Summary of findings
Among hospitalised patients, intravenous therapy is the most common invasive procedure. Intravenous therapy is associated with a phlebitis rate of between 2.3% (White 2001) and 60% (Gupta 2007) and an intravenous catheter-related bacteraemia (CRBSI) rate of approximately 0.1% (Maki 2006). Current guidelines recommend that "there is no need to replace peripheral catheters more frequently than every 72 to 96 hours to reduce risk of infection and phlebitis in adults" (O'Grady 2011), and most hospitals follow this recommendation. The 2011 recommendation carries a category rating of 1B (strongly recommended for implementation and supported by some experimental, clinical or epidemiological studies). In support of the rating, the guideline cites two observational studies (Lai 1998; Tager 1983) and one RCT. The first observational study followed 3094 patients through their period of IV peripheral catheterisation and found that the phlebitis rate was 3.2% among those whose catheters remaining in situ for > seven days, compared with a rate of 4.1% and 3.9% for those whose dwell times were three and four days respectively (Tager 1983). The second observational study compared intravenous catheters left in place for 72 hours or 96 hours and found equivalent phlebitis rates (Lai 1998). The one RCT that was cited was designed to compare two types of catheter material, not dwell times (Maki 1991). The guideline also exempts children or patients with poor veins from the recommendation. In recent years, there have been improvements in catheter design and composition and more recent studies, including an earlier version of this review (Webster 2010), indicate that the recommendation may need to be revised.
Description of the condition
Peripheral vein infusion thrombophlebitis (PVT) is characterised by pain, erythema (redness of the skin), swelling, and palpable thrombosis of the cannulated vein (Monreal 1999). Diagnosis remains controversial and a number of grading systems have been proposed, although with limited validation testing performed. These include the Maddox scale (Maddox 1977) and the Baxter scale (Panadero 2002), which rank infusion thrombophlebitis according to the severity of clinical signs and symptoms. The scales are limited because not all symptoms may be present, or they may not always be present in the clusters described in the scales. Consequently, many investigators define PVT based on two or more of pain, tenderness, warmth, erythema, swelling, and a palpable cord (Maki 1991; Monreal 1999), even though it may be difficult to distinguish between pain and tenderness. More recently, a new definition for phlebitis has been proposed, one based on a more objective assessment of the insertion site (Rickard 2012). Although the precise pathogenesis of thrombus formation remains unclear, it is thought to be related to inflammation of the vein wall. Studies have been unable to demonstrate a high correlation between phlebitis and catheter infection and Maki has suggested that phlebitis may primarily be a physical response (Maki 1991). This was supported by Catney and colleagues when investigating the aetiology of phlebitis; they found that drug irritation, size of catheter, and the person inserting the catheter were all predictors (Catney 2001). Utrasonographic imaging has demonstrated thrombus formation in two thirds of catheterised veins studied and it has been suggested that catheter design may be implicated (Everitt 1997). Thus, possible causes of phlebitis are mechanical irritation from the catheter and the properties of the infusate or intravenously administered medications.
Description of the intervention
The intervention under consideration is replacing an intravenous peripheral catheter only if there are clinical indications to do so. Clinical indications include blockage, pain, redness, infiltration, swelling, leakage, and phlebitis.
How the intervention might work
Each time a catheter is inserted, the patient's skin integrity is breached and a potential portal for pathogens is provided. For example, Uslusoy found a significant relationship between the number of times infusions were inserted and phlebitis (Uslusoy 2008). Consequently, it may be prudent to limit the frequency of peripheral catheter changes as long as there is no clinical reason to do so. There is some support for this approach from observational studies that have compared outcomes between catheters remaining in situ for varying periods. In an adequately powered observational study, which included patients from medical wards and intensive care units, the investigators were unable to demonstrate any increased risk of phlebitis beyond the second day (Bregenzer 1998). Similarly, in a retrospective study of 784 intravenous catheter starts the rate of phlebitis on days one and two was 11.5%, dropping to 3.9% by day four (Homer 1998). The authors concluded that “there appeared to be less risk in continuing therapy beyond the third day than re-starting the therapy” (pp 304). Catney 2001 also failed to demonstrate any increase in phlebitis rates with the passage of time, with failure rates being less at 144 hours (1.9%) than at 72 hours (2.5%) (Catney 2001). Similarly, in a prospective investigation of 305 peripheral catheters there were 10 cases of infusion phlebitis amongst patients who had their catheter in situ for less than 72 hours whereas none were reported in patients where the dwell time was longer (White 2001). In the same study, there were three cases of post-infusion phlebitis; these all occurred amongst patients whose peripheral vein infusion catheter had been in place for less than 72 hours. Even among a high risk population of oncology and infectious diseases patients, phlebitis rates were no different when length of cannulation was dichotomised to three days or less and more than three days (Cornely 2002).
Why it is important to do this review
These observational studies create uncertainty around the US Centers for Disease Control (CDC) guidelines relating to peripheral intravenous catheter management. This uncertainty has led some hospitals to adopt the practice of re-siting only where there is evidence of inflammation or infiltration (personal communication). Included in the new CDC recommendations is a statement related to clinically-indicated (Cl I) replacement in adults, advising that this was an "unresolved issue" and referencing the previous version of this review (Webster 2010), which showed 'no difference' between the two approaches to re-siting. Making the guidelines even more difficult to rationalise is the recommendation for peripheral catheter replacement in children, which states "replace peripheral catheters in children only when clinically indicated" (O'Grady 2011). References supporting the 2011 recommendation were unrelated to dwell times (Band 1980; Maki 1973) and may indicate a mistake in the CDC's reference list (p61) (O'Grady 2011). Insertion of a peripheral intravenous catheter can be a painful and traumatic process and, if unnecessary, adds not only to a patient's discomfort but also has significant cost implications for the institution. There is a clear need to provide direction for clinicians through systematically reviewing existing studies.
To assess the effects of removing peripheral intravenous (IV) catheters when clinically indicated compared with removing and re-siting the catheters routinely.
Criteria for considering studies for this review
Types of studies
All randomised controlled trials (RCTs) comparing routine removal of peripheral IV catheters with removal only when clinically indicated were considered. Cross-over trials were not eligible for inclusion.
Types of participants
Any patient requiring a peripheral IV catheter to be in situ for at least three days for the administration of intermittent or continuous therapy (this may include patients in hospitals, nursing homes, or in community settings). Participants receiving parenteral fluids were excluded.
Types of interventions
Any duration of time before routine replacement versus clinically-indicated replacement will be included. Catheters made from any type of material (for example metal, plastic); non-coated or coated with any type of product (for example antibiotic, anticoagulant); or covered by any type of dressing (for example gauze, clear occlusive) were eligible.
Types of outcome measures
- Catheter-related blood stream infection (CRBSI) (defined as a positive blood culture from a peripheral vein; clinical signs of infection; no other apparent source for the bloodstream infection except the intravenous catheter; and colonised intravenous catheter tip culture with the same organism as identified in the blood)
- Thrombophlebitis (using any definition identified by the trial author)
- Cost (in terms of materials and labour associated with IV catheter-related insertion)
- All-cause bloodstream infection (defined as a any positive blood culture drawn from a peripheral vein while an intravenous catheter is in situ or for 48 hours after removal)
- Infiltration (defined as permeation of IV fluid into the interstitial compartment, causing swelling of the tissue around the site of the catheter)
- Catheter occlusion (identified by the inability to infuse fluids)
- Number of catheter re-sites per patient
- Local infection
Search methods for identification of studies
There was no restriction on language. If foreign language studies had been found, we intended to seek initial translation of abstracts for the application of the inclusion and exclusion criteria. Where necessary, the methods, results, and discussion sections would have been translated for inclusion in the review.
For this update the Cochrane Peripheral Vascular Diseases (PVD) Group Trials Search Co-ordinator (TSC) searched the PVD Specialised Register (last searched December 2012) and the Cochrane Central Register of Controlled Trials (CENTRAL) (2012, Issue 11), part of The Cochrane Library at www.thecochranelibrary.com. See Appendix 1 for details of the search strategy used to search CENTRAL. The PVD Specialised Register is maintained by the TSC and is constructed from weekly electronic searches of MEDLINE, EMBASE, CINAHL, AMED, and through handsearching relevant journals. The full list of the databases, journals, and conference proceedings which have been searched, as well as the search strategies used, are described in the Specialised Register section of the Cochrane Peripheral Vascular Diseases Group module in The Cochrane Library (www.thecochranelibrary.com).
Searching other resources
We contacted researchers and manufacturers in order to obtain any unpublished data. Reference lists of potentially useful articles were also searched.
We also searched the following clinical trials registries using the terms peripheral intravenous catheter and phlebitis.
- ClinicalTrials.gov (http://clinicaltrials.gov/).
- World Health Organization International Clinical Trials Registry Platform (ICTRP) (http://apps.who.int/trialsearch/).
Data collection and analysis
Selection of studies
Titles and abstracts identified through the search process were independently reviewed by JW, SO, and CR. Full reports of all potentially relevant trials were retrieved for further assessment of eligibility based on the inclusion criteria. As the review authors were also the investigators on some of the included trials, assessment was allocated to a review author who was not an investigator. Differences of opinion were settled by consensus or referral to a third review author. There was no blinding of authorship.
Data extraction and management
Following PVD Group recommendations, two review authors independently extracted data to a pre-tested data extraction form. Disagreements were resolved by discussion and, where necessary, by a third review author. We contacted authors of published and unpublished trials for additional information.
We extracted the following main sets of data from each included study:
- lead author, date;
- study participant inclusion criteria;
- country where the research was conducted;
- participants' gender and age;
- study design, randomisation processes, allocation concealment;
- intervention descriptions;
- intervention setting (hospital, home, residential aged care facilities);
- numbers of participants in each trial arm, withdrawals and dropouts;
- outcome measures, time(s) at which outcomes were assessed
The first review author entered the data into RevMan, with another review author checking the data entry accuracy.
Assessment of risk of bias in included studies
Two review authors independently assessed the included studies using the Cochrane Collaboration tool for assessing risk of bias (Higgins 2011a). This tool addresses six specific domains, namely sequence generation, allocation concealment, blinding, incomplete outcome data, selective outcome reporting, and other issues (for example extreme baseline imbalance). Disagreements between review authors were resolved by consensus or referral to a third review author. We contacted the investigators of included trials to resolve any ambiguities.
Measures of treatment effect
For individual trials, effect measures for categorical outcomes included risk ratio (RR) with its 95% confidence interval (CI). For statistically significant effects, the number needed to treat (NNT) or number needed to harm (NNH) was calculated. For continuous outcomes the effect measure we used was mean difference (MD) or, if the scale of measurement differed across trials, standardised mean difference (SMD), each with its 95% CI. For any meta-analyses (see below), for categorical outcomes the typical estimates of RR with their 95% CI were calculated; and for continuous outcomes the mean difference (MD) or a summary estimate for SMD, each with its 95% CI, were calculated. Data were analysed using the Cochrane Collaboration's Review Manager (RevMan) 5 software.
Summary of findings tables
To assess the overall body of evidence, we developed a 'Summary of findings' table for the four primary outcomes (catheter-related bloodstream infection; phlebitis; all-cause bloodstream infection; and cost) using GRADEprofiler. The quality of the body of evidence was assessed against five principle domains: 1) limitations in design and implementation; 2) indirectness of evidence or generalisability of findings; 3) inconsistency of results, for example unexplained heterogeneity and inconsistent findings; 4) imprecision of results where confidence intervals were wide; and 5) other potential biases, for example publication bias or high manufacturer involvement (Schűnemann 2011).
Unit of analysis issues
It is inadequate merely to compare longer and shorter dwell time intravenous devices (IVDs) on crude incidence of complications; this does not take into account the cumulative daily risk inherent with IVD use. There is clearly a ‘per day risk’ that is present, and grows with each day of IVD treatment, regardless of how many IVDs are used over the period of therapy. This cannot be extrapolated to mean that restricting (removing) individual IVDs will reduce overall risk. That is, an IVD in situ for seven days has seven days of exposure to risk compared with an IVD in use for only three days, but if the patient requires therapy for seven days in total then using multiple catheters over the period may not reduce risk but merely divide the same risk between multiple catheters. Appropriate time comparisons need to be made using statistics such as Kaplan-Meier analysis, logistic regression, or Cox proportional models. It is vital that the patient is used as the unit of measurement (denominator for comparison), not the IVD. If a patient requires therapy for example for five days, the patient may have one catheter used for the entire time or alternately multiple IVDs used over the five days. If the multiple catheters are viewed independently they may appear to have lower risk per catheter but the total risk for the patient over the five days may be the same.
We dealt with this by only including studies where data were available per patient rather than per catheter. Where data were not originally analysed in this format we contacted the investigators (for example Van Donk 2009) to get these data. For comparison, we have also included an analysis of phlebitis per catheter days where this information was available.
Cross-over trials were not eligible. There were no cluster randomised trials.
Dealing with missing data
If any outcome data remained missing despite our attempts to obtain complete outcome data from authors, we assessed the risk of bias of the missing data and decided if the missing data were at 'low' or 'high' risk of bias according to our risk of bias criteria (Higgins 2011a). if data were considered to be missing at random, we analysed the available information. If standard deviations were missing, we planned to impute them from other studies or, where possible, compute them from standard errors using the formula SD = SE X √¯N where these were available (Higgins 2008).
Assessment of heterogeneity
We explored clinical heterogeneity by examining potentially influential factors, for example intervention dwell time, care setting, or patient characteristics. We assessed statistical heterogeneity using the I
Assessment of reporting biases
Reporting bias was assessed using guidelines in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011a). Where sufficient study data were available for individual outcomes, funnel plots were developed and inspected for evidence of publication bias.
Where appropriate, results of comparable trials were pooled using a fixed-effect model and the pooled estimate together with its 95% CI were reported. We conducted a narrative review of eligible studies where statistical synthesis of data from more than one study was not possible or considered not appropriate.
Subgroup analysis and investigation of heterogeneity
We planned to analyse potential sources of heterogeneity using the following subgroup analyses.
- Type of randomisation (truly randomised versus not reported).
- Concealment of allocation (adequate versus not reported).
- Blinding (patients and clinicians blinded versus open-label).
- Statement of withdrawals and losses to follow up in each group (stated versus not stated).
- Intermittent versus continuous infusion.
We planned to perform sensitivity analyses to explore the effect of the following criteria.
- Concealment of allocation.
- Size of studies (< 100 patients versus at least 100 patients).
- Duration of follow up.
- Unpublished studies.
Description of studies
Results of the search
For this update there were 10 additional citations which were considered potentially relevant following screening of the search results. Two of these were publications of two unpublished trials which were included in the original review (Rickard 2010; Rickard 2012). There was one additional included study (Nishanth 2009) and two additional excluded studies (Nakae 2010; Rijnders 2004). The remaining five citations did not relate to studies using peripheral catheters. Authors of all included trials were asked for additional information. Responses were received in all cases. No additional trials were found in our search of trials registries.
Because three of the authors of this review were also investigators in trials under consideration, we allocated the assessment of those trials to review authors who were not investigators for those particular studies.
The seven trials involved a total of 4895 participants, with individual trial sizes ranging between 42 and 3283. One trial was carried out in England (Barker 2004), one in India (Nishanth 2009), the remaining five trials were Australian (Rickard 2010; Rickard 2012; Van Donk 2009; Webster 2007; Webster 2008). Five of the trials were conducted in single-centre, acute inpatient settings (Barker 2004; Nishanth 2009; Rickard 2010; Webster 2007; Webster 2008), one was a multi-centre trial in three Australian hospitals (Rickard 2012), and one was undertaken in a community setting (Van Donk 2009).
In six trials (Barker 2004; Nishanth 2009; Rickard 2010; Rickard 2012; Webster 2007; Webster 2008) patients were included if they were receiving either continuous infusions or intermittent infusions for medication therapy, whereas the catheters in the Van Donk 2009 trial were used for intermittent medication therapy only. In five trials (Rickard 2010; Rickard 2012; Van Donk 2009; Webster 2007; Webster 2008) the comparison was between routine care (planned three-day changes) and clinically-indicated changes. Barker 2004 and Nishanth 2009 compared 48-hour changes with removal for clinical indications such as pain, catheter dislodgement, or phlebitis.
Five of the trials (Barker 2004; Rickard 2010; Rickard 2012; Webster 2007; Webster 2008) used a standard definition of two or more of the following: pain, warmth, erythema, swelling, or a palpable cord. Barker 2004 and Nishanth 2009 further classified phlebitis as either mild, moderate, or severe depending on the area of erythema (Barker 2004) or on the number of symptoms (Nishanth 2009). Van Donk 2009 included the same symptoms as other trials but scored them as either one or two depending on the severity. A score of two or more was classified as phlebitis, consequently a patient may have had only one symptom, for example pain, to receive a positive diagnosis.
Power calculations were reported by Nishanth 2009; Rickard 2010; Rickard 2012; Webster 2007; Webster 2008; and Van Donk 2009 but not by Barker 2004. All of the studies had institutional ethical approval.
The table Characteristics of excluded studies contains the reasons for excluding nine trials. In summary, two were very small studies involving the administration of peripheral parenteral nutrition. Neither trial compared straightforward routine replacement with clinically-indicated removal (Kerin 1991; May 1996). One trial (Panadero 2002) compared one group that used the same catheter both intraoperatively and postoperatively with a group using two catheters, one during surgery and one postoperatively. The Haddad 2006 trial compared 72-hour changes with 96-hour changes, and the Cobb 1992; Eyer 1990; Nakae 2010; and Rijnders 2004 trials involved central venous catheters. The other excluded study was not an RCT (Arnold 1977).
Risk of bias in included studies
|Figure 1. Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.|
|Figure 2. Risk of bias summary: review authors' judgements about each risk of bias item for each included study.|
Generation of random allocation sequence
Sealed envelopes were used for allocation concealment by Barker 2004; Nishanth 2009; and Van Donk 2009; the remaining four trials used a central telephone or computer-based service (Rickard 2010; Rickard 2012; Webster 2007; Webster 2008).
It was not possible to blind either the participants or the healthcare providers in any of the trials.
The chief investigator assessed outcomes in the Barker 2004 and the Nishanth 2009 trial. In the Van Donk 2009; Webster 2007; and Webster 2008 trials, assessment was made by nurses caring for the patient or by a dedicated IV service nurse. None of the nurses were blinded to the group allocation but nor were any of them associated with the trial. In the Rickard 2010 and Rickard 2012 trials, outcome assessment was undertaken by a dedicated research nurse who was also aware of the allocation.
Incomplete outcome data
A flow chart was not provided by Barker 2004, so the numbers screened and eligible were unclear, nor were any dropouts reported. There was also an imbalance in the number of participants reported by group in this trial, which may indicate either a failure in the randomisation process in such a small trial or incomplete reporting. The number of protocol violations by group was not reported. There was complete reporting in the other six trials, all of which provided a flow of participants through each stage and used intention-to-treat analysis (Nishanth 2009; Rickard 2010; Rickard 2012; Van Donk 2009; Webster 2007; Webster 2008). In the Webster 2007; Webster 2008; and Van Donk 2009 trials, approximately one third of the participants had protocol violations and in the Rickard 2012 trial, protocol violations occurred in 16% of the participants. Primarily these were in the routine replacement groups, where catheters were not replaced within the specified time period, reflecting day to day clinical practice.
Study protocols were available for five trials (Rickard 2010; Rickard 2012; Van Donk 2009; Webster 2007; Webster 2008) and reporting followed pre-planned analyses. Barker 2004 and Nishanth 2009 reported on the expected primary outcomes.
Other potential sources of bias
In the Barker 2004 trial there were two definitions of phlebitis, one of which stated that two symptoms were necessary; yet it appears that erythema alone was diagnosed as phlebitis, with severity based on the area of inflammation. The extreme results in the Nishanth 2009 trial, where 100% of participants in the clinically-indicated group developed phlebitis compared with 9% in the two-day change group, suggests that chance or other unknown bias affected results in this small trial.
Effects of interventions
Routine changes versus clinically-indicated changes
Catheter-related bloodstream infection (Analysis 1.1)
Catheter-related bloodstream infection was assessed in five trials (4806 patients) (Rickard 2010; Rickard 2012; Van Donk 2009; Webster 2007; Webster 2008). There were no reported CRBSIs in three of these trials (Rickard 2010; Van Donk 2009; Webster 2007). When results from the remaining two trials were combined there was a 39% reduction in the CRBSI rate favouring the clinically-indicated group (clinically-indicated 1/2365; routine change 2/2441). The RR was 0.61 but the confidence intervals were wide, creating uncertainty around the estimate (95% CI 0.08 to 4.68; P = 0.64) (Figure 3).
|Figure 3. Forest plot of comparison: 1 Clinically-indicated versus routine change, outcome: 1.1 Catheter-related bloodstream infection.|
Phlebitis (Analysis 1.2 and Analysis 1.3)
All of the included studies reported incidence of phlebitis (4895 patients). When results of all trials were combined, heterogeneity was 65%. Consequently, we conducted a sensitivity analysis and removed the two trials with less than 100 participants, both of which used a two-day replacement schedule (Barker 2004; Nishanth 2009). Removing the two trials eliminated the heterogeneity (I
|Figure 4. Forest plot of comparison: 1 Clinically-indicated versus routine change, outcome: 1.2 Phlebitis.|
We also analysed the data by number of device days and, again, no differences between groups were observed (RR 1.03, 95% CI 0.84 to 1.27; P = 0.75) ( Analysis 1.3; Figure 5). In the two trials using a two-day replacement schedule compared with clinically-indicated changes (Barker 2004; Nishanth 2009), heterogeneity was over 60% so results were not combined. In the first of these two trials Barker 2004 reported that 11/26 (42.3%) participants in the clinically-indicated group developed phlebitis compared with 1/21 (4.8%) in the two-day change group. Nishanth 2009 diagnosed all of the participants in the clinically-indicated group (21/21; 100.0%) with phlebitis and 2/21 (9.5%) in the two-day group.
|Figure 5. Forest plot of comparison: 1 Clinically-indicated versus routine change, outcome: 1.3 Phlebitis per device days.|
All-cause bloodstream infection (Analysis 1.4)
One trial assessed this outcome (Rickard 2012). There was no difference in the all-cause bloodstream infection rate between the two groups (clinically-indicated: 4/1593 (0.02%); routine change 9/1690 (0.05%); P = 0.21) (Figure 6).
|Figure 6. Forest plot of comparison: 1 Clinically-indicated versus routine change, outcome: 1.4 All-cause bloodstream infection.|
Infiltration (Analysis 1.5)
A total of four trials assessed infiltration in 4606 participants (Rickard 2010; Rickard 2012; Webster 2007; Webster 2008). Infiltration of fluid into surrounding tissues was reported less often in the routine change group (452/2346; 19.3%) compared with the clinically-indicated group (518/2260; 22.9%). The RR was 1.17 (95% Ci 1.05 to 1.31; P = 0.004) (Figure 7).
|Figure 7. Forest plot of comparison: 1 Clinically-indicated versus routine change, outcome: 1.5 Infiltration.|
Local infection (Analysis 1.6)
Among the four trials measuring local infection (Rickard 2010; Rickard 2012; Webster 2007; Webster 2008) no differences were found between groups (clinically-indicated 2/2260 (0.09%); routine replacement 0/2346 (0.0%); RR 4.96, 95% CI 0.24 to 102.98; P = 0.30) (Figure 8).
|Figure 8. Forest plot of comparison: 1 Clinically-indicated versus routine change, outcome: 1.6 Local infection.|
Catheter blockage (Analysis 1.7)
Five of the seven trials, reporting on 4806 participants, were included in this analysis (Rickard 2010; Rickard 2012; Van Donk 2009; Webster 2007; Webster 2008). Rates of catheter failure due to blockage were similar between groups (clinically-indicated 398/2395 (16.6%); routine replacement 377/2441 (15.40%); RR 1.25, 95% CI 0.91 to 1.71; P = 0.16) (Figure 9).
|Figure 9. Forest plot of comparison: 1 Clinically-indicated versus routine change, outcome: 1.7 Blockage.|
Mortality (Analysis 1.8)
|Figure 10. Forest plot of comparison: 1 Clinically-indicated versus routine change, outcome: 1.8 Mortality.|
Cost (Analysis 1.9)
In each of the three trials measuring this outcome (4244 participants) (Rickard 2012; Webster 2007; Webster 2008) cannulation costs, measured in Australian dollars (AUD), were lower by approximately AUD 7.00 in the clinically-indicated group (MD -6.96, 95% CI -9.05 to -4.86; P ≤ 0.00001) (Figure 11).
|Figure 11. Forest plot of comparison: 1 Clinically-indicated versus routine change, outcome: 1.9 Cost.|
Summary of main results
This systematic review analysed catheter-related bloodstream infection, phlebitis, other reasons for catheter failure, and cost with the intention of comparing routine catheter changes (at between two and four days) with replacing the catheter only if clinical signs were apparent.
The primary outcomes of this review suggest that patients are not adversely affected if the catheter is changed based on clinical indications rather than routinely, as recommended by the US Centers of Disease Control (O'Grady 2011). The rate of catheter-related bloodstream infection was similar in both groups, between 0.0% and 0.3%, and comparable to that previously reported in prospective studies (Maki 2006). A marginal but non-significant increase in the phlebitis rate in the clinically-indicated group was apparent when data were analysed by patient but became less perceptible when data were analysed per 1000 device days, which is a more clinically useful measure. In addition, most cases of phlebitis are mild in nature, requiring no treatment or removal of the catheter. There was no indication in our review that phlebitis was a precursor to bloodstream infection.
Catheter failure due to blockage was more frequent in the clinically-indicated group. This could be expected; all catheters will fail eventually and will need to be replaced if treatment is ongoing. The outcome is not clinically meaningful, it is simply an indicator of the longer dwell times in the clinically-indicated group. Since the ‘treatment’ for a blocked catheter is replacement of the catheter, it would not be of any benefit to the patient to replace the catheter earlier since it would not reduce the need for replacement, and would instead increase the chance of re-cannulation. Many catheters do not fail over the course of IV treatment, even with extended dwell times.
Cost was significantly less, around AUD 7, in the clinically-indicated group. This result was based on three studies and results were consistent and intuitively logical (fewer catheters, less clinician time and equipment). Although, this is a seemingly small amount, it corresponds to approximately 11% of catheter-related expenditure, which may represent a considerable saving to organisations with high use (Figure 11).
Overall completeness and applicability of evidence
Trials included in this systematic review directly addressed the review question and we were able to conduct a number of meta-analyses. Apart from the Barker 2004 and Nishanth 2009 trials, results from the other five trials were quite similar. Participants were representative of those usually managed in health care. They included patients in both acute and community settings and measured outcomes important to clinicians and patients, providing useful external validity. It has been suggested that insertion and management by an IV team may explain the inefficacy of routine replacement to prevent complications (Maki 2008), yet we saw no effect in trials that had significant numbers inserted by an IV team (Webster 2007; Webster 2008) or trials where insertion was by the general medical and nursing staff (Rickard 2010; Rickard 2012). In all of the trials except for Barker 2004 and Nishanth 2009 standard guidelines were followed for the control group, that is catheters were changed at between 72 and 96 hours, reflecting usual care. In the Barker 2004 and Nishanth 2009 trials, catheters were changed every 48 hours. None of the trials, except Rickard 2012, were powered to report on phlebitis alone, and some of the trials were very small. For example, the studies that showed statistically lower phlebitis rates in the clinically-indicated group (Barker 2004; Nishanth 2009) involved just 47 and 42 people respectively and showed differences between the control and intervention groups that were quite dissimilar to all of the other studies. Consequently, results of these two trials should be interpreted with caution, particularly results from the Nishanth 2009 trial where all patients in the clinically-indicated group developed phlebitis compared with none in the two-day change group. It seems unlikely that these results would have occurred by chance but correspondence with trial authors shed no further light on these extreme results. There are no other published papers showing phlebitis rates of 100%.
Five of the seven included trials were conducted in Australia; this imbalance is difficult to understand. It would be useful to see similar studies from other healthcare systems to test the robustness of results from this review.
Quality of the evidence
Limitations in study design and implementation
Risk of bias was assessed according to six components: sequence generation, allocation concealment, blinding, selective outcome reporting, incomplete follow up, and other potential biases. All of the studies avoided selection bias and ensured allocation concealment. The methodological quality of most of the RCTs was high with one exception. It was not possible to blind the primary outcome in any of the trials. Blinding was not possible because it was necessary to identify the catheter as either 'routine change' or 'clinically indicated', to prevent inadvertent routine replacement of catheters in the intervention group. It is unclear if this had any bearing on outcomes but the review authors argue that it is unlikely (Figure 1; Figure 2). In the Barker 2004 and Nishanth 2009 trials, the investigator was directly involved in diagnosing phlebitis; in all of the other studies either medical staff, ward nurses, IV therapy staff, or research nurses evaluated the outcomes. As one author noted, it is routine practice to record reasons for removal of an intravenous catheter in the medical record, and it is unlikely that such entries would be falsified based on group allocation (Webster 2008).
Indirectness of evidence
All of the trials compared routine changes with clinically-indicated changes. However, five trials used a three to four-day change schedule and two trials changed catheters every two days. Consequently, three to four-day results may provide indirect evidence for two-day changes, conversely two-day changes provide indirect evidence for a three to four-day change schedule. Additionally, only one study (Nishanth 2009) included patients who were from a developing country and who were "usually asthenic, many underhydrated/dehydrated on admission" (personal correspondence), so the evidence may be regarded as indirect for these types of patients.
Unexplained heterogeneity or inconsistency of results
When we combined results of studies that investigated the effect of different catheter replacement schedules on phlebitis, the heterogeneity was high. This was probably due to the different schedules for the routine catheter changes or population differences, or both. Small sample sizes may also have contributed to the extreme results, which caused the heterogeneity.
Imprecision of results
Confidence intervals were wide in the pooled outcomes of catheter-related bloodstream infection, local infection, and mortality (Figure 3; Figure 8; Figure 10) indicating a high level of uncertainty around the effect size. Further research is therefore very likely to have an important impact on the confidence in the estimate of effect for these outcomes.
We feel confident that our comprehensive electronic searches identified all existing, published, randomised controlled trials addressing the review question.
Potential biases in the review process
Although the authors were investigators in one or more of the included trials, clearly described procedures were followed to prevent potential biases in the review process. A careful literature search was conducted and the methods we used are transparent and reproducible. None of the authors has any conflict of interests.
Agreements and disagreements with other studies or reviews
Our results concur with several prospective observational studies, which found no additional risk in extending IVD dwell times (Bregenzer 1998; Catney 2001; Homer 1998; White 2001). We believe the reason for this is the similarity in the mean dwell times between the intervention and control arms. Each of the included studies were pragmatic trials and, in real life, many catheters are not changed within the prescribed time frames. For example, in three-day protocols the 72-hour period may occur in the middle of the night; or a decision may be made to leave an existing catheter in place if the patient is due for discharge the following day or if they are thought to have poor veins. Conversely, the catheter may need to be removed early in any clinically-indicated group if the patient's catheter becomes blocked or infiltration or phlebitis occurs, or the patient is discharged within a couple of days of catheter insertion.
Our results also support the CDC guidelines for peripheral catheter replacement in children, which state "replace peripheral catheters in children only when clinically indicated" (O'Grady 2011). Similarly, in a guideline for timing peripheral intravenous replacement (Ho 2011) findings from the original version of this review were replicated (Webster 2010).
Implications for practice
The review found no difference in catheter-related bloodstream infection or phlebitis rates whether peripheral intravenous catheters are changed routinely every 72 to 96 hours or when clinically indicated. The consistency in these results, which now include a very large multi-site study, indicate that healthcare organisations should adopt a clinically-indicated replacement policy. This would provide significant cost savings and would also be welcomed by patients, who would be spared the unnecessary pain of routine re-sites in the absence of clinical indications. Busy clinical staff would also reduce time spent on this intervention. To minimise peripheral catheter-related complications, the insertion site should be inspected at each shift change and the catheter removed if signs of inflammation, infiltration, or blockage are present.
Implications for research
Any future trial should use standard definitions for phlebitis and be sufficiently large to show true differences. Based on results from the meta-analysis in this review, at least 2500 participants would be required in each arm of any future trial to show a lowering of phlebitis rates from 8% to 6% (α = 0.05 and 80% power). Neither pain nor satisfaction were measured in any of the reviewed studies and would be a useful addition to any future trial. Although costs were estimated in some of the included trials, a careful economic analysis of routine versus clinically-indicated replacement would be helpful for healthcare administrators. There was also some evidence from this review that different results may occur when the population is drawn from a developing country. Consequently, trials conducted in a wider variety of healthcare systems would add to the external validity of the review's results.
We are grateful to Marlene Stewart, PVD Review Group Managing Editor, for her support and speedy responses, and to the editors Mr Paul Tisi and Dr Jackie Price for their useful comments.
Data and analyses
- Top of page
- Summary of findings [Explanations]
- Authors' conclusions
- Data and analyses
- What's new
- Contributions of authors
- Declarations of interest
- Sources of support
- Differences between protocol and review
- Index terms
Appendix 1. CENTRAL search strategy
#1 MeSH descriptor: [Infusions, Intravenous] explode all trees 8402
#2 MeSH descriptor: [Catheters, Indwelling] explode all trees 877
#3 MeSH descriptor: [Catheterization, Peripheral] explode all trees 626
#4 catheter* 12181
#5 cannul* 1612
#6 intravenous near/2 (therapy or treatment) 2950
#7 #1 or #2 or #3 or #4 or #5 or #6 23117
#8 MeSH descriptor: [Equipment Contamination] explode all trees and with qualifiers: [Prevention & control - PC] 241
#9 MeSH descriptor: [Catheter-Related Infections] explode all trees and with qualifiers: [Prevention & control - PC] 66
#10 MeSH descriptor: [Phlebitis] explode all trees and with qualifiers: [Prevention & control - PC] 734
#11 *phlebitis 1910
#12 infiltrat* 3359
#13 occlusion 5295
#14 infect* 58672
#15 MeSH descriptor: [Equipment Failure] explode all trees 1354
#16 #8 or #9 or #10 or #11 or #12 or #13 or #14 or #15 68931
#17 #7 and #16 5063
#18 MeSH descriptor: [Device Removal] explode all trees 224
#19 change or routine or resit* or re-sit* replace* or remov* 139184
#20 #18 or #19 139184
#21 #17 and #20 in Trials 939
Last assessed as up-to-date: 11 December 2012.
Contributions of authors
JW conceived the idea for the review. JW and SO wrote the protocol. CR critically reviewed the protocol before final submission.
JW selected trials for inclusion, assessed methodological quality of trials and extracted data. JW entered the data, developed the analysis plan for the update and drafted the review update.
SO arbitrated on the selection of trials, assisted with data extraction, assessed methodological quality and assisted in drafting the final review.
CR selected trials for inclusion, assessed methodological quality of trials, extracted data, assisted with interpreting results and drafting of the final review.
KN assessed methodological quality of trials, extracted data, and commented on the review update.
Declarations of interest
CR's department has received a grant in aid for a research project from a manufacturer of peripheral intravenous catheters (Becton Dickinson, Australia). The sponsor had no involvement in study design, execution, analysis or publication. The research project was unrelated to the topic of this review.
Sources of support
- No sources of support supplied
- Chief Scientist Office, Scottish Government Health Directorates, The Scottish Government, UK.The PVD Group editorial base is supported by the Chief Scientist Office.
Differences between protocol and review
The primary outcome was changed to catheter-related bloodstream infection; all-cause bloodstream infection was added as a secondary outcome. This was done to more closely differentiate between the two outcomes.
The methodological quality assessment of the included studies has been updated to the Cochrane Collaboration tool for assessing risk of bias (Higgins 2011a).
Medical Subject Headings (MeSH)
Catheter-Related Infections [*prevention & control]; Catheterization, Peripheral [adverse effects; economics; *instrumentation]; Catheters, Indwelling [adverse effects]; Device Removal [*standards]; Guideline Adherence; Incidence; Phlebitis [epidemiology; etiology]; Randomized Controlled Trials as Topic; Time Factors
MeSH check words