Central venous access devices (CVADs) are essential for managing critically ill patients in the intensive care unit (ICU). Despite their utility and widespread use (Climo et al, 2003; Timsit et al, 2020), CVADs are associated with multiple complications, including central line-associated bloodstream infection (CLABSI) (Timsit et al, 2018; Xiong and Chen, 2018), with an incidence rate as high as 6.9 per 1,000 catheter days (Holton et al, 2006; Takashima et al, 2018; Wichmann et al, 2018). Central line-associated bloodstream infections have significant attributable mortality and increase in-hospital length of stay (LOS), readmissions, and costs (Haque et al, 2018). Recent literature found CLABSI to have an odds ratio up to 2.75 for in-hospital deaths (Ziegler et al, 2015; Huerta et al, 2018). In the United States, CLABSI has the second-highest attributable LOS of any hospital-acquired infection, at 15.7 days (Zimlichman et al, 2013). Furthermore, CLABSI patients are readmitted to the hospital at a higher rate than non-CLABSI patients (Khong et al, 2015) and increase healthcare costs by an average of US $45,814 per case (Khong et al, 2015; Stevens et al, 2014; Zimlichman et al, 2013).
Multiple sources of contamination contribute to the incidence of CLABSI (Rupp and Karnatak, 2018). Pathogens gain access to the luminal surface of a CVAD via colonization of the catheter hub, hematogenous seeding, or contaminated infusate (Rupp and Karnatak, 2018). Ample literature has prompted evidence-based practices to reduce the incidence of CLABSI in ICU settings (Park et al, 2017; Bannatyne et al, 2018). In light of these interventions, hub colonization has emerged as the dominant route of infection (Rupp and Karnatak, 2018). Once pathogens gain access to the CVAD, the lumen acts as a nidus of colonization, and bacteria enclose themselves in a protective matrix called a biofilm (Ielapi et al, 2020). Pathogens growing in a biofilm are particularly resistant to antibiotics and host defences (Ielapi et al, 2020). Frequent handling of a colonized catheter hub risks displacing the biofilm, potentially leading to bacterial invasion of the catheter's inner lumen (Leistner et al, 2013).
Various interventions have been suggested to decrease the incidence of CLABSI, such as antimicrobial-impregnated catheters and antibiotic locking solutions (Velasquez Reyes et al, 2017). Some studies demonstrate a reduction in CVAD colonization when using impregnated catheters.
However, results vary based on setting, and widespread use is not recommended (Bell and O'Grady, 2017; Chong et al, 2017; Timsit et al, 2018). A recent randomized controlled trial compared antimicrobial-impregnated CVADs with standard CVADs in a pediatric ICU setting, without a statistically significant difference in outcome (Gilbert et al, 2019). Antibiotic lock solutions have also been investigated to decrease the incidence of CLABSI (Vassallo et al, 2015) with little efficacy and a tendency to cause antibiotic resistance (Bueloni et al, 2019).
Recent studies have demonstrated topical chlorhexidine (di)gluconate's (CHG's) effective role in reducing CLABSI (Pallotto et al, 2019; Scheier et al, 2021). Chlorhexidine (di)gluconate is generally safe and serious allergic reactions are rare. The antimicrobial efficacy of CHG has been shown in vitro and in preclinical animal trials (Kowalewska et al, 2018). Chlorhexidine (di)gluconate significantly reduced Staphylococcus aureus contamination in vitro, decreasing bacterial load by 6 log10 colony-forming units (CFU), and 3-4 log10 CFU/lumen in Yorkshire swine with a CVAD inserted into the jugular vein (Kowalewska et al, 2018). This preclinical work informed the potential of a device that delivers CHG locking solution (CHGLS) to decrease incidence of CLABSI. This device instills CHG into CVADs to sterilize the inner catheter lumen. This trial also aims to establish the feasibility of a large randomized control trial (RCT) of a chlorhexidine (di)gluconate lock solution by determining uptake of the device and rates of eligible patients, consent, and compliance.
Materials and methods
A detailed protocol describing the study methods was published and is publicly accessible (Zamir et al, 2020).
Study design
This pilot study of the CHGLS was a two-arm, parallel-group, open-label, single-centre, feasibility RCT. Approval was granted by the Hamilton Integrated Research Ethics Board (HiREB #3654). This trial was registered at ClinicalTrials.gov (NCT03309137) before the enrolment of our first participant.
Setting
Participants were recruited from three participating ICUs at the Hamilton General Hospital (HGH), a large academic hospital and major cardiac surgical centre in Hamilton, Ontario, Canada. In April 2018, the trial was expanded from one cardiac surgical ICU to include two medical-surgical ICUs due to the slow recruitment rate.
Participants
Eligible patients were greater than 18 years of age and admitted to the ICU with any CVAD expected to remain in situ for at least 72 hours. Most patients had either a large bore introducer catheter, a 3-lumen catheter, or a peripherally inserted central catheter (PICC). Participants were recruited within 72 hours of ICU admission. Exclusion criteria included patients expected to be discharged from the ICU within 36 hours of admittance; care limited to palliative measures or prognosis deemed hopeless by the attending physician; patients were receiving antibiotics for a known or suspected infection; patients with a chronic indwelling CVAD in situ; and patients with a known CHG allergy. Patients were screened daily for eligibility. Patients or their substitute decision-maker (SDM) were approached by research staff for informed consent.
The target sample size was 100 participants, with 50 in each group. The sample size was not derived statistically due to the exploratory nature of pilot trials.
Randomization, concealment and blinding
Randomization was conducted in blocks of two, four, six, and eight in a fixed 1:1 ratio using the statistical computer software, R v3.4.2. A third-party member used the list generated to prepare 100 opaque security envelopes containing a note that indicated either CHG or standard care. Upon obtaining informed consent, the research coordinator or assistant opened the security envelope that corresponded with the participant's study ID to determine allocation. Due to the lack of a placebo device, clinical, and research staff were not blinded to treatment allocation.
Interventions
The standard care group continued to operate as per hospital-approved protocol; when venous catheters were not infusing, they were flushed, instilled with 0.9% normal saline (NS) solution, and capped with a luer lock adapter. Participants allocated to the CHGLS group received a set volume of CHG in the catheter lumens in addition to usual care. Venous catheters that were not infusing were maintained with the addition of a pre-specified volume of CHG solution (see video and description previously published; Zamir et al, 2020). This was achieved by attaching a sterile CHG adapter onto a sterile saline syringe and pushing a pre-specified volume (indicated by the type of lumen) of CHG-infused solution into the catheter lumen (Zamir et al, 2020). The catheters were then capped with a luer lock adapter. Before re-accessing the venous catheter, all fluids were aspirated to clear the lumen of CHG. Device use was discontinued when participants were discharged from the ICU, or all venous catheters were removed.
A single set of blood cultures was drawn from all CVADs at baseline and every 48 hours for participants in both groups. Blood cultures were analyzed for infectious organisms in the core laboratory at the HGH. Bedside nurses documented flushing and locking procedures in study-specific logs, which were assessed daily by the research team for compliance.
Prior to the study start-up, clinical staff were educated on the protocol, device, and intervention. Educational activities included group and one-to-one staff training sessions, one-page handouts circulated within the ICU, and an accessible video detailing the study protocol (Zamir et al, 2020). Two nurse champions were identified to provide support to bedside staff.
Outcomes
The primary outcome of this pilot study was the feasibility of conducting a large, full-scale RCT. The contributing outcomes were (1) consent rate, measured as the proportion of patients and SDMs approached who consented to participate in the trial, (2) recruitment rate, measured as the proportion of eligible patients who were randomized, (3) protocol adherence, measured using tracking logs for flushing and locking procedures, and (4) staff comfort with the trial protocol, measured via serial surveys of the ICU nursing staff.
Secondary clinical outcomes related to the antimicrobial efficacy of CHGLS and prevention of central line infections were also assessed. Secondary outcomes were (1) central line colonization, defined as a positive central line culture and negative peripheral venous specimen, (2) bacteremia, defined as a positive venous specimen and correlation with central line cultures, and (3) clinical endpoints, including adverse events, length of stay (LOS) in ICU, hospital LOS, ICU mortality at 28 days, and hospital mortality at 28 days. The sample size was not powered to capture the statistical significance of clinical outcomes.
Survey development
To determine the comfort level of ICU nurses with the CHGLS, we distributed a paper-based survey to nurses involved in the trial. We developed a nine-item survey to assess four domains: (1) ease of study-related tasks, (2) time for device use, (3) compliance with the study protocol, and (4) effectiveness of educational activities and support materials. These domains were selected after consultation with the ICU educator and two nurse champions. Questions were dichotomous, multiple-choice, or five-point Likert scales and respondents were invited to provide demographic data. The survey was pilot tested and assessed for clinical sensibility before distribution using a convenience sample of six ICU RNs. Feedback was provided on the appropriateness, redundancy, and completion time of the tool. The survey was distributed in all participating ICUs at nurse huddles and directly to nurses caring for trial participants.
Statistical analysis
Descriptive statistics were used to analyze feasibility outcomes and baseline demographics. Continuous data were tested for normality using histograms, normal Q-Q plots, skewness, kurtosis, and the Shapiro-Wilk test. Normally distributed data were summarized using mean and standard deviation. Non-normal data were summarized using medians and interquartile ranges, and categorical variables were summarized using percent proportions. Group differences were tested using the chi-square test for homogeneity, Mann–Whitney U test, and Levene's test of homogeneity. All statistical tests were conducted using SPSS version 24 (IBM Corp, Armonk, NY). No subgroup analyses or interim analyses were performed, and we adhered to the intention to treat analysis.
Results
Primary feasibility outcomes
Between November 2017 and July 2019, 100 patients were randomized to the study, with 50 allocated to each group. All randomized participants were included in the analysis, the intention-to-treat principle was followed. The trial was discontinued once the target sample size was achieved. Of the 966 ICU days of data collected, 504 were from participants allocated to CHGLS and 462 from participants assigned to standard care. Baseline characteristics were statistically similar between groups (Table 1).
Table 1. Patients’ baseline and demographic characteristics by treatment group
| Characteristic | Standard care (n=50) | Chlorhexidine (di)gluconate Lock (n=50) | P value |
|---|---|---|---|
| Gender, male, n (%) | 33 (66.7) | 27 (54.0) | 0.221 |
| Age, median [IQR] | 71.5 [60, 77] | 68.5 [49, 76] | 0.207 |
| BMI, median [IQR] | 30.6 [25.3, 34.3] | 30.5 [26.8, 34.4] | 0.589 |
| ICU admission, n (%) | |||
| West** | 34 (68) | 28 (56) | 0.327 |
| East | 10 (20) | 12 (24) | 0.915 |
| South | 6 (12) | 10 (20) | 0.239 |
| Diagnosis at ICU admission, n (%) | |||
| Non-operative | 9 (18) | 9 (18) | 1 |
| Post-operative | 41 (82) | 41 (82) | 1 |
| Pre-existing comorbidities, n (%) | |||
| Diabetes mellitus (no insulin) | 12 (24) | 10 (20) | 0.629 |
| Diabetes mellitus (insulin) | 5 (10) | 4 (8) | 0.727 |
| Congestive heart failure | 11 (22) | 7 (14) | 0.289 |
| Ischemic heart disease | 4 (8) | 5 (10) | 0.727 |
| Chronic lung disease | 1 (2) | 2 (4) | 0.558 |
| Liver disease | 1 (2) | 1 (2) | 1 |
| Malignancy | 4 (8) | 4 (8) | 1 |
| Chronic renal insufficiency | 3 (6) | 3 (6) | 1 |
| Chronic dialysis | 0 (0) | 1 (2.0) | 0.315 |
| Other immunocompromised state | 5 (10) | 7 (14) | 0.538 |
| Chronic health index of zero, n (%) | 23 (46) | 23 (46) | 1 |
IQR= interquartile range, BMI=body mass index, ICU=intensive care unit
*ICU West is a 17-bed cardiac surgical ICU, while ICUs East and South are medical-surgical ICUs with 16 and 13 beds respectively
Consent rate
Of the 122 patients and SDMs approached, 100 (82.0%) consented to participate in the trial, and an average of 1.1 participants were enrolled per week. All study participants were followed up for the duration of their ICU and hospital stay, with three participants (one CHGLS and two standard care) lost to follow-up at 28 days.
Recruitment rate
Over the 21-month recruitment period, 3,848 patients were screened, of whom 173 (4.5%) met the inclusion and none of the exclusion criteria. Reasons for ineligibility were central lines expected to be removed within 72 hours (n=1,975), no CVAD in situ (n=782), and patients receiving antibiotics for a known or suspected infection (n=676). Of the 173 patients who met our eligibility criteria, 122 (70.5%) were approached by a member of the research team. Eligible patients were missed when research staff or SDMs were unavailable (n=42) or when the attending physician disagreed with patient suitability for enrollment (n=5). The reasons for non-inclusion can be found in Figure 1.
Protocol adherence
Tracking logs were retrieved from the medical charts of 59 participants at study completion. Tracking logs were missed when participants were transferred from the ICU within the HGH, and the research team was unable to locate the tracking log (n=25); participants died in the ICU, and their tracking log was sent to the morgue along with their medical chart (n=9); participants were transferred to a different ICU (n=7).There were 32 tracking logs retrieved for participants allocated to CHGLS. Of the 447 times participants’ venous catheters were accessed, CHGLS was used as indicated by the study protocol 408 (91.3%) times for a median of 10.5 [4.25-22.25] uses per patient. The standard of care for central line flushing is twice daily or after infusion and was similar in both arms of the study. The reasons CHGLS was not used per protocol can be found in Figure 2. There were 27 tracking logs retrieved for participants allocated to standard care; participants’ venous catheters were flushed 437 times for a median of 12 [IQR, 2-22] times per patient.
Level of comfort with protocol intervention
The survey response rate was 51% (n=51). A summary of respondents’ demographic data can be found in Table 2. On a five-point Likert scale from ‘very difficult’ to ‘very easy’, most respondents found CHGLS ‘very easy’ or ‘easy’ (34 [66.7%]) to use, and only two respondents (3.9%) indicated that using CHGLS was ‘difficult.’ Most respondents (42 [82.4%]) felt comfortable using the device, and most nurses (44 [86.3%]) indicated they were able to access the research team for help when necessary. The CHGLS device was not found to be labour intensive, only two respondents (5%) required more than five minutes to use the device, and no respondents indicated it ‘significantly increased’ their workload. Some of the educational and supportive activities implemented by the research staff were deemed ineffective by a significant proportion of survey respondents; a summary can be found in Figure 3. Survey respondents shared their opinions about CHGLS in an open-ended format; many commented that the device was easy to use (n=17), and several were keen on being part of a study to reduce bloodstream infections (n=5).
Table 2. Survey respondents demographic characteristics
| Characteristic | Respondents n (%) |
|---|---|
| Gender, male (n=43)* | 3 (7) |
| Employment status (n=43) | |
| Full-time | 34 (79) |
| Part-time | 9 (21) |
| Years of nursing experience (n=44) | |
| 1–5 | 3 (7) |
| 6–10 | 6 (14) |
| 11–15 | 9 (20) |
| >15 | 26 (59) |
| Years of ICU experience (n=43) | |
| 1–5 | 13 (30) |
| 6–10 | 6 (14) |
| 11–15 | 4 (19) |
| >15 | 20 (47) |
| Allocation of participants treated (n=51) | |
| Chlorhexidine (di)gluconate lock | 38 (75) |
| Standard care | 13 (25) |
| Number of participants treated (n=51) | |
| 1 | 37 (73) |
| 2 | 8 (16) |
| 3 | 5 (10) |
| 4 | 1 (2) |
*Some respondents elected not to completed the demographics portion of the questionnaire.
**Respondents that cared for participants from each treatment group were characterized as chlorhexidine (di)gluconate lock
Secondary clinical outcomes
The study was not powered to determine the statistical significance of clinical outcomes however there was an a priori plan with the funder to test efficacy. No adverse drug events, serious adverse drug events, or unintended effects were reported throughout the duration of the trial. A total of 277 blood cultures were drawn from CVADs between both groups, 138 from participants allocated to the standard care group and 139 from participants assigned to CHGLS. The proportion of central line colonization was significantly higher in the standard care group, 40 (29%) versus 26 (18.7%) in the CHGLS (P=0.009). However, there was no statistically significant difference between groups in the proportion of participants who developed bacteraemia during their ICU stay, with nine (17.6%) in the standard care group compared to three (6.1%) in the CHGLS group (P=0.065). No difference was found between groups for ICU length of stay, hospital length of stay, or mortality at 28 days. A summary of clinical outcomes can be found in Table 3.
Table 3. Comparison of clinical outcomes between study groups
| Outcome measure | Standard care (n=50) | Chlorhexidine (di)gluconate lock (n=50) | P value |
|---|---|---|---|
| Clinical end points | |||
| LOS in ICU, median [IQR] | 6 [4, 13] | 7 [3, 16.5] | 0.649 |
| LOS in hospital, median [IQR] | 13 [8, 31] | 18 [10, 30] | 0.485 |
| Death at 28 days, n (%) | 12 (24%) | 9 (18%) | 0.649 |
| Bacteremia, n (%)* | 9 (17.6) | 3 (6.1) | 0.065 |
| Proportion of CVAD colonization, n (%)** | 40 (29.0) | 26 (18.7) | 0.009 |
LOS=length of stay; IQR=interquartile range; ICU=intensive care unit; CVAD=central venous catheter
*Bacteraemia was defined as a positive stab culture and correlation with a positive central line culture
**CVAD colonization was defined as a positive central line culture with a negative peripheral poke. A total of 138 blood cultures were drawn from the CVADs of participants in the standard care group with 139 from the chlorhexidine (di)gluconate lock group
Discussion
This CHGLS pilot trial is the first human RCT to investigate a CHG locking device for the prevention of central line infection in ICU patients. Participant recruitment was slower than anticipated, with an average of 1.1 participants enrolled per week. The consent rate for the trial was high and exceeded our initial target of 80%. For participants allocated to CHGLS, tracking logs indicate the device was used per protocol 91.3% of the time. However, this method of recording adherence has limitations and is vulnerable to human error. Intensive care unit nurses felt comfortable using the device and found the device easy to use. Furthermore, using the CHGLS device was not overly time-consuming or burdensome and did not considerably change the workload of ICU nurses. The proportion of participants with CVAD colonization was significantly higher in the standard care group, but this pilot study was not powered to determine if this translated to a reduction in CLABSI.
To our knowledge, no previous trial investigating a medical device to reduce central line infection in the ICU has conducted a comprehensive evaluation of device uptake and adherence of ICU nurses. Past research has demonstrated that collaboration between research staff and hospital nurses optimizes the quality and implementation of the catheter-related protocol (Johnson et al, 2016), and continuous feedback from ICU nurses is critical in interventional trials to decrease catheter-related infections (Musu et al, 2017; Tyson et al, 2020). Our survey findings highlighted an opportunity to improve educational and supportive materials, including initial in-servicing, as it was deemed not effective by 34% of respondents. In future trials, we suggest that in-servicing occur in tandem with patient enrollment for the trial duration to optimize protocol adherence.
Various interventions have been investigated to prevent central line infections in ICU patients (Landry et al, 2010; Vassallo et al, 2015; Lai et al, 2016; Bell and O'Grady, 2017; Chong et al, 2017; Velasquez Reyes et al, 2017; Bueloni et al, 2019; Gilbert et al, 2019), with limited success and potentially harmful complications, such as increased thrombotic events and antibiotic resistance, limiting widespread adoption (Landry et al, 2010; Liu et al, 2014). An RCT by Storey et al investigated the comparative efficacy of CHG-impregnated versus non-CHG-impregnated PICCs (Storey et al, 2016). These authors found no significant difference in the incidence of CLABSI between groups, contradicting the findings of a previous quasi-experimental study of the CHG PICC (Rutkoff, 2014). Limited clinical efficacy has also been demonstrated for catheters impregnated with silver nanoparticles, which have been shown to have no significant effect on CVAD colonization or CLABSI incidence in an ICU setting (Antonelli et al, 2012). Ethanol lock solution has been a recent focus of investigation for reduction of CLABSI (Pérez-Granda et al, 2014); however, it has been predominately trialled in patients with short-term hemodialysis catheters (Souweine et al, 2015) and has been shown to increase catheter occlusion (Wolf et al, 2018). In light of the scarcity of clinically effective innovations, our findings demonstrate the preliminary efficacy of a CHG locking device for reducing central line colonization rates in ICU patients and a signal of reduced bloodstream infections.
The strengths of this trial include the RCT design, diverse patient population, and embedded questionnaire. Well-designed RCTs are rarely conducted to assess medical devices (Neugebauer et al, 2017), but must address the unique barriers facing protocol uptake and adherence (Jenks et al, 2016;Valencia et al, 2016). This pilot trial was conducted in three ICUs at a large academic hospital, including cardiac surgery, trauma, and neurological injury. Survey data on device uptake of ICU nurses will help optimize compliance in a larger RCT. This study has several limitations, including lack of blinding, generalizability beyond the ICU, and accurate documentation. The treating team, participants, and outcome assessors were not blinded to study allocation, introducing bias. Future trials should consider implementing a sham device as a comparator to the CHGLS device. The generalizability of this trial may be limited because this was a single centre and the high proportion of excluded patients. In addition, this pilot may not be applicable to settings outside of the ICU, where contact with central lines may be less frequent. Our clinical findings should be corroborated in a larger, multicentre RCT. At study completion, 39 tracking logs were not retrieved, presenting a significant source of missing data. Literature indicates that documentation errors and omissions are frequent in the ICU (Artis et al, 2019; Eltaybani et al, 2019). Consequently, future trials should consider electronic documentation systems, which have been shown to reduce errors and do not require physical retrieval (Thongprayoon et al, 2016).
Conclusion
The use of CHGLS in an ICU setting is feasible and safe. Recruitment was slower than anticipated, the target consent rate was achieved, and bedside nurses felt comfortable using the device. The proportion of CVAD colonization was significantly lower among CHGLS patients. Findings from this trial will inform the conduct of a larger RCT and provide preliminary data on the clinical efficacy of a novel CHG locking device.