Albumin (g/L)
>3.8	40	54
<3.8	60	46
Creatinine (μmol/L)
>6.0	50	46
<6.0	50	54
Urea (mmol/L)
>35	60	35
<35	40	65
Ferritin (μg/L)
>200	40	28
<200	60	72
C-reactive protein (mg/L)
>8.0	80	65
<8.0	20	35

Other Infections

Additional abnormal blood culture results are shown in Table 3. Eight blood cultures were positive for coagulase-negative Staphylococcus in five patients, with evidence of a localized skin infection (of the AVF) in one patient, but no other clinical manifestations of systemic or local infection in the others. Follow-up cultures remained positive in two patients, and these patients received treatment. All of these infections occurred before introduction of mupirocin cream prophylaxis. There were three nonstaphylococcal infections: α-hemolytic Streptococcus, Enterobacter cloacae, and Enterococcus faecalis. In the latter two infections, a urinary source was identified, and all episodes of bacteremia were successfully treated with intravenous antibiotics.

Discussion

Infection remains a major cause of morbidity and mortality in the HD population. The BHC technique has generally been regarded as safe, provided that rigorous skin cleansing and cannulation protocols are used (9,13,14). However, few studies to date have reported long-term safety data with BHC. In this retrospective prepost comparison, we documented a significant number of bacteremic episodes in patients using BHC before the implementation of a program-wide antimicrobial cream prophylaxis protocol, with subsequent episodes documented only in patients who were nonadherent to the prescribed regimen. In addition, 4 of 10 of these infection episodes were complicated by metastatic spread, a potentially life-threatening complication of SAB. To our knowledge, this study represents the first-ever report of severe systemic infections associated with BHC.

Reporting of infection rates with BHC has generally been anecdotal (9) and limited by the small cohort sizes (14). In the study presented here, the overall rate of infection was 0.23/1000 AVF-days in the preintervention period, which was similar to that observed in a recent multicenter Canadian cohort of CHD patients using rope-ladder cannulation (15). Although this study reported 5 bacteremic episodes per 10,000 treatments (or 0.21 episodes/1000 AVF-days), measures were not taken to confirm AVF as the source of infection, and the actual SAB rate related to AVF may have been closer to that observed in our CHD cohort (0.005 infections/1000 AVF-days). Comparing the rates of SAB in patients in our study cohort with those using rope-ladder cannulation with full-care CHD, a 46-fold difference was observed. Although the CHD subjects were unmatched, on average they were older (66.5 ± 6.7 years) and the burden of comorbid illness was likely higher than that of the study cohort (16,17). Therefore, the observed SAB rate in the study cohort was more likely attributable to the BHC technique itself (or its concomitant use with nocturnal HD) than to case-mix differences. A comparison with patients receiving home nocturnal HD using rope-ladder cannulation would have been more informative. Because all patients in this study cohort used BHC, this was not possible. Unfortunately, such comparisons would be prone to confounding by indication. After initiation of the topical mupirocin protocol, no further systemic infections with S. aureus or any other pathogen occurred among patients adherent to the prescribed regimen. It is also noteworthy that no AVF failures occurred during the study period, which is in keeping with other reports of improved access longevity with BHC (18).

The mechanism by which infection risk is increased in BHC is not known, but several factors may contribute. It is well established that uremia itself is associated with impaired immune function (19). Interactions between systemic and local factors may contribute to the increased infection risk associated with BHC. Local factors may include breeches in cannulation technique, skin colonization, loss of skin integrity, frequency of cannulation, and needle movement within the needle tracts (especially during sleep). Because the total number of events in our study was relatively small, it was not possible to undertake a meaningful quantitative analysis of predictors of infection. Because our program did not provide alternate-night nocturnal HD until March 2004, we could not determine the infection rate associated with alternate-night HD in the absence of mupirocin cream prophylaxis and could therefore not evaluate HD frequency as a predictor of infection risk. Among the variables reported in Table 5, only an elevated C-reactive protein at baseline appeared to be associated with infection risk. It is possible that this patient group had subclinical infections that later evolved into overt infection. The cannulation and cleansing techniques used by our program were identical to those used elsewhere (9,20,21), thus it is unlikely unique local practices contributed to the observed risk of infection.

Antimicrobial cream prophylaxis has been studied in the peritoneal dialysis (PD) and CHD populations with generally favorable results. A meta-analysis of ten studies demonstrated combined risk reductions of 80% (95% CI = 65% to 89%) and 63% (95% CI = 50% to 73%) in HD and PD patients, respectively (22). A subsequent study found that antimicrobial cream prophylaxis in patients using CVCs for HD reduced infection rates and possibly improved patient survival (23). Despite these findings, adoption of this practice has been variable. This is likely due in part to concerns regarding the emergence of antimicrobial resistance. The incidence of mupirocin resistance among patients with regular application to PD catheter exit sites has been variable across centers and countries, with cumulative incidence rates ranging between 2.7% and 32% (24–26). Our program did not routinely screen for mupirocin resistance. However, because no infections occurred on treatment with mupirocin, it is unlikely that the prevalence high-level mupirocin resistance was significant. Low-level mupirocin resistance, although generally more common, is of limited clinical significance because the mupirocin concentration in a 2% preparation (2 g/L) safely exceeds the minimal inhibitory concentration (8 to 64 mg/L) for low-resistance organisms (27). Although mupirocin resistance can result in failure of prophylaxis, it is not independently associated with the emergence of MRSA or failure of systemic therapy for infections (28). Although it is reassuring that no MRSA infections or suspected cases of mupirocin resistance were encountered during our 11 years of observation, risk factors such as nasal carriage rates and local antimicrobial resistance patterns preclude generalization of our results to other programs.

Some methodological issues merit discussion. Because of a prepost design, we cannot exclude confounding due to changes in patient adherence and cannulation technique, which may have reduced infection rates over time. This design was chosen because it allowed data for all patients followed during either or both study periods to be used. Alternatively, a cohort study design could have compared distinct patient groups followed during the two study periods. However, this would not have altered the conclusions drawn. Secondly, given the zero-event rate, the OR for the as-treated analysis (35.3, 95% CI = 2.0 to 626.7) was estimated by manual computation with 0.5 added to the event rates. CIs were wide because of highly variable follow-up times. Future studies with more events and longer follow-up will provide more precise estimates of treatment efficacy. The use of alternative statistical tests, such as Fisher's exact test, would have been equally valid and would not have changed the treatment effect estimates or the conclusions drawn. Third, because of the retrospective nature of this study, missing data (missed infections or lapses in fistula use) may have resulted in an underestimate of the infection risk with BHC. Finally, confirmation of these findings with a randomized controlled clinical trial is desirable. However, given the potentially devastating consequences of infection and the relatively innocuous nature of the intervention, some practitioners may not feel there is sufficient equipoise to justify this approach. More observational data may be needed before the dialysis community will embrace a randomized trial. Comparisons between various antimicrobial preparations may be of value as well.

In addition to allowing for self-cannulation, proposed benefits of BHC include decreased pain, less hematomas, improved hemostasis, increased overall patient satisfaction, resolution of aneurismal dilation (29), and improved access survival (13,18). It therefore seems likely that its use will continue to increase, particularly if home dialysis continues to proliferate. Given the relatively small number of patients currently using BHC, there would be considerable value in prospectively capturing outcome data in national registries so that more rigorous studies of infection rates and predictors of infection can be undertaken and thereby inform future clinical decision-making and research. In the meantime, rigorous surveillance by patients and their healthcare professionals is warranted.

In conclusion, in this retrospective prepost study, we report a significant risk of SAB with potentially fatal metastatic complications in patients receiving home nocturnal HD with BHC. Patients considering this method should be counseled with respect to the potential risks. While awaiting more rigorous studies of efficacy and safety, topical mupirocin prophylaxis should be considered for patients using this cannulation technique.

Disclosures

None.