In 2010, changes were made to the way vascular access was established in trauma patients within the NHS. A more structured approach to traumatic injury management with trauma centres and networks was introduced (NHS Confederation, 2010). These included changes to paramedic practice within NHS ambulance services and European resuscitation guidance (Soar et al, 2015; Joint Royal Colleges Ambulance Liaison Committee, 2016). One change to practice was the advocation of intraosseous (IO) access in critically unwell adults—this required using new devices for the insertion of IO cannulae.
A device commonly found in UK paramedic practice is the EZ-IO. It is a reusable driver paired with disposable cannulae of differing lengths for use in various ages and insertion sites, excluding the sternal site. This represents a constraint which may limit the use of this intervention and result in decreased survivability should distal long-bone sites be contraindicated.
An initial scoping review of the existing literature showed a dearth of research focusing on the quasi-scientific assessment of intraosseous devices in UK pre-hospital practice. A brief database search using the terms ‘pre-hospital’ and ‘intraosseous’ found only nine articles from the past decade—none of which were directly related to UK practice. The lack of evidence combined with the wide range of devices that are available indicated a significant opportunity for research. Therefore, a literature review was conducted to find existing research, critically appraise it, synthesise it, and draw a conclusion that can be applied to paramedicine.
The aim of this article was to examine evidence comparing devices for: success at first insertion attempt; time of insertion; ease of use; and flow rates. Device cost was also considered. Additionally, the article aims to either support or refute the current device of choice and any need for further research. The article focused on adult participants for two reasons:
Buelow's (2006) four-step process was employed to devise the research question that underpins this review. These four steps involved: defining the clinical problem; identifying existing literature and any shortcomings within it; defining the purpose of the research, what it aims to do and why; and finally, the research question is defined. This is based on the PICO approach by Sackett et al (1997) whereby the participant groups (P), intervention (I), comparator (C) and outcome (O) are defined to finalise the research question.
Methods
The following medical literature databases were searched: Allied and Complementary Medicine Database (AMED); British Nursing Index (BNI); CINAHL Plus; Cochrane Library; Medline; and PubMed. A range of search terms were included to find the most relevant literature and all were combinations of the primary term ‘intraosseous’ with ‘ease’, ‘insertion’, ‘site’, ‘success’ and, ‘stern*’ (the asterisk allowed for the search of associated terms, such as sternum and sternal). The primary focus of the search was randomised controlled trials (RCTs) or meta-analyses. However, given that the intervention is still relatively new, invasive, and targeted at a rare group of the critically ill or injured, the inclusion of lower forms of evidence was appropriate to ensure maximal reporting.
Searches were conducted and article titles were screened for relevance These articles were evaluated against the criteria laid down in a bespoke data extraction table to ensure the literature is examined for all features relevant to the research question. Additional data were also identified using a ‘snowballing’ technique to find relevant references within the selected articles—reducing the chances of missing a relevant trial. The data were assessed against the first two criteria: articles less than 10 years old and participants over the age of 18. The literature was then appraised for the relevance of its context and participant groups, the design and methodologies used and, finally, the outcomes of the trial. Articles that met the study criteria were selected for more comprehensive critical appraisal.
For accuracy and replicability, the articles were appraised using the NHS Critical Appraisal Skills Programme (CASP) frameworks (Public Health Resource Unit, 2006). By applying this framework which is relevant to the trial methodology, a comprehensive assessment of the literature and its suitability to inform practice were completed.
Results
Initial searches yielded 2262 potentially relevant titles. Of these, 47 were identified as potentially suitable and were examined in depth (Table 1).
| Search number | Search term | Number returned |
|---|---|---|
| 1 | ‘Intraosseous’ | 6052 |
| 2 | 1 AND ‘access’ | 961 |
| 3 | 1 AND ‘ease’ | 63 |
| 4 | 1 AND ‘insertion’ | 361 |
| 5 | 1 AND ‘site’ | 462 |
| 6 | 1 AND ‘success’ | 313 |
| 7 | 1 AND ‘stern*’ | 102 |
| Total return of articles | Titles screened for relevance | 2262 |
AMED, BNI, CINAHL Plus, Cochrane Library, Medline, PUBMED
A provisional appraisal of the articles was carried out at the data extraction stage. From the 47 articles, 18 were selected as most relevant according to the extraction criteria.
Of these 18 articles, 7 were RCTs and the remaining 11 were observational studies (Table 2). There was a mixture of direct comparison of devices for efficacy/flow rates and observation of individual devices performance. No systematic reviews were identified. Owing to a lack of paramedic-specific evidence, articles were included if they used non-paramedic investigators on the basis that the intervention is the same.
| Brenner et al | Demir et al | Derikx et al | Hammer et al | Kurowski et al | Leidel et al | Szarpak et al | |
|---|---|---|---|---|---|---|---|
| Methodology | 1b | 2b | 2b | 2b | 1b | 2b | 2b |
| Population | Doctors and paramedics deploying devices into cadavers | Doctors and nurses deploying devices into mannequins | Critically ill or injured adults, device deployed by investigator | Medical students deploying devices into cadavers | IO familiar paramedics deploying into mannequins | Critically ill or injured adults, device deployed by investigator | Paramedics deploying devices in mannequins |
| Population size | 39 – Man |
30 – BIG |
26 – BIG |
27 – EZ-IO T |
107 – EZ-IO T |
20 – BIG |
37 – EZ-IO T |
| Devices studied | Man EZ-IO T | BIG EZ-IO T | BIG EZ-IO T | EZ-IO T EZ-IO H FAST | EZ-IO T BIG Man | BIG EZ-IO H | EZ-IO T NIO |
| 1st pass success rate | EZ-IO = 97.8% |
N/A | BIG = 92.3% |
EZ-IO T = 91% |
BIG = 92% |
BIG = 80% |
EZ-IO T = 100% |
| Time | EZ-IO = 32 ±11s |
N/A | BIG = 2.8 ± 1.2s |
EZ-IO T = 17 ± 7s |
EZ-IO T = 3.1 ± 0.9m |
BIG = 2.2 ± 1.0m |
EZ-IO T = 26s |
| Ease of insertion | Rated 1–6 |
Safe use at 12 months |
Rated on 100 mm VAS |
Rated 1–6 |
Rated sequentially Big = 1st |
N/A | Rated 1–5 |
| Maximal flow rates (ml/min) | N/A | N/A | N/A | EZ-IO T = 226 ± 173 |
N/A | N/A | N/A |
T = tibia, H = humeral (methodology graded using Oxford CEBM system)
The studies selected featured the following IO access devices:
Ethics
Key ethical considerations in trials of this nature are freedom from industry sponsor or influence; approval of the trial by an ethics committee; and the appropriate enrolment and consent of participants. This last point is especially important when this participant group may not be able to provide prior informed consent (Sherwin, 2015).
The ethics of the trials were found to be largely sound; all declared no conflict of interest but one did not mention it. There was appropriate ethical approval; consent gained in all trials; and funding was declared in all but three. This may have been significant as the devices are relatively expensive and influence from the pharmaceutical industry could have added bias to results (Sismondo, 2008).
Outcomes
Success of insertion
Semi-automatic devices were more successful at insertion than manual devices. However, there wasn't sufficient evidence beyond this to differentiate between the semi-automatic devices. Brenner et al (2008) showed that the EZ-IO tibia had higher success rates and fewer complications than the manual device. The EZ-IO pairs a reusable battery-powered driver with IO needles of various lengths to account for the anatomy and physiology of the individual patient; the needles may be sited in a variety of locations, most commonly the tibial tuberosity just distal to the knee and the medial aspect of the humeral head. This article will refer to the area of insertion as EZ-IO tiba and EZ-IO humeral where appropriate. Kurowski et al (2014) also found that both BIG and EZ-IO tibia were significantly more successful than manual devices but there was little difference between the BIG and EZ-IO.
Demir et al (2016) showed that the BIG was slightly more successful than the EZ-IO tibia but the study was underpowered and there was a high p-value given to the marginal difference. Hammer et al (2016) showed that when comparing FAST with EZ-IO tibia and humeral, again, there was little difference in success rates but the study was underpowered, using only 27 cadavers so the significance of this finding may be misleading.
Leidel et al (2010) found no significant difference between the EZ-IO tibia and the BIG. Another study by Szarpak et al (2016) showed no difference in success between the NIO and the EZ-IO tibia; however, again, lack of power may render this result inaccurate. The inconclusive evidence in the RCTs is added to by a handful of observational studies, none of which are conclusive. Byars et al (2011) showed good success rates using the FAST but the study was underpowered. Frascone et al (2007) also found that the FAST had a good success rate but that the EZ-IO tibia was slightly better.
Both Reades et al (2011) and Vassallo et al (2014) showed that the EZ-IO tibia was more successful than when used at the humeral site, although the study was underpowered. Santos et al (2013) supported this high success rate with the EZ-IO tibia, whereas Wampler et at (2012) contradicted the generally poor performance on the EZ-IO humerus with a high success rate.
Time of insertion
There was little difference in insertion times across the studies, some articles favouring one device over another. Differences in reporting parameters limit the comparability of data. Brenner et al (2008) found little difference between the EZ-IO tibia and the manual IO, at around 33 seconds each. Demir et al (2016) found that the BIG took half as long to insert as the EZ-IO, but the times were still very closely matched. In addition, the short insertion times were unlikely to have any clinical significance. Similarly, well-matched insertion times for the EZ-IO tibia, humerus and FAST were found by Hammer et al (2016). Again, Szarpak et al (2016) found that the NIO took half the time to insert compared with the EZ-IO tibia. All these studies recorded times in seconds, whereas the following studies recorded times of several minutes—suggesting different timing methodologies.
Leidel et al (2009) found that the EZ-IO tibia was faster than the BIG but the study was underpowered. Kurowski et al (2014) showed the opposite with BIG being faster than EZ-IO tibia and both faster than a manual IO. The observational study by Byars et al (2007) found mean insertion times for the FAST of around 1 minute, which is again in contrast to the approximately 30-second times reported by Hammer et al (2016).
Ease of use
Beyond indicating that any semi-automatic device was easier to use than a manual device, the data lack consistency and it is difficult to draw conclusions. Brenner et al (2008) found that the EZ-IO was rated as easier to use than the manual IO, but there was only a small overall difference. In Demir et al (2016), the BIG was rated significantly easier to use than the EZ-IO at either possible site.
Derikx et al (2014) observed that 12-month skill retention in the BIG was worse than in the EZ-IO tibia, which could be extrapolated to suggest that the BIG was harder to use and to retain the necessary skills. Hammer et al (2016) found that participants favoured the EZ-IO tibia over the FAST in a ratio of around 2:1—no participant in this study preferred using the EZ-IO in the humeral site. Kurowski et al (2014) and Brenner et al (2008) both found that the manual IO was the hardest to use, followed by the EZ-IO tibia and then the BIG as easiest.
Szarpak et al (2016) found that when compared, the NIO scored around half as well as the EZ-IO tibia on a Likert scale for difficulty. The trend of moderate performance in the EZ-IO in previous studies are contrasted by Schalk et al (2011) who found in an observational study that all participants rated the EZ-IO tibia 10/10 for satisfaction with its ease of insertion and handling.
Flow rates
In general, the more proximal the insertion site, the greater the flow rate. In the study by Hammer et al (2016), flow rates measured in unfixed cadavers, under gravity and then under pressure, were well matched in the EZ-IO tibia and humerus but surpassed by the FAST. However, this study was limited given the use of only three cadavers. The comparable rates of the EZ-IO humerus and tibia were replicated by Ong et al (2009) but again, this study was underpowered. Pasley et al (2014) found that in cadavers, flow rates were greatest in the sternum using the FAST, followed by the EZ-IO humerus, with a larger gap this time between the humerus and the tibia which had the lowest flow rate. A small observational study found that the proximal tibia had greater flow rates than the distal tibia when using the EZ-IO (Tan et al, 2012).
Cost
Using prices from a commercial UK-based supplier (SP Services, correct as of July 2016) the device costs are listed exclusive of VAT:
Discussion
The evidence described here was lacking in quality; there were no meta-analyses, and the RCTs (Brenner et al, 2008; Leidel et al, 2010; Derikx, 2014; Kurowski et al, 2014; Demir et al, 2016; Hammer et al, 2016; Szarpak, 2016) suffered from under-powering and methodological flaws that made direct comparison difficult. All but two articles were graded as 2b for this reason (according to the Oxford Centre for Evidence-Based Medicine (CBEM) grading system). No article was able to stand alone and influence practice. However, the findings from the overall review may inform a direction for change of practice and indicate a need for further research.
None of the trials used a truly random sample of the population, either to allocate the devices or select the participants. Instead, all authors used convenience sampling which, while being more practical, risks the sample not being truly representative of the population (Newell, 1996). All studies were limited by a lack of detailed reporting on how participants were randomised, although some authors declared the use of randomisation software or blind envelopes (Leidel et al, 2010; Kurowski et al, 2014; Szarpak et al, 2016). This suggests that reasonable effort was made to reduce introduction of bias at this phase.
No trial was blinded, likely because of the distinct nature of the interventions and lack of ability to blind investigators to which devices they were using. There would be the potential to introduce observer bias, which is a confounding factor in this kind of research. However, lack of disclosure of conflicts of interest goes some way to reducing the considered risk of bias. No information was given about blinding at the analysis stage.
Study power was poorly addressed in all RCTs; only one author reported calculating the participants needed for adequate power but failed to meet this target (Demir et al, 2016). The other trials didn't report power calculations but, between them, had an average of only 54 participants; suggesting a lack of statistical significance in the population sizes, given the only power calculation demanded 200 participants.
These studies are limited in their application to UK practice owing to small sample sizes and conflicting results. However, the findings offer a body of evidence that may inform practice as well as indicate a need for further study.
Success of insertion
The only clear conclusion that could be drawn from the results was that a semi-automatic device was more likely to be successful than a manual one. It can be suggested that an increased scope for user error was the cause for this finding. No other conclusion could reliably be drawn as the studies generally found limited differences in success rates between the semi-automatic devices and there was little consistency in the conclusions; with articles contradicting each other. This may have been because of a lack of adequately-powered studies or related to factors such as operator confidence or study design.
Time
It is difficult to draw an accurate conclusion about device insertion time as there are clear disparities in reporting across the studies. With results ranging from a few seconds to several minutes, it is clear that there were no consistent start or stop points.
Use of any device will require time spent on identifying landmarks for insertion. In addition, the manual devices were understandably longest to insert given the hand-driven nature. The EZ-IO may also suffer increased insertion times given the need to select and then match an appropriate needle to the driver. In practical terms, the spring-loaded devices should take the least time because of the lower step number during insertion; however, the evidence is inconclusive and this theory couldn't be supported. These studies don't indicate if any difference in the insertion times may be clinically significant; with approximately 2 minutes of variation across the devices, there may be little impact on clinical outcomes.
Ease of use
The manual device was rated poorest in the two studies that investigated it. Both studies that looked at the EZ-IO on the humeral site showed that it was rated poorly for ease of use, this may have been because of difficulties in identifying anatomical landmarks at the humeral head, but wasn't explicitly stated. The remaining studies compared the EZ-IO tibial site, the BIG, the FAST and the NIO favourably; with various articles showing different devices as easiest to use but the margins of difference were small. Selecting the correct needle for the EZ-IO may introduce an opportunity for confusion or error that doesn't impact other devices.
Flow rates
The sternum appears to have the greatest flow rates; this is as a result of the sternum's anatomy and proximity to the great vessels of the chest. The relatively small differences in results are unlikely to be clinically meaningful.
Cost
The manual device costs the least, as may be expected; this is followed by the spring-assisted BIG; NIO; and FAST in ascending order, with the EZ-IO driver alone significantly more costly than any other device, presumably owing to its more complex design. Cost will fluctuate based on supplier and bulk contracts but the examples given act as good baselines for unit cost.
Conclusion
This literature review identified that any semi-automatic IO access device is more likely to be successful and easier to use than a manual counterpart. The manual device also appears slowest to insert but it is not possible to make a definitive conclusion on which semi-automatic device is the fastest to insert given shortcomings in the evidence and the varied methods used.
The evidence supports that an insertion site closer to the heart (i.e. the sternum) will have the greater flow rate but whether the increased rate of flow here is clinically relevant can't be determined at this stage. Finally, in terms of device cost and given the lack of evidence to suggest superiority, the EZ-IO may not offer value for money as it is significantly more expensive compared with the spring-actuated devices. Further and larger randomised trials are needed to provide definitive answers to these questions.