Type of anaesthesia for hip fracture surgery – the problems of trial design

The UK National Institute for Health and Clinical Excellence (NICE) has recently proposed that a randomised controlled trial (RCT) be performed to assess the clinical and cost effectiveness of regional versus general anaesthesia on postoperative morbidity in patients undergoing surgery for hip fracture [1]. Such a recommendation is likely to stimulate large research grant applications to the National Institute for Health Research (NIHR) in future.

To date, strong evidence for the clinical and cost benefit of one mode of anaesthesia compared with the other has been minimal, even though this would seem to be an important research question to answer, given the current and projected increases in postoperative mortality, bed occupancy and cost associated with hip fracture within an aging population [2].

In theory, an RCT should be relatively straightforward to design: randomly assign a suitably large number of patients requiring hip fracture surgery to receive either general or regional anaesthesia, and measure suitable cost and clinical outcomes to determine whether there are statistical differences between the groups. Comparative trials of anaesthesia in other types of surgery have been completed before [3], but the fact that an RCT involving hip fracture patients has not been recently attempted is informative, and indicates that the design of such an RCT is considerably more complex than it appears at first sight. Furthermore, given the current evidence base, it may not even be possible to consider the definitive RCT for a number of years until preparatory trials have been concluded; in the current financial climate, grant-awarding bodies might be better advised to spend their money on programmes of research, rather than a single ‘headline’ RCT.

Four main problems exist in trying to design such an RCT, solutions to which are not immediately (if at all) apparent. It is interesting to note that other researchers have identified similar problems in the design of trials of general versus regional anaesthesia [4]. Firstly, equipoise (i.e. genuine uncertainty about which mode of anaesthesia is ‘best’) is exceedingly unlikely, potentially introducing recording and reporting bias. Equipoise for such a trial needs to be both scientific and clinical. Meta-analyses [5, 6] and guidelines [7, 8] have previously suggested a marginal superiority of regional anaesthesia over general anaesthesia in terms of early mortality and acute postoperative confusion. Given the paucity of data to suggest that one technique is convincingly better, there is probably scientific equipoise overall.

Clinical equipoise is harder to achieve. Although approximately half of patients nationally receive general anaesthesia and half regional anaesthesia, institutional bias towards one type of anaesthesia is prevalent [9, 10]. Furthermore, there are a myriad of patient, context and operator factors that influence the decision whether to administer general or regional anaesthesia. These would render randomised intervention unethical if it was contrary to what the anaesthetist considered to be in the patient’s best interests, and highly problematic in terms of recruitment or retention of participants.

Controlling for patient variability through extensive inclusion or exclusion criteria would also be likely to limit the applicability of results; for example, about 4% of hip fracture patients taking oral anticoagulants would normally receive general rather than regional anaesthesia; by excluding such patients (who might be expected to have underlying ischaemic or valvular heart disease) from study enrolment, measured outcomes for the general anaesthesia study arm might be improved. Similarly, general anaesthesia outcomes could be improved by excluding the ∼8% of hip fracture patients who have moderate or severe aortic stenosis from enrolment, for whom regional anaesthesia is relatively contraindicated [11]. Another example concerns the age of participants included, as increasing age is associated with increased early and late mortality after restorative surgery [12, 13] and possibly more detrimental cerebral effects after exposure to general anaesthesia [14].

Secondly, there are very real problems in determining which primary and secondary outcomes should be measured. Thirty-day postoperative mortality is commonly used and has the advantage of being easy to define and measure, but previous studies have failed to find any difference associated with anaesthetic technique [6], and up to 50% of mortality may not be preventable by modification of treatment [15]. This is not surprising given the temporal disconnection between intervention (anaesthesia) and outcome (death within thirty days of anaesthesia), a problem highlighted recently in the FOCUS study, in which – unsurprisingly – the administration of a mean additional 1.2 units of blood after hip fracture surgery based on a liberal or restrictive transfusion threshold was not associated with a difference in mortality at sixty days [16]. Five-day postoperative mortality is more temporally related to anaesthesia, but this outcome occurs rarely (< 2%) and would require a very large sample size to detect any significant difference between techniques. Similarly, length of hospital stay is a defined, easy-to-measure endpoint, but again is temporally disconnected from the administration of anaesthesia, and subject to organisational factors.

Although we consider that death and length of stay are important secondary outcomes, we suggest that there should be greater emphasis placed on measuring functional recovery amongst survivors as the primary outcome, but this too is problematic, given the dearth of validated measures of recovery. Even then, measures such as return to mobility [17], return to home [18] and postoperative delirium [19] are likely to be influenced by non-anaesthetic, organisational factors such as availability of orthogeriatricians, nurses and physiotherapists, and also tend to focus on only one aspect of recovery. Indeed, the functional outcome of primary importance is probably yet to be developed, and would require the validation of a composite score of recovery comprising both objective (for example, analgesia requirements, return to mobility, and confusion) and subjective (patient reported) measures. Indeed, to our knowledge, no research has been published asking patients exactly which outcomes are important to them or their relatives/carers after anaesthesia for hip fracture surgery. Furthermore, any outcome measures would need to be adjusted for casemix, as anecdotal evidence suggests that regional anaesthesia is preferred for sicker patients; scoring systems are available, including the ASA and POSSUM scores, but these appear to be less predictive of postoperative mortality than the Nottingham Hip Fracture Score [12], which itself requires multicentre validation before it is recommended for general use.

Thirdly, ‘general’ and ‘regional’ in the context of anaesthesia for hip fracture surgery define a vast array of techniques that are known to be used by individual anaesthetists, and are consequent to the lack of evidence for any ‘best’ general or regional anaesthetic [9]. Recent AAGBI guidance suggests that the mode of anaesthesia is less important than the manner in which it is delivered, sympathetically to patients’ age and comorbidities [8]. In order to detect a 1% difference in 30-day mortality between regional and general groups (α = 0.05, power 80%), we estimate that approximately 11 000 patients would be required per group, which would certainly allow for subgroup analysis of outcome by specific type of anaesthesia. An alternative would be to standardise the anaesthetic technique, requiring the enrolment of many fewer patients. However, as indicated above, anaesthesia for hip fracture invariably has to be tailored to individual patients’ needs, depending on comorbidities and contraindications. Even then, no standardised best technique of general or regional anaesthesia has yet been identified: for example, it remains unclear which local anaesthetic, of what baricity, concentration and volume [20], and with which additives, provides the most cardiostable combination for regional anaesthesia (if indeed, cardiostability is what is most desirable in a regional anaesthetic technique). Similarly, the contribution to outcome of contemporaneous anaesthetic management of, for example, temperature, analgesia, monitoring, hypotension and fluid management are still unclear.

In addition to those already highlighted, there are further recruitment issues that would hamper the design of an RCT, relating to the inclusion or exclusion of patients with cognitive impairment in light of the Mental Capacity Act 2005 [21]. The Act indicates that research should not involve patients without mental capacity if it can be adequately performed involving participants with capacity, and strictly applied, this would exclude the 25–40% hip fracture patients with pre-operative cognitive impairment. Anecdotally, recruitment problems would be significant – at Peterborough, a small, randomised trial of the effect of anaesthesia technique on outcome has only been able to recruit ∼15% of patients admitted to the hospital with hip fracture over five years. Applied nationally, within a two-year study period, 3000 patients per treatment limb (as suggested by NICE) would require the involvement of 45 hospitals recruiting 165 patients per year without recruitment fatigue, when evidence from the FOCUS study suggests that a more realistic figure is around eight patients per hospital per year.

In conclusion, a prospective, randomised trial would not seem to be possible. Is there a credible alternative? We suggest that a large, prospective observational study should be considered, which would largely overcome the problems inherent in any trial of such a complex, multifactorial intervention as ‘anaesthesia’, as described above. Such a trial would also be expected to eliminate the Hawthorne effect of improved individual outcome merely through trial enrolment, although we recognise that an institutional Hawthorne effect may result from prospective data collection, analysis and interpretation. Practically, a large, observational study is achievable within the framework of data collection that already exists in England, Wales and Northern Ireland, in the form of the National Hip Fracture Database (NHFD) [22]. We accept that there are limitations to observational studies, but provided that recording of data is accurate (which is increasingly possible with improved technology and data guarantors, as evidenced by both the recent European Surgical Outcomes Study and National Anaesthesia Projects), observational studies appear to be as accurate as RCTs in studies of adverse events (such as mortality and delirium, for example), and avoid the ‘spectrum’ bias related to inclusion/exclusion criteria that would limit the applicability of RCT findings to the general hip fracture population [23]. Moreover, although data collectors are already employed by each hospital to collect NHFD datasets, the cost of employing an additional data collector at each hospital (∼£3 million for a one-year study), would be similar in scale to a NIHR grant for a RCT, and might have the advantage of collecting other data that answer specific NICE research questions concerning orthogeriatric and orthopaedic care, e.g. falls and secondary prevention, possibly offsetting the research expense by identifying potential reductions in inpatient stay (∼£16.5 million per inpatient day per annum). To this end, we propose a role for the Health Services Research Centre, supported by the National Institute of Academic Anaesthesia (NIAA), in co-ordinating such a study, as part of a wider national research strategy into anaesthesia for the elderly undergoing emergency surgery, traditionally the most vulnerable of inpatients, but latterly increasingly numerous and with potentially the most to gain from this research.

Conflicts of interest

SW is a member of the AAGBI Hip Fracture Guidelines Working Party, advised NICE during development of CG124, is a Council member of the Age Anaesthesia Association (whom he represents at the NHFD), is national research co-ordinator for the Hip Fracture Perioperative Network, and is an Editor of Anaesthesia. RG chaired the AAGBI Hip Fracture Guidelines Working Party, was a member of the NICE guideline development group for CG124 and founded the Hip Fracture Perioperative Network. He is also a Honorary Secretary elect of the AAGBI. IM is a member of the NICE topic expert group for Quality Standards for hip fracture, a member of the NIAA Research Council and a NIHR grant award holder for trials in hip fracture.