Neuraxial anesthesia for orthopedic surgery: systematic review and meta-analysis of randomized clinical trials

CONTEXT AND OBJECTIVE: Taking the outcome of mortality into consideration, there is controversy about the beneficial effects of neuraxial anesthesia for orthopedic surgery. The aim of this study was to compare the effectiveness and safety of neuraxial anesthesia versus general anesthesia for orthopedic surgery. DESIGN AND SETTING: Systematic review at Universidade Federal de Alagoas. METHODS: We searched the Cochrane Central Register of Controlled Trials (Issue 10, 2012), PubMed (1966 to November 2012), Lilacs (1982 to November 2012), SciELO, EMBASE (1974 to November 2012) and reference lists of the studies included. Only randomized controlled trials were included. RESULTS: Out of 5,032 titles and abstracts, 17 studies were included. There were no statistically significant differences in mortality (risk difference, RD: -0.01; 95% confidence interval, CI: -0.04 to 0.01; n = 1903), stroke (RD: 0.02; 95% CI: -0.04 to 0.08; n = 259), myocardial infarction (RD: -0.01; 95% CI: -0.04 to 0.02; n = 291), length of hospitalization (mean difference, -0.05; 95% CI: -0.69 to 0.58; n = 870), postoperative cognitive dysfunction (RD: 0.00; 95% CI: -0.04 to 0.05; n = 479) or pneumonia (odds ratio, 0.61; 95% CI: 0.25 to 1.49; n = 167). CONCLUSION: So far, the evidence available from the studies included is insufficient to prove that neuraxial anesthesia is more effective and safer than general anesthesia for orthopedic surgery. However, this systematic review does not rule out clinically important differences with regard to mortality, stroke, myocardial infarction, length of hospitalization, postoperative cognitive dysfunction or pneumonia.


INTRODUCTION
Neuraxial anesthesia combined with postoperative epidural analgesia can reduce the physiological stress attributed to surgery and the incidence of postoperative complications. 1 There are some advantages to neuraxial anesthesia, such as reductions in the incidence of deep-vein thrombosis and pulmonary embolism, and in the need for blood transfusion. 2Although there are advantages, there is some potential for neurological damage and a great degree of hypotension, which may make it less acceptable. 1 Rodgers et al. published a systematic review demonstrating that neuraxial anesthesia can decrease mortality over the course of the follow-up time, compared with general anesthesia, but these results cannot be used in orthopedic surgery because most of the procedures involved consisted of abdominal surgery. 3uraxial anesthesia is used routinely in orthopedic surgery, but there is controversy regarding the beneficial effects of this anesthetic technique, taking into consideration the length of follow-up and its correlation with mortality and postoperative morbidity. 4though some systematic reviews have analyzed neuraxial anesthesia for orthopedic surgery, the reviewers looked for the magnitude of the effect in a separated manner, without pooling knee, hip, femur and ankle results in a meta-analysis.6][7][8] In this context, we proposed to answer the research question: what is the effectiveness and safety of neuraxial anesthesia, in comparison with general anesthesia for orthopedic surgery?

OBJECTIVE
The purpose of this systematic review was to compare the effectiveness and safety of neuraxial anesthesia versus general anesthesia for orthopedic surgery.

Protocol
A protocol was developed a priori and is available from the corresponding author if it needs to be analyzed.This research was conducted in accordance with the Cochrane Collaboration Handbook. 9The journals, institutions where the studies were conducted or researchers did not influence our results.The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement was followed in relation to reporting items for systematic reviews and meta-analyses. 10

Eligibility criteria
Types of participants: The patients included in this review were 18 years old or over.They presented orthopedic disorders below the umbilical scar, and were treated surgically.Patients who underwent orthopedic surgery performed together with other types of surgery were excluded.
Types of studies: Randomized controlled trials (RCTs) were included.Data from studies published twice were collected from the original article with the best description.Studies with incomplete outcome data description were excluded from the meta-analysis.

Types of interventions:
the intervention group was neuraxial anesthesia.The control group was general anesthesia.Use of a catheter in neuraxial anesthesia techniques was not an exclusion criterion.

Identification of studies
The following databases were searched:  .In addition, the reference lists of the studies included were also searched.There were no restrictions on any language, date and document format.The search strategies used in Medline via PubMed were adapted and used for CENTRAL.We used the terms anesthesia and orthopedic surgeries for Lilacs.We used the terms anesthesia and urology for SciELO.The search strategy for EMBASE was 'general anesthesia'/exp OR 'spinal anesthesia'/exp OR 'epidural anesthesia'/exp AND rand* AND 'orthopedic surgery'/exp [embase]/lim.The search strategy for PubMed can be seen in Table 1.

Selection of studies
Titles, abstracts, or both, identified through the search strategy for all databases, were independently reviewed by two investigators (Barbosa FT and Rodrigues CFS).RCTs identified in accordance with our eligibility criteria were obtained in order to read the full text.A standardized form was developed by the authors and was used to collect data.Discordances were resolved through consensus meetings.

Methodological quality and risk-of-bias assessment
The study validity of the RCTs was independently assessed by two authors (Barbosa FT and Rodrigues CFS).Discordances were resolved through consensus meetings.The risk of bias was determined by means of the Rob table, as recommended by the Cochrane Handbook. 9The Rob table analyzes sequence generation, allocation sequence concealment, blinding, incomplete outcome data, selective outcome reporting and other sources of bias.
Each item was judged subjectively, looking for bias.Three answers were possible: low risk of bias, high risk of bias, or unclear risk of bias.This instrument generated a figure showing the risk-of-bias summary for each study included.

Outcomes
The primary outcome was mortality.Mortality was defined as a fatal event during surgery or occurring within one year afterwards. 3e secondary outcomes were stroke, myocardial infarction, length of hospitalization, quality of life, degree of satisfaction, postoperative cognitive dysfunction, blood transfusion requirements and pneumonia.We considered stroke to be a loss of brain function caused by a disturbance in brain blood supply.
Myocardial infarction was considered to be a loss of cardiac function caused by a disturbance in coronary blood supply.Length of hospitalization was the duration of hospital stay.Quality of life was the aspect of life that was influenced by physical wellbeing or mental status. 9Degree of satisfaction was the patient's reaction to the healthcare received. 11Postoperative cognitive dysfunction was a state of mental confusion after orthopedic surgery.Blood transfusion requirement was considered to be the number of blood units transfused.Pneumonia was lung infection with changes in pulmonary radiography that started at least 48 hours after surgery.

Data analysis
For dichotomous outcomes, the odds ratio and 95% confidence interval (CI) were calculated using a random-effect model (REM).When the effect was absent, the risk difference (RD) and 95% confidence interval were calculated using REM.For continuous outcomes, the mean and standard deviation were used to generate the mean difference (MD) and 95% CI using REM.
The Rev Man 5 statistical package (Cochrane Collaboration) was used to perform meta-analyses. 12I 2 statistical heterogeneity was assessed by using heterogeneity tests, i.e. the standard chi-square test (P-value < 0.10 or < 10%) and the I 2 test (I 2 > 50% was statistically significant).

Study selection
A flow diagram demonstrating the process for selecting relevant articles is outlined in Figure 1.In total, 5032 titles and abstracts were screened.We analyzed 4591 titles after running the search strategy, and 32 papers were identified as relevant through this process.  Fift40,43 The reasons for their exclusion can be seen in Figure 1.,44 We also analyzed 441 titles from the reference lists of these 17 studies included, but did not find any additional studies (Figure 1).
Sequence generation.Two studies included in this systematic review were considered to have low risk of bias. 38,44The other studies did not present the method used to generate the allocation sequence.
Allocation concealment.The method used by the authors to conceal the allocation sequence was described correctly in four of the studies included. 18,25,27,29The other studies did not report any details regarding allocation concealment that would make it possible to determine whether this process had been free of bias.
Blinding of participants, personnel and outcome assessors.
One study was considered to present high risk of bias because although the authors reported that the investigator was blinded to randomization, they did not describe the condition of the participants and the outcome assessors. 30The other studies included did not report sufficient details to determine whether they presented high or low risk of bias, or they reported that the outcome assessors were blinded.
Incomplete outcome data.In the study by Valentin et al., the result relating to blood loss was reported in a figure, but this presentation did not allow the numbers to be seen and readers were left to estimate the amount of blood loss. 41In the study by Juelsgaard et al., usable Holter data were obtained from 43 patients out of the 54 participants. 30lective reporting.In two of the studies included, the authors reported results relating to outcomes that were not described in the method. 29,34her sources of bias.Eight studies received the classification of high risk of bias.In the study by Berggren et al., a scale used to analyze postoperative confusion was modified by the authors and the validation process for this new scale was not tested. 15In the study by Bigler et al., the t test was used in the data analysis, but these data presented asymmetric distribution. 16In the study by Davis et al., the participants may not have received the same intervention, because the anesthesia was administered by "the duty registrar or consultant anesthetist". 22In another study

Outcomes
The characteristics of the selected RCTs analyzed and their outcomes are shown in Table 2. 15,16,18,22,23,25,27,29,30,34,35,37-39,41,42,44 It was Berggren et al. 15 Bigler et al. 16 Casati et al. 18 Davis et al. 22 Davis et al. 23 Eroglu et al. 25 Hole et al. 27 Jones et al. 28 Juelsgaard et al. 30 McKenzie et al. 34 McLaren et al. 35 Nielson et al. 37 Racle et al. 38 Riis et al. 39 Valentin et al. 41 White et al. 42 Williams-Russo et al. 43 Random  not possible to pool the data for degree of satisfaction, blood transfusion requirements or quality of life.The data on the degree of satisfaction and blood transfusion requirements were not described correctly in the studies included.The authors did not report any data about quality of life in the studies included.
Pneumonia: This was analyzed in three studies. 15,38,42Two types of surgery were conducted: hip and femur surgery.There was no statistically significant difference between the types of surgery (OR = 0.61; 95% CI: 0.25 to 1.49; P = 0.28; 167 participants) (Figure 9).

Sensitivity analysis
None of the authors reported dropouts but they did report the frequencies with which the outcomes occurred.The results did not change when we analyzed only the studies with the same length of follow-up time.
One study was responsible for statistical heterogeneity in the mortality analysis. 35However, there was no statistically significant difference when the data of this study was not taken into consideration (RD = -0.01;95% CI: -0.02 to 0.01; P = 0.45).
The analysis on stroke showed statistical heterogeneity in a test for subgroup differences.The study by Berggren et al. 15 was responsible for this, because more patients in the neuraxial anesthesia group presented hypotension than was seen in the general anesthesia group, and four patients were unresponsive to treatment.

Knee surgery
Williams-Russo et al. 43 12        poor quality.Random sequence generation, allocation concealment and blinding were not described correctly, were omitted or were not conducted appropriately.The results from the studies included may therefore be limited.Appropriate reporting of the methodological items for designing and conducting studies is important for ensuring quality in a systematic review. 9This systematic review can be considered to present good current evidence because we pooled data from studies with the same surgical procedures, analyzed methodological flaws in the studies included before pooling the results and explored occurrences of heterogeneity.The poor quality of the studies included may place limitations on our results.
[6]8 We demonstrated the same result as in these previous studies, but the length of follow-up among our included studies reached six months in only one study and was less than three months in the others.
Stroke was analyzed in three RCTs, with no differences between the groups. 15,22,38These were small studies, and our stroke rate was 2.7% (7/259), with no statistically significant difference between the groups.Although meta-analysis can improve statistical power, our sample size was small.In this systematic review, there were no subgroup analyses.There was an analysis on the types of surgery, and this was done in all the metaanalyses.The test for the difference between the groups in the stroke analysis showed an I 2 test result of 57.5%.We reviewed the articles included in this analysis again and noted that there was an event in the study by Berggren et al. 15 that was not seen in any other study.These authors reported cases with arterial hypotension that did not resolve with the treatment used, which may have been responsible for the greater frequency of stroke in the neuraxial anesthesia group.On the other hand, they did not report whether this event was present in the other group, how the treatment was performed or whether the habitual treatment used in clinical practice was used in these cases.
The cardiac protective effect from neuraxial anesthesia seems to be a matter of controversy in noncardiac surgery. 45Parker et al. analyzed patients who underwent hip fracture surgery and found that the rate of myocardial infarction was 1% (5/505) in the neuraxial anesthesia group. 8Our rate for this outcome was 1.7% (5/291) with no statistically significant difference, but those authors analyzed thoracic epidural anesthesia.
Length of hospitalization was reported in three studies. 23,38,44cfarlane et al. reported that neuraxial anesthesia produced a beneficial effect for total knee arthroplasty. 5They concluded that neuraxial anesthesia can facilitate rehabilitation and can reduce hospital stay.However, we did not observe this result.
Postoperative cognitive dysfunction was analyzed in six studies. 15,18,38,39,42,44Each study included used a different analysis method: Organic Brain Syndrome Scale, 15 Mini Mental State Examination, 18 mental changes, 38 psychologist and attention test, 39 mental confusion 42 and neurophysiological tests. 44The results relating to this outcome are questionable.
Pneumonia was reported in three studies. 15,38,42Furthermore, a previous systematic review showed that this outcome was less common following neuraxial anesthesia than following general anesthesia (OR 0.37; 95% CI: 0.15 to 0.89).However, the scenario analyzed was vascular surgery, and one study in that review reported more events in the general anesthesia group than were seen in other studies. 46In that study, there were more smokers in the general anesthesia group and bias may have occurred.
Our results were unable to prove the same effect from neuraxial anesthesia in relation to orthopedic surgery.
For future research, attention needs to be paid to the way in which random sequence generation, allocation concealment and blinding are reported.Appropriate use of these methodological criteria can improve the quality of systematic reviews. 5These topics need to be reported in sufficient detail for readers to be able to judge whether the results are good enough to be reproducible in clinical practice.The mortality rate can help to elucidate the effectiveness and safety of neuraxial anesthesia in comparison with general anesthesia.Making the assumptions of 5% mortality in the general anesthesia group, 1% mortality in the neuraxial anesthesia group, 80% power and 5% significance level, it will be necessary to have 284 participants in each group for future studies that analyze mortality.More RCTs with adequate numbers of patients and external and internal validity are needed.
The implications for clinical practice are that, so far, it is not possible to say whether neuraxial anesthesia is more effective and safer than general anesthesia for orthopedic surgery.Each patient should be analyzed individually, and anesthesiologists should take into account their own previous experiences and hospital working conditions.

CONCLUSION
So far, the evidence available from the studies included is insufficient to prove that neuraxial anesthesia is more effective and safer than general anesthesia for orthopedic surgery.However, this systematic review does not rule out clinically important differences with regard to mortality, stroke, myocardial infarction, length of hospitalization, postoperative cognitive dysfunction or pneumonia.

5 )Figure 1 .
Figure 1.Flow diagram summarizing the process for selecting original articles.

by
Davis et al., the follow-up time may not have been the same for all participants.23In the study by Hole et al., patients of different ages received different premedication, and the surgical technique may not have been the same because thirteen surgeons participated in this study.27In the study by McKenzie et al., statistical tests and significance level were not described.34In the study by Valentin et al., there were more ill patients in one group.41In the study by White et al., the authors reported that special attention was given to any respiratory problems during the postoperative period, but they did not report what this attention comprised.42

Figure 2 .
Figure 2. Methodological quality summary: risk of bias for each study included.

Figure 7 .
Figure 7. Forest plot including the six studies that analyzed postoperative cognitive dysfunction.

Figure 8 .
Figure 8. Forest plot including the three studies that analyzed pneumonia.

Figure 9 .
Figure 9. Risk of bias in studies included.

Table 2 .
Characteristics of the randomized controlled trials that compared the kinds of anesthesia for orthopedic surgery and their contribution to the meta-analysis