A randomised trial of non-invasive cardiac output monitoring to guide haemodynamic optimisation in high risk patients undergoing urgent surgical repair of proximal femoral fractures (ClearNOF trial NCT02382185)

It is estimated that the number of patients that will sustain a hip fracture will reach 100,000 per annum by 2033 with a cost to the UK health services alone of some £ 2.7 billion (White & Griffiths, 2011). Outcomes for this group remain poor with a mortality of 6% at 1 month, increasing to 20–33% (Brauer et al., 2009; Hip Fracture Anaesthesia Sprint Audit Project, n.d.) at 1 year (The National Hip Fracture Database, 2016). Complication rates for this group also remain significant with between 20 and 60% of patients having significant post-operative complications (Roche et al., 2005; Bartha et al., 2013). These complications are not only associated with increased length of stay and healthcare costs, but also reduced long-term survival (Lawrence et al., 2002; Khuri et al., 2005).

Attempts to improve outcomes through fluid optimisation protocols have been mixed (Bartha et al., 2013; Sinclair et al., 1997; Venn et al., 2002) with the most recent work (Moppett et al., 2015) suggesting that fluid optimisation using an arterial waveform-based system did not reduce complications or mortality, although a recent meta-analysis shows a trend towards both. The authors suggested that the role of tighter arterial pressure control should be addressed as part of any future intervention. Given that the incidence of hypotension in the Anaesthesia Sprint Audit of Practice (ASAP) (White et al., 2016) was significant with 56% of patients undergoing neck of femur (NOF) repair having a systolic blood pressure of less than 90 mmHg, and one third having a mean arterial pressure (MAP) of less than 55 mmHg, and the association of hypotension with adverse events (Salmasi et al., 2017; Sun et al., 2015), this may be a valid treatment strategy in this group.

Within this group of elderly patients, a high-risk group exists. The Nottingham Hip Fracture Score (NHFS) is a validated scoring system that predicts patients at increased risk of both 30-day and 1-year mortality (Wiles et al., 2011; Maxwell et al., 2008). This high-risk group, with an increased mortality rate, may also have an increased complication rate, and hence any effective intervention may have a greater impact on these outcomes.

The aim of the study was to investigate the role of haemodynamic optimisation (stroke volume optimisation and blood pressure control) in high-risk patients defined by a NHFS ≥ 5 undergoing urgent surgical repair of proximal femoral fractures and the effects of this on complications. Secondary outcomes included the length of stay, the incidence of hypotension and the use of vasopressors and fluids.

We performed a multicentre randomised controlled observer-blinded trial.

This study was approved by the National Research Ethics Service (NRES) committee Yorkshire and the Humber Leeds West (14/YH/1170). Written informed consent was obtained by those individuals that held capacity; otherwise, if a patient met the inclusion criteria but lacked the capacity to consent, we followed the process outlined in the Mental Capacity Act and approved by the NRES committee. The trial was registered prior to patient enrolment at clinicaltrials.gov (NCT02382185).

Two hundred forty-one patients were recruited and randomised from November 2015 to September 2017. Consort flow chart for the study is shown in Fig. 1. Patient demographics, surgical and anaesthetic details are shown in Table 1.

Haemodynamic optimisation of high-risk patients undergoing proximal neck of femur fracture repair using non-invasive technology did not result in a reduction in post-operative complications or mortality. On the per-protocol analysis, a reduction in mean length of stay was seen.

Whilst the lack of effect is at odds with the original trials using goal-directed fluid therapy in this patient group (Sinclair et al., 1997; Venn et al., 2002), it mirrors the effects seen in the more recent trials where a clinically significant risk reduction was seen that failed to reach significance (Bartha et al., 2013; Moppett et al., 2015).

This trial differs from the more recent ones performed by Bartha and Moppett in various ways; firstly, we used the NHFS to identify those at higher risk of morbidity and mortality. The assumption was that any treatment effect seen from haemodynamic optimisation would be greater in this group. However, it may also be that due to the multiple comorbidities, poor cardiac function and poor outcomes in this group, the outcome is minimally modifiable. Secondly, we chose to enrol patients undergoing both spinal and general anaesthesia, whereas in the work by Moppett and colleagues, only patients undergoing spinal anaesthesia were included, and over 90% of surgery was performed under regional anaesthesia in the work by Bartha. Whilst there is no data that suggests that one form of anaesthesia is related to better outcomes in this patient group (White et al., 2016), it is a potential confounder when comparing the studies. This was stratified for, however, in the randomisation procedure. The inclusion of a population that had a mixture of general and neuraxial anaesthesia is more reflective of the practice seen in the UK and hence makes the results more generalisable. Thirdly, we chose to use a non-invasive measurement of blood pressure and stroke volume in the form of the ClearSight™ system as opposed to an arterial line based system that was used in the 2 recent trials. This approach was chosen as invasive monitoring is rarely used in this patient group and would represent a significant change from practice. The ClearSight™ technology provides the same information from a finger cuff and hence may have greater uptake by the profession as opposed to the increased usage of invasive monitoring. Whilst both systems utilise arterial waveform analysis, the ClearSight™ system reconstructs this from a finger artery waveform and uses pulse contour analysis rather than pulse power analysis as seen in the LiDCO systems. The technology failure rate of 15% in this study is of significance and must be bore in mind by future researchers, with the majority of failure due to poor signal acquisition due to cold peripheries and osteoarthritic fingers. On the per-protocol analysis, a reduction in LOS was seen, confirmed by the meta-analysis, and hence if a suitable technology could be made to work effectively in this group, then there may be a benefit. Finally, and similarly to the work by Bartha, we chose to include a step in the intervention arm that aimed to control for blood pressure. Hypotension has been shown to be associated with acute kidney injury and myocardial injury after non-cardiac surgery (MINS), and a value below 65 mmHg has been suggested as putting individuals at a higher risk of these events (Salmasi et al., 2017; Sun et al., 2015). In addition, hypotension has been associated with harm in this specific patient group (White et al., 2016). We chose to use a treatment value of below 30% from baseline, which has a similar relative risk of injury (Salmasi et al., 2017). Whilst the incidence and duration of hypotension was less in the intervention group, there was no statistical significance in the AUC for MAP below 30% of baseline [0.35 (0.36) vs. 0.27 (0.29), p = 0.067] representing a protocol compliance failure for the intervention group overall. Despite this, there was no association between any degree or magnitude of hypotension and outcome in this trial, and the incidence of complications, including kidney injury or cardiovascular complications, was not different between the groups. However, MINS was not specifically screened for and this study was not designed to detect an outcome difference due to variations in MAP. The currently accepted value of 65 mmHg is also based on population data that does not include the group studied in this trial who have a significant incidence of hypertension and in perhaps whom a higher cutoff value relating to outcomes may be seen in this population.

The complication rates in this trial were similar to those seen in the more recent trials. Bartha, who also treated hypotension as part of the protocol, reported a RR of 0.79 (95% CI 0.54–1.16), similar to the OR of 0.73 (95% CI 0.43–1.24) in this trial, whilst Moppett who did not control for baseline blood pressure showed RR of 0.91 (95% CI 0.64–1.28) for patients developing one or more complications. Whilst our groups were well matched in terms of comorbidities, there was a significant difference in stroke volume, with the intervention group having a lower SV at the start of surgery. This may reflect poorer underlying cardiovascular function, and hence this group may have been more compromised compared to the control group leading to a reduced treatment effect. Fluid volumes given were also different between the trials with the intervention group receiving 875 ml of crystalloid compared to 1500 ml in the work by Moppett and colleagues and 1078 ml in the Bartha trial. A concern is that this may relate to an increased use of vasopressor. The use of vasopressors is not without risk, and pure alpha agonists such as phenylephrine are associated with a reduction in cardiac output and cerebral blood flow (Cannesson et al., 2012; Ogoh et al., 2011), depending on preload responsiveness, which may translate into worse outcomes. However, we showed no difference in the amount given between the groups.

A reduction in mean and not median length of stay was detected in patients treated on a per-protocol basis. Whilst this analysis was defined a priori, caution must be used in the interpretation of this result as the trial was not powered to detect this and the effect size is borderline. The mean length of stay is more reflective of the economic impact that an intervention may have on a population, and indeed if this effect can be repeated then due to the magnitude of reduction seen, this would have a significant impact for the NHS. In addition, there was a noticeable reduction in the upper quartile limit for the median stay in the intervention group from 32 to 22 days in keeping with a mean reduction. It may well be that these metrics represent a reduction in the variation of care, which in itself has been suggested to lead to improved outcomes.

An updated meta-analysis showed a significant effect on morbidity with a risk ratio of 0.81 (0.67–0.98) whilst the effects on mortality remained non-significant and unchanged. It is interesting to note that none of the 4 studies included have suggested harm from the intervention; however, given the potential impact that even a small reduction in either LOS or complications would have on such a large population, and the signal seen in meta-analysis, then a large pragmatically designed multicentre trial that addresses both fluid management and blood pressure control is now warranted.

The study has limitations: primarily in that it is underpowered. We expected a complication rate of 60% in the control group and this was only 51%. This may reflect temporal improvements in perioperative care based on guidelines from groups such as the National Institute for Healthcare and Excellence, the National Hip Fracture Database and the Association of Anaesthetists of Great Britain and Ireland amongst others, and additionally, the trial was conducted over the best part of 2 years and there is a chance of a significant Hawthorne effect. Despite this, it remains the largest randomised controlled trial of haemodynamic optimisation of hip fracture patients and adds to the literature and updated meta-analysis. We chose to use the same basket of complications as those used in the Bartha trial in order that direct comparison may be made; however, it may be that these complications are too generic and more procedure specific and patient-relevant end points are needed. Finally, protocol compliance may have been suboptimal as there was no significant difference in patients with a MAP of less than 30% of baseline and a relatively small increase in SV. This may reflect the relatively short nature of the surgery and hence a limited opportunity of optimising patients.

The strengths of the study include a homogenous population, the inclusion of MAP control in the algorithm, the exclusion of lower risk patients and the fact that the intervention was performed in the context of best practice for fractured neck of femur care.

The updated meta-analysis suggests a possible reduction in complications, and the relative reduction in complications appears consistent throughout the three more recent trials. Given the significant and growing volume of surgery undertaken, and the economic impact on health care, larger studies should address a treatment strategy where both fluid and blood pressure management are addressed.