Introduction

Hip fractures in elderly people are almost always fixed surgically.1 A variety of anesthetic techniques can be used for these procedures. In a recent comparative effectiveness review that examined all available strategies (anesthesia- and non-anesthesia-based) to decrease pain in hip fracture, we found that important outcomes, such as survival and major morbidity, are not affected by the choice of anesthetic technique.2,3 Nevertheless, we did find evidence that the use of supplemental peripheral nerve blockade was generally associated with reductions in postoperative pain intensity, delirium, and length of hospital stay (Fig. 1).

Fig. 1
figure 1

Meta-analysis of nerve blockade vs no blockade – acute pain intensity.2 Reproduced with permission from reference2

Full size image

The hip joint and adjacent structures are amenable to nerve blockade by a number of different approaches and techniques. While our findings were consistent with an earlier review which indicated that nerve blocks reduce perioperative pain after hip fracture surgery,4 it is not known which, if any, of these methods is most effective.

By combining the results of randomized controlled trials (RCTs), conventional meta-analysis allows inferences to be made that might be invisible in the individual source studies because of a lack of statistical power. The vast majority of anesthesia trials, as in other areas of medicine, compare one or two types of active treatment with a reference treatment, placebo, or no treatment. While this method yields robust data about the absolute performance of each treatment, it is not methodologically valid simply to collect the aggregate effect sizes of each block and determine the best treatment by choosing the biggest number, because this approach would ignore any data we have from direct comparisons.5

How then can we determine best practice? The difficulty of performing an omnibus multi-armed RCT to find the single best option becomes obvious if, for the sake of discussion, we accept that there are (at least) seven different anatomic approaches to the nerves innervating the hip, three different technical methods of nerve location, catheter vs single-shot options, and several different local anesthetics that can be administered in any number of concentrations and doses, with many different additives, and for a variety of durations. Such a trial would be prohibitively large, complex, and costly.

Multiple treatment comparison (MTC) is a recent development in evidence synthesis that provides inferences on the comparative effectiveness of interventions that may never have been directly evaluated in clinical trials.6 This analytic method, while computationally complex, is conceptually simple. If there are three treatments, A, B and C, the relative effectiveness of A vs C can be inferred even if the only available direct comparisons are A vs B and B vs C. Multiple treatment comparisons have previously been used in cardiology,6 pediatric emergency medicine,7 and respirology.8

We applied MTC to the RCT data obtained during our previous comparative effectiveness review on pain management intervention in hip fracture patients. Our specific aim was to determine whether there is evidence that one or more nerve block approaches are more effective than others when used as a supplement to standard care for hip fracture.

Methods

We followed an a priori research protocol using recognized methodological approaches for conducting systematic reviews (http://effectivehealthcare.ahrq.gov/search-for-guides-reviews-and-reports/?pageaction=displayproduct&productID=368).

Search strategies

We searched 25 electronic databases (including MEDLINE, CINAHL, EMBASE, the Cochrane Database of Systematic Reviews, and Web of Science) and clinical trials registered from January 1990 to December 2010. We conducted hand searches of scientific meeting proceedings and the reference lists of reviews and included studies for RCTs and nonrandomized controlled trials (NRCTs) that a) were published from 1990 to 2010; b) focused on adults aged ≥ 50 yr who were admitted to hospital with acute hip fracture due to low-energy trauma; and c) examined the use of any peripheral nerve block. In order to preserve as much statistical power as possible, we made no analytic distinction between studies in which blocks were placed before, at the time of, or immediately after surgery, or between studies that used single-shot, catheter based, or intermittent injection techniques. No language restrictions were applied to the articles searched. A detailed description of our search strategy is available in the full report.3

Study selection

Two reviewers independently screened titles, abstracts, and the full text of potentially relevant articles. They extracted data, assessed methodological quality, and rated the body of evidence. Discrepancies were resolved by consensus or by third-party adjudication. We extracted study characteristics, inclusion and exclusion criteria, participant characteristics, interventions, and outcomes. We used the Cochrane Collaboration’s tool9 to assess risk of bias. Potential publication bias was explored graphically through funnel plots for comparisons for which meta-analyses were conducted and when there were at least ten studies in the analysis. Additionally, if bias was suspected, publication bias was quantitatively assessed using the Begg adjusted rank correlation test and the Egger regression asymmetry test.10

We selected the following outcomes of interest a priori: acute postoperative pain intensity, delirium, other adverse events, myocardial infarction, stroke, renal failure, and 30-day mortality.

Analysis

We conducted a MTC using a Bayesian network model in a single analysis.6,11,12 Results are expressed as standardized mean differences (continuous outcomes) or odds ratios (binary outcomes) with 95% credible intervals (CrIs). Credible intervals are Bayesian versions of confidence intervals and, for the purposes of this report, can be interpreted in the same way as confidence intervals. The analysis also yields the probability that each of the treatments tested is the best. For the outcomes of acute postoperative pain intensity and delirium, we conducted the MTC using a Bayesian network model to compare all interventions simultaneously and to use all available information on treatment effects in a single analysis. The analysis yields the probability that each of the treatments tested is the best, and delivers effect size estimates with narrower confidence intervals than conventional methods. For the one node where we had both direct and indirect evidence for a treatment comparison, we sought evidence of statistical inconsistency.13

Technical details about the conduct of the MTC are given in Appendix A.

Results

The search strategy identified 9,357 citations. Of these, 21 RCTs comprising 1,422 patients were eligible for inclusion (Fig. 2). One of these studies14 yielded non-analyzable data, leaving 20 studies for analysis. Eighteen pairwise comparisons were possible for acute pain intensity, five for delirium, four for mortality, and two for other adverse events.

Fig. 2
figure 2

PRISMA flow diagram

Full size image

Detailed assessment of each included study’s risk of bias is given in Appendix B. Overall, 1 (5%) RCT15 was assessed as low risk of bias, 9 (43%)14,16-23 had high risk of bias, and the remaining 11 (52%)13,24-33 had unclear risk of bias. The risk of bias relating to incomplete outcome data and selective outcome reporting was assessed as low in the majority of trials, (14/21trials [67%] and 18/21 [86%], respectively). Most trials were assessed as being at unclear risk of bias for sequence generation (15/21 trials [71%]), concealment allocation (16/21trials [76%]), and other bias (13/21 trials [62%]).

The Table lists key characteristics of the eligible studies. They were published from 1991 to 2010 and ranged in size from 14-209 participants. The range of mean age reported for participants was from 59-86 yr, with the majority being female (74%). Most studies (n = 15) were conducted in Europe. The most commonly studied approaches were the femoral and the three-in-one block. In 11 trials, electrical nerve stimulation was used to confirm nerve location, while the remainder used clinical landmarking. None of the trials featured ultrasound guidance. In the majority of trials, the comparator treatment was systemic analgesia or standard care. The definition of these varied from study to study but generally featured systemic opioids titrated to effect with or without a scheduled or on-request regimen of non-opioid analgesics.

Table Patient and study characteristics
Full size table

Fig. 3 shows the number of studies that were examined in each pairwise comparison of acute postoperative pain intensity. Fig. 4 compares the performance of the nerve blocks at reducing acute pain intensity. We were able to compare acute pain intensity between seven nerve block strategies using data from 16 studies comprising 1,089 subjects. In 15 studies, pain was measured on a 10-cm visual analogue scale or an 11-point numeric rating scale,13,16,17,20-25,27,28,30-33 and in one study, a four-point verbal rating scale was used.18 Pain was also measured at a variety of intervals (ranging from hourly to daily). In most cases, this was a single score on the first postoperative day. When faced with a choice, we extracted data for the epoch that showed the most improvement for the intervention arm. For each nerve block, the reduction in pain compared with standard care is presented along with the 95% CrIs for that estimate. Since the constituent studies used a variety of scales to measure pain, the measure is a standardized mean difference (SMD) and should be interpreted in terms of unit effect size. While there is no definitive clinical way to interpret this, the most common interpretation is that an effect of 0.2 is “small”, an effect of 0.5 is “medium”, and an effect of 0.8 is “large”.34 For all but one intervention, the CrIs cross the null line, indicating a non-significant statistical difference between the interventions, each other, or standard care. The exception was the combination of lateral femoral cutaneous and obturator nerve blockade. This treatment was compared with obturator block alone and to fascia iliaca block in two trials13,22 of 165 patients. The estimate for pain intensity change was SMD -2.0 (95% CrI -3.81 to -0.25), and it had the highest probability of being the best choice for pain relief. Fascia iliaca block, studied in three trials22,30,31 comprising 453 patients, was the next best choice on probability grounds, but its effect size for pain relief did not reach statistical significance.

Fig. 3
figure 3

Number of studies available for each pairwise comparison of postoperative pain intensity. “Standard Care” refers to placebo, usual care, systemic analgesia, spinal anesthesia, or single-shot epidural analgesia, depending on the study

Full size image
Fig. 4
figure 4

Comparative efficacy of nerve blocks against acute postoperative pain. Data are standardized mean difference (95% credible intervals). PB = probability of being the best option

Full size image

Fig. 5 compares the performance of nerve blocks at reducing postoperative delirium. The criteria used for the diagnosis of delirium were not reported in three studies.15,19,30 In one study,16 the Mini Mental State Examination (MMSE) was used, but the change in MMSE necessary to define delirium was not provided. In one other trial,31 daily MMSE and two other instruments were used to diagnose delirium against the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV) criteria. Fascia iliaca was the only block that had a statistically significant association with reduction in postoperative delirium (Estimate SMD -0.20; 95% CrI: -0.07 to -0.83).

Fig. 5
figure 5

Comparative efficacy of nerve blocks against delirium. Data are standardized mean difference (95% credible intervals). PB = probability of being the best option

Full size image

Discussion

We previously described a robust association between the use of peripheral nerve blockade in hip fracture surgery and improvements in acute pain, delirium, and length of hospital stay.2,3 In the current analysis, all available data were compared simultaneously using MTC, and the combination of obturator and lateral femoral cutaneous blockade yielded the best reductions in acute pain. Fascia iliaca blockade was associated with the best reductions in delirium.

The apparent magnitude of the effect of these techniques (SMD = -2.0) on perioperative pain is clinically relevant. In addition, delirium is a significant cause of morbidity and cost following hip fracture, and the results of this analysis show a > 70% reduction in delirium associated with fascia iliaca blockade.35 It should be noted that most of the trials included in this synthesis excluded patients with preoperative cognitive impairment, arguably the group in whom delirium is most likely to occur and hardest to detect.36

Each of the peripheral nerve blocks studied represents something of a clinical compromise because the hip joint transmits sensation via branches of the femoral, obturator, superior gluteal, and sciatic nerves, and the nerve to the quadratus femoris,37 and no single peripheral injection site allows all of these nerves to be reached. In addition, pain following hip surgery is not only generated by the hip joint but also by the soft tissues that are disrupted as part of the surgical approach. This may partially explain why the fascia iliaca block— which blocks the lateral cutaneous nerve of the thigh more reliably than the femoral or three-in-one approach — and the combined obturator/lateral cutaneous nerve of the thigh block performed better than other methods.

There are obvious hazards when comparing studies of different blocks that used local anesthetics of different durations, continuous vs single-shot techniques, and different additives. Since the pain of soft tissue dissection and osteosynthesis does not disappear on the first postoperative day, intuition would suggest that blocks performed with longer-acting agents or blocks that involved continuous or patient-controlled infusions of local anesthetics would be more beneficial for our outcomes of interest. Regrettably, there were inadequate data to compare single-shot and continuous forms of each individual block directly. We accept that we may have missed important nuances by choosing to combine blocks by type, whether performed before, during, or after surgery.

Regional anesthesia is a rapidly evolving area of anesthesia practice, and the fact that ultrasound guidance was not used in any of the included trials may tempt some readers to dismiss this entire analysis as irrelevant. In our view, the relevant question is not whether ultrasound permits safer or more reliable blockade of the chosen target but whether it changes the efficacy of any given block technique more than any other. We would argue that the major nerves to the hip joint are, for the most part, anatomically consistent and amenable to reliable clinical landmarking and show easily visible responses to electrical nerve stimulation. Consequently, while ultrasound will improve the reliability and safety of these blocks, it may end up that it does not create new differences in efficacy between them.

The limitations of this study should be acknowledged. We were unable to review 25 studies from our original comparative effectiveness review searches either because we could not secure a translation or we could not find the original copy. The impact of this shortcoming in this study is mitigated by the fact that only three of those 25 titles appeared to be controlled studies of nerve blocks (one each femoral, lumbar plexus, and three-in-one block), while one other seemed to be a review article about nerve block techniques. Our assessment of the methodological quality on study publications was performed independently using the risk of bias tool, and we did not contact authors to verify the methods used. Although the methods were poorly reported in some trials, they may have been adequately conducted.

One of the main limitations of MTC is the presence of statistical inconsistency, that is, when there is no agreement between direct and indirect evidence. This was not an issue in our analysis as the only node that had both types of evidence showed consistency between the direct and indirect evidence. In MTC, we also assume that unobserved treatment effects are missing at random and that trials in two different comparisons are exchangeable. These assumptions are similar to those made in a standard meta-analysis. As with conventional meta-analysis, this method is designed to respect the randomization of the original studies and the assumptions that are made about the characteristics of those studies. Threats to the validity of an MTC analysis are similar to those in conventional meta-analyses.

Our analysis was also limited by the paucity of available data. More research, particularly studies including patients with cognitive impairment and studies comparing multiple nerve blocks simultaneously would improve our ability to make good therapeutic decisions in this important clinical area. We were further constrained by the lack of standardization in the reporting of pain as an outcome, which made it impossible to render the aggregate changes in pain intensity as clinically meaningful values. Finally, although pain and function are correlated, most studies focused on pain relief and did not evaluate the effects of an intervention on the patient’s ability to mobilize after surgery, a factor linked to recovery levels after hip fracture.38

In conclusion, using MTC, we found that the combination of obturator and lateral femoral cutaneous blockade yielded the best reductions in acute pain following surgery for a hip fracture, while fascia iliaca blockade was associated with the best reductions in delirium. Multiple treatment comparison, a tool to use when simultaneously comparing the effectiveness of multiple treatments, provides a useful guide for anesthesia providers seeking the best treatment when faced with a body of RCTs in which each trial investigates one treatment.