SEARCH

SEARCH BY CITATION

Summary

  1. Top of page
  2. Summary
  3. Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. Competing interests
  8. References

We performed an evaluation of non-Luer spinal devices supplied by four manufacturers or suppliers: Polymedic; Pajunk; Sarstedt; and Smiths. For each supplier, 100 evaluations were performed using a 25-G 90-mm spinal needle, 3-ml syringe, 5-ml syringe and filter needle; for comparison, 100 evaluations were performed with our standard Luer equipment. The non-Luer devices were associated with more qualitative problems compared with the Luer devices, for example, poor feel of dural puncture (9–32% vs 10%, respectively), poor observation of cerebrospinal fluid in the hub (3–27% vs 0%), and connection problem of the syringe to the spinal needle (7–33% vs 0%). There was also more frequent failure to achieve the spinal injection due to equipment-related causes (4–7% vs 0%, respectively). Median (IQR [range]) numeric satisfaction scores for the spinal needles were: Luer 10 (9–10 [7–10]); Polymedic 7 (4–8 [0–10]; Pajunk 7 (5–8 [0–10]); Sarstedt 7 (6–8 [0–10]); and Smiths 9 (7–10 [0–10]) (p < 0.0001). Satisfaction scores for all spinal equipment were: Luer 10 (9–10 [5–10]); Polymedic 8 (6–8 [0–10]); Pajunk 7 (5–7 [1–9]); Sarstedt 8 (6–8 [0–10]); and Smiths 8 (8–9 [2–10]) (p < 0.0001). Between 21% and 75% of non-Luer evaluations were rated with satisfaction worse than the usual Luer needle compared with 0–10% rated better, depending on the needle type. Between 22% and 76% of non-Luer evaluations were rated with satisfaction worse than the usual Luer equipment compared with 0–14% rated better. Specific concerns included poor feel of tissue planes and observation of cerebrospinal fluid (Polymedic), difficulty with connection of the syringe to the spinal needle and trocar removal (Pajunk), poor feel of tissue planes and needle flexibility (Sarstedt) and difficulty with connection of the syringe to the spinal needle (Smiths). We could not demonstrate a short-term learning curve for the new devices. Decisions on purchasing and implementation of the new non-Luer equipment will have to acknowledge that clinicians may have greater technical problems and reduced satisfaction compared with the current equipment.

In November 2009, the National Patient Safety Agency (NPSA) in the UK issued two Patient Safety Alerts. These required NHS organisations to change spinal, epidural and regional anaesthesia equipment so that the connectors could not attach to intravenous (Luer) equipment. The target date for implementation was 1 April 2011 for Part A (spinal equipment) and 1 April 2013 for Part B (epidural and regional anaesthesia equipment). Part A was updated in January 2011 with a new implementation date of 1 April, 2012 [1]. Similar legislation is also intended in California, USA.

A similar Patient Safety Alert was issued in 2007 requiring an alteration to the dimension of enteral feed connectors so that they would not connect to Luer equipment [2]. Although successfully implemented, the coordination of the implementation posed problems. The change to non-Luer spinal needles is likely to pose a greater challenge. Besides the logistical issues related to the changeover of a large number of device items in multiple clinical areas, the design of the equipment is crucial as it not just a different connection between components of otherwise familiar equipment; many suppliers are also using newly designed syringes. Unlike the connection between enteral feed components, spinal anaesthesia involves considerable dexterity while reinserting the trocar into a fine bore needle or attaching the injectate syringe while precisely maintaining the needle tip’s position. Furthermore, ∼40% of spinal anaesthetics are performed for obstetric procedures [3], with a significant minority performed in extremely urgent cases.

There is little published information on the non-Luer spinal equipment that has been developed in response to the NPSA Patent Safety Alert. Some devices have been studied in the laboratory or on manikins, including the InterVene Spinalok® (InterVene Ltd, Chesterfield, UK), the B-Link Neurax® (B-Link (UK) Ltd, Leeds, UK) [4, 5] and the BD UniVia-6 (Becton Dickinson UK Ltd, Oxford, UK) systems [6]. Lawton et al. performed >350 clinical evaluations of the B-Link Neurax system in five trials [5]. Bristol and Weston NHS Purchasing Consortium buys equipment for both large acute healthcare Trusts in Bristol – University Hospitals Bristol NHS Foundation Trust and North Bristol NHS Trust. We wished to assess the clinical performance of a wide range of new spinal needles, with the aim of introducing a single design into both Trusts before 1 April 2012.

Methods

  1. Top of page
  2. Summary
  3. Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. Competing interests
  8. References

This clinical evaluation was registered with the University Hospitals Bristol and North Bristol Trust Audit Committees. The Local Research Ethics Committee (LREC) determined that this should be considered a service evaluation as the change to non-Luer equipment is an externally mandated process with a fixed deadline, and, therefore, there was no requirement to gain consent from patients. It was carried out in University Hospitals Bristol and at Southmead Hospital. The purchasing specialist for both Trusts organised the supply of equipment for the clinical evaluations. Suppliers were asked to provide equipment at a price equivalent to the current purchase cost for the Luer equipment, although in a number of cases some of the items were given free of charge.

We evaluated non-Luer needles from Polymedic (manufactured by TeMeNa, Carrières-sur-Seine, France; supplied in the UK by Polymedic Medical Ltd, Southampton, UK), Pajunk (Pajunk UK Medical Products Ltd, Prestwick, Northumberland, UK), Sarstedt (Sarstedt Ltd, Leicester, UK) and Smiths (Smiths Medical, Ashford, Kent, UK). Polymedic, Pajunk and Sarstedt use the Surety® connector designed by InterVene Ltd together with syringes manufactured by InterVene Ltd. Smiths use the Portex® CorrectInjectTM connector with their own syringes. Equipment was evaluated on a like-for-like basis. Our standard equipment is a 25-G 90-mm spinal needle with a non-cutting tip (Vygon [UK] Ltd, Swindon, UK at Southmead Hospital and Smiths Medical, Ashford, Kent, UK at University Hospitals Bristol) together with a 3-ml lock syringe (Plastipak; Becton Dickinson UK Ltd). For each non-Luer evaluation, we specified a 25-G spinal needle with a non-cutting tip, 3-ml and 5-ml lock syringes and a filter needle, in addition to Luer equipment for skin infiltration and other components for the sterile procedure. All spinal needles were supplied with an introducer needle. The standard procedure for the anaesthetic drug preparation was to draw up hyperbaric bupivacaine into the 3-ml syringe, draw up saline into the 5-ml syringe and dissolve and draw up powdered diamorphine from an ampoule, and then draw up diamorphine solution from the 5-ml syringe into the 3-ml syringe using the same filter needle [7]. This method ensured that the filter needle dead space was accounted for.

Evaluations were performed between April 2011 and March 2012 by anaesthetists providing the clinical service, all of whom had at least one year’s experience of anaesthesia. A full explanation was provided before the first use of non-Luer equipment. Each non-Luer evaluation series was performed separately, lasting ∼2 months. We evaluated 100 procedures for each non-Luer supplier. These were carried out in daylight hours on cases that were predicted to be straightforward clinically. Patients with spinal abnormalities, extreme obesity, a history of difficulty with regional anaesthesia or clinical urgency were excluded. The position of the patient during spinal insertion was chosen by the anaesthetist according to clinical preference. The majority of spinal anaesthetics were performed for caesarean section, although there were some non-caesarean and non-obstetric cases included. One hundred evaluations using Luer equipment were performed on patients having similar surgery, although there was no exclusion of difficult cases.

We used a standard evaluation form [8]. There were ‘yes/no’ answers for: good feel of dural puncture; trocar easy to remove; free aspiration of cerebrospinal fluid (CSF); easy to see CSF in hub; leak of injectate; connection problem of syringe to spinal needle; disconnection problem of syringe to spinal needle; needle movement related to connection or disconnection; needle bent; and failure to achieve procedure. Other items recorded were the estimated time from trocar withdrawal until observation of CSF and the number of needles used. Numeric rating scale scores were recorded for satisfaction with the needle and for satisfaction with all equipment. Categorical scores were recorded for satisfaction with the needle and with all equipment in comparison with the usual (Luer) needle. Data were compared using Wilcoxon signed rank test (Minitab, State College, PA, USA).

Results

  1. Top of page
  2. Summary
  3. Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. Competing interests
  8. References

Sixty-seven, 20 and 13 Luer evaluations were performed by trainee, consultant and non-consultant specialist doctors, respectively. The corresponding figures for non-Luer evaluations were 178 (45%), 180 (45%) and 42 (11%), respectively. The qualitative data and comparative satisfaction scores are shown in Table 1. Figures 1 and 2 show the numeric satisfaction scores for the needle and all the spinal equipment. There were significant differences (p < 0.0001) between needles and between all spinal equipment for both the comparative (vs Luer) and absolute satisfaction scores.

Table 1.   Equipment characteristics during needle evaluations and satisfaction compared with Luer devices during non-Luer evaluations. There were 100 evaluations per supplier; the denominator is < 100 if evaluations were excluded (for example, if there was a failure to achieve injection). Values are number/denominator (proportion), number (proportion) or median (IQR [range]).
 LuerPolymedicPajunkSarstedtSmiths
  1. There were significant differences in satisfaction ratings between needles (p < 0.0001) and between all spinal equipment (p < 0.0001).

  2. CSF, cerebrospinal fluid.

Good feel of dural puncture90/100 (90%)79/94 (84%)65/95 (68%)72/96 (75%)89/98 (91%)
Trocar easy to remove99/100 (99%)88/98 (90%)64/99 (65%)87/99 (88%)98/99 (99%)
Free aspiration99/100 (99%)78/88 (89%)86/91 (95%)81/86 (94%)87/92 (94%)
Easy to see CSF in hub100/100 (100%)70/96 (73%)89/96 (93%)91/96 (95%)94/97 (97%)
Leak of injectate1/100 (1%)1/93 (1%)8/95 (8%)5/96 (5%)1/96 (1%)
Connection problem0/100 (0%)12/94 (13%)31/95 (33%)10/96 (10%)7/96 (7%)
Disconnection problem0/100 (0%)1/94 (1%)2/94 (2%)0/96 (0%)0/96 (0%)
Needle movement related to connection/disconnection0/100 (0%)6/94 (6%)8/95 (8%)2/96 (2%)4/96 (4%)
Needle bent1/100 (1%)2/100 (2%)11/100 (11%)13/100 (13%)3/100 (3%)
Failure to achieve procedure with this needle0/100 (0%)9/100 (9%)5/100 (5%)8/100 (8%)9/100 (9%)
Time from trocar withdrawal until CSF seen; s1 (0–1 [0–5])2 (2–3 [0–12])1 (0–2 [0–6])1 (0–1.5 [0–10])1 (0–1 [0–5])
Delay in appearance of CSF (≥4 s)1/91 (1%)13/90 (14%)10/91 (11%)12/94 (13%)3/97 (3%)
Number of needles used
 196 (96%)96 (96%)86 (86%)92 (92%)95 (95%)
 24 (4%)4 (4%)12 (12%)8 (8%)4 (4%)
 ≥32 (2%)1 (1%)
Satisfaction compared with usual spinal needle
 Significantly better 0 (0%)0 (0%)0 (0%)1 (1%)
 Better 5 (5%)0 (0%)7 (7%)9 (9%)
 Same 32 (32%)23 (24%)41 (41%)69 (69%)
 Worse 43 (43%)55 (58%)45 (45%)19 (19%)
 Significantly worse 19 (19%)16 (17%)7 (7%)2 (2%)
 No comment 1600
Satisfaction compared with usual spinal equipment
 Significantly better 0 (0%)0 (0%)2 (2%)1 (2%)
 Better 4 (4%)0 (0%)11 (11%)12 (12%)
 Same 42 (45%)22 (24%)44 (44%)64 (64%)
 Worse 38 (40%)60 (64%)39 (39%)22 (22%)
 Significantly worse 12 (13%)11 (12%)3 (3%)0 (0%)
 No comment 4711
image

Figure 1.  Numeric satisfaction scores for spinal needles. LU, Luer; PO, Polymedic; PA, Pajunk; SA, Sarstedt; SM, Smiths. Values are median (line), IQR (box) and range (whiskers).

Download figure to PowerPoint

image

Figure 2.  Numeric satisfaction scores for all spinal equipment. LU, Luer; PO, Polymedic; PA, Pajunk; SA, Sarstedt; SM, Smiths. Values are median (line), IQR (box) and range (whiskers).

Download figure to PowerPoint

There was no difference in the estimated median time from withdrawal of the trocar from the spinal needle until observation of CSF in the hub of the needle for different needle types in the sitting and lying positions. However, delayed appearance of CSF (defined as estimated time to appearance ≥ 4 s) occurred in 1/57 and 0/35 cases for Luer, 7/58 and 6/38 for Polymedic, 8/44 and 1/54 for Pajunk, 9/57 and 3/41 for Sarstedt, and 1/59 and 1/37 for Smiths in the lying and sitting positions, respectively. The position was not recorded in one Smiths case where CSF appeared after 5 s.

To gauge the effect of learning and familiarisation with the equipment, we counted the number of abnormal responses to the nine ‘yes/no’ questions on the evaluation form. There was no consistent change in the sum of abnormal responses or the VAS score from the first to the last evaluation when an individual anaesthetist used a particular needle on five or more occasions.

A summary of free text comments that were included on the evaluation forms is shown in Table 2. Positive comments related primarily to the spinal needle and also to the syringe–spinal needle connection. The negative comments on Polymedic related to the spinal needle, connection, introducer needle or other in 100, 12, 1 and 5 cases, respectively. The corresponding numbers for the other manufacturers were 87, 68, 3 and 5 for Pajunk; 74, 11, 4 and 2 for Sarstedt; and 10, 10, 0 and 3 for Smiths.

Table 2.   Summary of free-text comments. The five most common comments for each needle are included. Values are number.
NeedlePositive comments Negative comments 
  1. CSF, cerebrospinal fluid

LuerTotal0Total3
Dissatisfaction with introducer needle2
Delay in CSF flashback1
PolymedicTotal27Total118
Satisfaction with syringe to spinal needle connection12Delay in CSF flashback26
Good feel of tissues8Poor/altered feel of tissues24
Good needle6Difficult to see CSF15
Good connection1Difficulty connecting syringe to spinal needle7
  Difficulty removing/reinserting trocar7
PajunkTotal2Total163
Satisfaction with connection1Difficulty removing/reinserting trocar40
Good feel of dural puncture1Difficulty connecting syringe to spinal needle30
  Concern regarding security of connection19
  Needle too flexible/bent17
  Poor/altered feel of tissues16
SarstedtTotal38Total91
Liked the flat section of needle/improves control13Needle too flexible/bent26
Easy to see CSF9Poor/altered feel of tissues19
Nice needle5Difficult to remove/reinsert trocar12
Good feel of dural puncture3Introducer needle too short4
Satisfied with connection3Difficult to connect syringe to needle4
SmithsTotal5Total22
Satisfaction with needle5Difficulty connecting syringe to spinal needle10
  More force to withdraw trocar4
  Poor/altered feel of tissues4
  Air entrainment in syringe3
  Needle too flexible/bent2

There were 31 cases of failure to perform the anaesthetic procedure with the non-Luer needle (Table 3). There were no reports of patient harm.

Table 3.   Detailed comments after failure to achieve spinal procedure.
  Technically satisfactory, but block failedChange to LuerPoor feelOther
  1. CSF, cerebrospinal fluid.

Polymedic1   
2  Patient moved
3  Clumsy plastic on thread of non-luer connector. Unable to twist syringe onto needle; needle moved; unable to aspirate CSF
4   
5  
6   
7 Stylet not well fixed and kept popping out
8[no comments]   
9  
Pajunk1   Difficult to re-insert trocar
2  Needle bent. Difficult to remove trocar
3 Difficult to reinsert trocar
4  
5  Difficult to remove trocar; abandoned procedure as did not have spare syringe to flush needle to determine if blocked
Sarstedt1  Prefer this make; I like the flattened part; easy to hold; easy to see CSF; block only went to T7 therefore epidural inserted 20 min after spinal
2  Flimsy needle.
3  Slow flow of CSF; pain on injection of spinal anaesthetic
4  Unsuccessful placement due to undiagnosed scoliosis; ultrasound used to successfully place (Luer) needle
5   Introducer too short
6 Flexible needle ++; needle bent on bone
7  CSF easy to see but needed to rotate needle before CSF came back, then immediate; unable to aspirate at end, block slow to come up, epidural sited, not really equipment-related
8 Patient felt faint on first pass
Smiths1  Good aspiration at beginning and end, but incomplete block; spinal repeated
2  
3  Initial CSF flashback then unable to aspirate CSF, repositioned and bloody tap; two further needles used - both bloody taps, therefore abandoned
4  Block inadequate, therefore epidural sited; may of course have happened with any other needle
5  First pass good dural pop, easy aspiration, but needed epidural as block left > right
6  Attempted at two interspaces in lateral position then sitting; two trial needles used
7   
8  Trocar tends to fall out of needle hub; difficult spinal; multiple contacts with bone, eventually sited, but no change in (preceding epidural) block, therefore general anaesthetic
9  Two ‘dry taps’, changed to normal set and had a very slow flow of CSF - may be patient-/operator-related and not equipment problem

Discussion

  1. Top of page
  2. Summary
  3. Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. Competing interests
  8. References

This work was undertaken to allow clinically informed purchasing decisions for our local hospitals. We wished to compare non-Luer spinal devices from as many suppliers as possible; in the event, time constraints limited this to just four. Although it might be argued that comparisons with the standard Luer equipment are not valid, we believe that it is important to put any problems or successes of new devices into the context of our present service. The equipment that was evaluated was found to be less satisfactory than the current Luer equipment, but it is clear that this is the beginning of a development process that has been going on for many decades in the case of the Luer connector.

Conducting the assessment in the format of a service evaluation was supported by our LREC, as well as by the NPSA (David Cousins, personal communication). Ideally, we would have designed an evaluation in which all equipment designs were tested in random order. This was not possible, as, even at the time of writing, some companies do not stock our required components. We usually use a 3-ml Luer-lock syringe for a standard injectate volume of 2.8 ml at caesarean section. The alternative 5-ml syringe is adequate for virtually all other anaesthetic spinal doses including those for non-obstetric practice. The other decision in the design of the evaluation was to include all anaesthetists providing the service, rather than restricting it to a core group of senior practitioners. This not only allowed a faster evaluation but also reflected the reality of medical practice where non-consultant anaesthetists make a large contribution to the workload.

The new devices were associated with more technical problems than the Luer equipment used routinely in our hospitals. The initial part of the spinal process requires tactile identification of a dural ‘pop’ by the needle tip and rapid observation of CSF. Identification of CSF was slower with the Surety non-Luer needles, and there were a number of comments in the free-text responses about delay or difficulty in seeing CSF with the Polymedic needle. We used a threshold of ≥ 4 s to define delayed CSF appearance. This was more frequent in the lying position for Pajunk and Sarstedt, but there was no positional influence for the other needles. Scott and Sharpe found differences in the flow of fluid between Luer and non-Luer needles and between the lateral and sitting positions in a simulator study [9]. The mean times for fluid to appear in the hub of the needle varied from 5.8 to 25.2 s in the groups that they studied. Their model clearly measures a different parameter from the one that we used (time from dural puncture until CSF seen), as the median time in our investigation ranged between 1 and 2 s in the different groups. We acknowledge that we estimated rather than measured the time till appearance of CSF, so this measurement lacks precision, and further clinical studies are required.

Leak of fluid from the connection between syringe and spinal needle is clearly important for low volume injections such as during spinal anaesthesia or chemotherapy. Countrywide, a significant proportion of anaesthetists use Luer-slip syringes for spinal injection, arguing that the spinal needle is less likely to move when the syringe is attached. The argument for the Luer-lock syringe is the reduced chance of leakage of injectate. We used lock syringes for both the Luer and non-Luer evaluations. A leak was found in 1% of Luer evaluations and between 1% and 8% of non-Luer evaluations. In no case of leak was a failure to achieve a satisfactory block recorded. In contrast, during informal evaluations using slip syringes that were not included in the preceding data, there were leaks on three out of six occasions using Luer equipment and on three out of eight occasions using a non-Luer system.

The failure rate during the evaluations requires comment. Failure may be defined in several ways; for instance, it may describe failure of spinal anaesthesia itself, with the requirement to perform a different anaesthetic such as an epidural or general anaesthetic. This type of failure attributed to equipment-related problems is currently rare [10], although design attributes of a needle may cause an unanticipated increase in failure only borne out by large-scale use [11]. There were no such cases in the Luer evaluations. In nine non-Luer evaluations, the anaesthetist was satisfied with the equipment, but an inadequate block followed. In 16 cases of failure, a ‘softer’ endpoint of completing the procedure with a standard Luer needle was documented, reflecting the anaesthetist’s unhappiness in continuing with the test needle. Unless novices perform a trial, there will always be preceding familiarity with a particular equipment design. In our case, we normally use 25-G spinal needles from Vygon and Smiths. In addition, 24-G spinal needles from Pajunk are also sometimes used at Southmead Hospital. We cannot exclude some bias related to unfamiliarity, but believe that this is a minor effect as the occasions when the anaesthetist changed to standard Luer equipment were evenly distributed between the four non-Luer needle types. Clearly, during the Luer evaluations, there was no option of change to a more familiar item of equipment in the event of difficulty.

The number of instances where more than one needle was used was greater with the Pajunk needle; more than one needle was required on 14 occasions. This often related to adverse comments on the flexibility of the needle. The data on cases requiring more than one needle are usually not readily available, but we believe that it is important for costing purposes. To achieve 100 procedures during different needle evaluations, we used between 104 and 118 needles (not including cases where a switch to a standard Luer needle was made). Although such data will have to be collected within individual units, it is clear that the unit purchase price for equipment is not the only factor in the overall cost of performing a procedure.

We produced separate satisfaction scores for the spinal needle alone and for all the equipment (needle and syringe and other equipment). The evaluation forms had to be sufficiently simple to be used in clinical practice, and therefore detailed ratings for each component were not feasible; furthermore, when assessing a connection between two components, it is their interaction that is the relevant issue. Although we assessed four different non-Luer needles, there were only two non-Luer connector types. In the event, satisfaction ratings for the needle and all equipment were closely correlated, but this could not be assumed in advance.

We found that the non-Luer systems scored less well than the Luer equipment, both on itemised difficulties and subjective satisfaction scores. We are confident that this is not a ‘learning curve’ effect. There was no consistent pattern of improvement in satisfaction scores for needles or equipment, or the number of items where there was a problem. There were particular problems that featured with different needles as shown in Table 2. Furthermore, some anaesthetists used all three Surety systems, but there was no consistent improvement in the equipment satisfaction scores, although these systems use the same syringes and connector design. With each individual anaesthetist performing only 5–10 evaluations, it may be argued that it would be difficult to detect a more prolonged learning curve. However, the scale of change in clinical practice is unprecedented. There are approximately 12 500 anaesthetists in the UK, most of whom will start using non-Luer equipment over the next 1–2 years, as well as a large number of other physicians performing spinal procedures. The number of practitioners involved means that a slow learning curve might have a significant clinical impact. Moreover, even if a hospital changes to one supplier of spinal equipment in 2012, there may be further changes when Part B concerning epidural equipment comes into effect [1]. We must assume that the time required for established anaesthetists and physicians to become familiar with non-Luer equipment will be several years.

A survey of anaesthetists showed a desire for the same connector to be used for both spinals and epidurals within a hospital, even at the expense of their own personal preference for equipment [12]. There is considerably more complexity regarding epidural infusion administration equipment and potential incompatibility between local anaesthetic containers, pump giving sets and the epidural catheter than there is for spinal equipment [13]. The solutions to the implementation of Part B, and the possible alignment with spinal equipment, are yet to be clarified.

In summary, we have found an increased frequency of technical problems with the non-Luer equipment that we assessed. These related to both the needle performance and the new connectors. We could not demonstrate a short-term learning curve for the new devices.

Acknowledgements

  1. Top of page
  2. Summary
  3. Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. Competing interests
  8. References

We thank the manufacturers for their co-operation and help. We also thank the operating theatre staff and anaesthetists who facilitated or participated in this study.

Competing interests

  1. Top of page
  2. Summary
  3. Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. Competing interests
  8. References

Some of the equipment for these evaluations was supplied at below-market price or free of charge. There was no other external funding. No competing interests declared.

References

  1. Top of page
  2. Summary
  3. Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. Competing interests
  8. References