Introduction

Total knee arthroplasty is amenable to various regional anesthesia (RA) techniques. Central neuraxial blockade (CNB) can provide excellent intraoperative anesthesia and prolonged postoperative analgesia. Peripheral nerve blockade (PNB) avoids many of the unwanted adverse effects of CNB and allows for targeted analgesia of the operative limb [17, 23, 85]. In recent years, the use of continuous PNB (CPNB) has escalated, because it has the advantage of prolonging postoperative analgesia compared with single-injection techniques [10, 33].

Despite a low rate of complications and published benefits in certain orthopaedic procedures, including superior postoperative analgesia, improved rehabilitation, and reduced length of hospital stay, there are disadvantages of RA [3–5, 12–14, 20, 58, 74, 80, 82]. There is an inherent block failure rate (reportedly between 0% and 67%), although this varies considerably with the particular block, operator experience, and method of nerve localization [19, 20, 62]. Operating room delays and a perceived risk of increased liability also are criticisms often directed at RA [46, 65]. Other limiting factors include the training required to develop the necessary technical skills for successful RA and, more recently, the expense of ultrasound equipment as this method of nerve localization increases in popularity. Finally, many patients are fearful of RA and may have misconceptions about the technique [53].

Although RA is undergoing a renaissance, the results of meta-analyses and randomized, controlled trials (RCTs) comparing general anesthesia (GA) and RA for major lower limb orthopaedic surgery often are conflicting [9, 15, 54, 67, 78, 84]. It is not uncommon for the results of large RCTs to disagree with each other and with those of meta-analyses [24, 31]. This latter effect can be the result of the inclusion of small studies, publication bias, sample heterogeneity between different trial populations, and meta-analysis bias [34, 49, 55]. More importantly, many trials included in recent meta-analyses were originally published more than 30 years ago and do not reflect modern anesthetic or surgical practice. During the past two decades, postoperative care of surgical patients has improved, new thromboembolic prophylaxis regimes have been introduced, and RA has advanced as a result of enhanced needle technology, block placement techniques, catheter design, and infusion pumps [26, 29, 71, 75].

We therefore performed a systematic review of the contemporary literature (published from 1990 onward) to ascertain if RA and/or regional analgesia were superior to GA and/or systemic analgesia for TKA. The specific questions we sought to answer were whether, when compared with GA and/or systemic analgesia, RA and/or regional analgesia for TKA decreased (1) mortality, (2) cardiovascular morbidity, (3) deep venous thrombosis (DVT) and pulmonary embolism (PE), (4) blood loss, (5) duration of surgery, (6) pain, (7) opioid-related adverse effects, (8) cognitive defects, and (9) length of stay. We also examined whether RA improved rehabilitation compared with GA.

Materials and Methods

We (GAP, RB) searched the electronic databases MEDLINE, EMBASE, and the Cochrane Central Register of Controlled Clinical Trials (from January 1990 to October 2008) using the following population search terms: “total knee replacement” OR “total knee arthroplasty” OR “knee operation”. These search results then were combined with “anesthesia” OR “analgesia” using the Boolean search operator AND. Only RCTs were included, and the search subsequently was limited to English language studies involving human adults. Each abstract then was screened to identify studies that had randomized patients to compare GA versus RA for surgery. RCTs comparing systemic versus regional techniques for postoperative analgesia also were included. The references of the resulting RCTs were examined for any relevant articles not identified in our original search. The specific outcomes sought in each article were (1) mortality, (2) cardiovascular morbidity (myocardial infarction, pulmonary edema, hypotension), (3) DVT, (4) PE, (5) blood loss, (6) duration of surgery, (7) pain (pain scores and morphine consumption), (8) opioid-related adverse effects (nausea, vomiting, pruritis, sedation, urinary retention), (9) cognitive defects, (10) length of stay, and (11) rehabilitation (knee flexion, extension, ambulation). It was noted whether each outcome was primary or secondary.

We excluded studies if surgery other than a joint arthroplasty was performed or if the knee and hip arthroplasties were treated as one study population and data for the patients undergoing knee surgery were not presented separately in the results [11, 27, 30, 81]. Studies using opioid only neuraxial techniques or in which regional analgesia was not administered on the day of surgery were excluded [47, 64]. Finally, studies were excluded if the primary outcome was not included in the list described above [7, 48].

We used a templated evidence-based medicine literature review form to assist in the systematic review of articles and in the data collected. Demographic data extracted for comparison included year of publication, author, total number of subjects, mean patient age, percent male, and comorbidity. The intervention (specific RA and/or regional analgesia technique) and comparator (GA and/or specific systemic analgesia technique) were recorded. Each outcome then was evaluated qualitatively for each intervention and comparator and the data recorded in tables. Because there were a limited number of studies with homogenous design for each outcome, meta-analysis was not performed.

Several criteria were used to assess the quality of each trial. The likelihood of methodologic bias of each RCT was assessed using the Jadad score, which assigns points based on three factors [39]. One point was given to randomized studies, an additional point was given if the method of randomization was described and appropriate, and one point was deducted if randomization was inappropriate. One point was given if a study was double-blind, and an additional point was given if the blinding procedure was described and appropriate. One point was deducted if blinding was inappropriate. One point was given if the numbers and reasons for withdrawals were described. The maximum score is 5; trials scoring 3 or more generally are regarded as having satisfactory methodologic quality. Allocation concealment, which helps eliminate selection bias, was assessed and defined as adequate, unclear, or inadequate. Finally, whether patient followup rates were less than 80% was recorded.

After abstraction of information, a level of evidence (see Guidelines for Authors) was assigned to the outcomes of each RCT (Level I is a high-quality RCT; Level II is a lesser-quality RCT, eg, less than 80% followup, no blinding, or improper randomization). Two authors (AJRM, RB) independently reviewed and scored each RCT using the aforementioned methodology.

Results

We identified 28 RCTs that compared either GA versus RA and/or systemic versus regional analgesia for TKA (Fig. 1). The 28 studies included 1538 patients. Fourteen of these had a Jadad score of 2 or less. Allocation concealment was unclear in 27 trials and inadequate in one. In one RCT, the dropout of participants was greater than 20% (Table 1). Eleven of the 28 RCTs were considered to provide Level I evidence. We summarized the 11 outcomes in each of the 28 trials and noted the Level of Evidence and the direction of difference between the two types of anesthesia (Table 2).

Fig. 1
figure 1

A flowchart shows the included and excluded studies.

Full size image
Table 1 Demographic data, design, and breakdown of Jadad score for each trial
Full size table
Table 2 Outcomes and levels of evidence for each study included in the review*
Full size table

Only one trial (comparing epidural anesthesia and GA) recorded mortality (during the first 8 weeks postoperatively) as a secondary outcome [83]; the authors observed no difference with one death in each group (Level II) [83]. There are no recent RCTs primarily designed to assess differences in mortality after GA versus RA for TKA.

Nine trials examined cardiovascular morbidity, always as a secondary outcome. In the three that recorded postoperative myocardial infarction and pulmonary edema, there was no difference between the two anesthetic techniques (Level II) [16, 73, 83]. The incidence of postoperative hypotension was recorded in eight studies (Table 3). In three of the six that compared epidural analgesia with either systemic analgesia or other methods of regional analgesia, there was more postoperative hypotension in the epidural group (Level I).

Table 3 General anesthesia versus regional anesthesia and/or systemic versus regional analgesia for TKA: cardiovascular morbidity*
Full size table

When chemical thromboprophylaxis was administered, there was no difference in the incidence of DVT (Table 4) or PE (Table 5) between GA and RA for TKAs (Level II) [16, 21, 57, 82]. In one Level I study, there was a decreased incidence of DVT in favor of RA; however, no chemical thromboprophylaxis was used [41].

Table 4 General anesthesia versus regional anesthesia and/or systemic versus regional analgesia for TKA: perioperative deep venous thrombosis*
Full size table
Table 5 General anesthesia versus regional anesthesia and/or systemic versus regional analgesia for TKA: perioperative pulmonary embolus*
Full size table

Of the five RCTs that addressed perioperative blood loss, none reported a difference in either blood loss or transfusion requirements with RA compared with GA (Level II) [6, 16, 41, 57, 63].

The duration of surgery was not influenced by the type of anesthetic for TKA (Level II) [6, 14, 16, 25, 45, 57, 59, 61, 63, 70, 72, 82, 83].

Our review identified 24 RCTs that compared systemic and regional analgesia for TKA, and in 21 of these trials, RA reduced pain scores and/or morphine consumption. Epidural analgesia (Level I), single-injection FNB (Level I) either with or without sciatic nerve block (Level II), continuous catheter-based FNB (CFNB) (Level II), and continuous psoas plexus block (Level II) were superior to systemic analgesia (Table 6) [2, 6, 14, 16, 18, 21, 25, 28, 35, 44, 56, 59, 61, 69, 70, 74, 76, 79]. One Level I study only reported an analgesic benefit (reduced pain scores) from FNB compared with systemic analgesia when the FNB was combined with an obturator nerve block [50]. Obturator block alone was of no value (Level II) [45]. When reported, the analgesic benefit after single-shot PNBs compared with systemic analgesia lasted as much as 48 hours. The effect of CFNB on pain scores varied in different studies from a reduction in the recovery area only to as much as 48 hours. Epidural analgesia was effective in reducing pain scores for as much as 10 days when the infusion was continued for 7 days postoperatively [21]. Only one study included a comparison of single-injection FNB with CFNB for postoperative analgesia after TKA [35]. In this study, placement of a catheter provided little additional benefit, although this finding has since been countered in the literature [68, 80].

Table 6 General anesthesia versus regional anesthesia and/or systemic versus regional analgesia for TKA: postoperative pain*
Full size table

Eighteen trials reported opioid-related adverse effects, and although all were analyzed as secondary outcomes, there was evidence that FNB (Level I), FNB plus obturator block (Level I), FNB plus sciatic block (Level I), CFNB (Level II), and epidural analgesia (Level II) reduced opioid-related adverse effects (Table 7) [21, 35, 50, 56, 76, 79]. Specifically, postoperative nausea and vomiting and sedation were less frequent in the RA groups. Urinary retention, however, was, in two studies, greater in patients who received epidural analgesia compared with patients who received PNB and systemic analgesia (Level II) [14, 74]. Pruritis also was increased, compared with systemic analgesia, in one of the three studies in which epidural opioids were combined with local anesthetic in the infusion (Level II) [6]. In 10 of the 18 studies, there were no differences in adverse effects between the GA and RA groups (Level II), although frequently there were trends toward a benefit of RA [1, 2, 16, 25, 28, 35, 45, 61, 69, 70].

Table 7 General anesthesia versus regional anesthesia and/or systemic versus regional analgesia for TKA: adverse effects*
Full size table

In the two studies that examined short-term (1 week) or long-term (3 and 6 months) cognitive function, the anesthetic technique made no difference (Level II) [63, 83].

Of the 12 RCTs that examined length of stay, one Level I and two Level II studies found CFNB or FNB can reduce length of hospital stay by up to 1 day and/or length of rehabilitation center stay by up to 13 days (Table 8) [14, 74, 79]. The remainder of the studies reported no difference, but these were all Level II evidence.

Table 8 General anesthesia versus regional anesthesia and/or systemic versus regional analgesia for TKA: length of stay*
Full size table

Among the 14 studies that investigated postoperative rehabilitation for TKA, six reported RA improved this process compared with GA (Table 9) [14, 25, 69, 74, 79, 82]. There was Level I evidence that range of motion and ambulation were improved by either FNB or CFNB. Epidural analgesia and CFNB can help to attain rehabilitation milestones earlier than intravenous patient-controlled analgesia [17].

Table 9 General anesthesia versus regional anesthesia and/or systemic versus regional analgesia for TKA: rehabilitation*
Full size table

Discussion

The aim of this systematic review was to examine the best available evidence comparing GA and/or systemic analgesia versus RA and/or regional analgesia for TKAs. The specific questions we sought to answer were whether, when compared with GA and/or systemic analgesia, RA and/or regional analgesia for TKAs decreased mortality, cardiovascular morbidity, DVT and PE, blood loss, duration of surgery, pain, opioid-related adverse effects, cognitive defects, and length of stay. We also wanted to know if RA improved rehabilitation.

Before further considering the implications of our review, we accept there are several limitations. First, for practical reasons, we chose to include only English language trials. Although this may have introduced bias, Juni et al. [42] suggested excluding trials not published in English has little effect on summary treatment effect estimates. Second, we found 14 of the 28 RCTs evaluated here had Jadad scores of 2 or less. However, not all of these trials were of poor methodologic quality. For example, when studying PNBs for systemic analgesia, sham nerve blocks often are not administered for ethical reasons. Regardless, the Jadad score is effectively reduced automatically by 2 points for the lack of a double-blinded study design. Third, we identified no large (n > 1000) trials. The studies included in our review had sample sizes varying from only 20 to 262 patients. In trials with small numbers of subjects, the absence of significant differences in secondary outcomes must be interpreted with caution, because these studies are often inadequately powered to detect such differences [52]. Lack of evidence is not the same as evidence of absence and therefore secondary outcomes are highlighted in the summary tables. These shortcomings also are reflected in the level of evidence scores (ie, by definition, Level II). Finally, the purpose of this review was not to provide recommendations on the preferred mode of anesthesia for TKA. To do so would have required an assessment of harm and consideration of other information such as costs, quality of life, and feasibility. One major concern with RA is the risk of nerve injury compared with GA. This is difficult to quantify and has been addressed elsewhere [13]. As with any anesthetic technique, patient preference and expertise of the anesthesiologist also must be considered.

The lack of difference in mortality between GA and RA for TKA is unsurprising to us given the safety of modern anesthetic and surgical practices. Much greater numbers than those included in the one RCT that we identified would be required to show any difference. In a large meta-analysis comparing GA and CNB for various types of surgery, Rodgers et al. [67] reported overall mortality was reduced by 1/3 (odds ratio, 0.70; 95% confidence interval, 0.54–0.90) in patients allocated to CNB. When each surgical group in this meta-analysis was analyzed individually, there was decreased mortality only in the orthopaedic group [67]. Overall mortality was reduced regardless whether CNB was continued postoperatively. Conversely, combined intraoperative GA and CNB negated the mortality benefit of CNB alone.

Three RCTs examined cardiovascular morbidity other than hypotension in TKA and compared GA and RA. The lack of difference in these trials could have been the result of inadequate numbers of patients. In the meta-analysis by Rodgers et al. [67], there was a reduction in the incidence of myocardial infarction in the epidural group, although the confidence interval just reached zero. This difference was detected only when all surgical groups were combined and did not specifically apply to orthopaedic patients alone. Additional RCTs with large numbers are needed to examine whether RA reduces serious cardiovascular morbidity in TKA. Although three trials detected an increased frequency of hypotension in patients undergoing epidural analgesia compared with other methods of regional or systemic analgesia, there was no information regarding whether this resulted in any other morbidity. Two of these trials were included in a recent meta-analysis, which concluded epidural analgesia for TKA caused more hypotension than PNBs (odds ratio, 0.19; 95% confidence interval, 0.08–0.45) [23].

Although a recent meta-analysis showed the incidence of DVT in THA was reduced by using RA compared with GA, there are no similar meta-analyses for TKA [54]. In the only study we found that had a decreased incidence of DVT with RA, patients did not receive chemical thromboprophylaxis [41]. Four studies reported no difference in the incidence of DVT, but two were inadequately powered and the others were of poor quality [16, 21, 57, 82]. In a subset of patients in one of these trials, there was no difference in plasma markers of thrombin generation or fibrinolytic activity between the GA and RA groups [72]. Another study, however, measured laboratory parameters of coagulation and reported an increase in coagulability in the GA group compared with RA [48]. Unfortunately these authors did not seek to formally diagnose DVT or PE in their study. It has been suggested CNB may decrease the incidence of DVT either directly by enhancing lower extremity venous blood flow or indirectly by facilitating postoperative rehabilitation after TKA, but additional work is required to ascertain whether RA offers any additive benefit when used in combination with contemporary preventive strategies such as routine thromboprophylaxis and rapid postoperative mobilization. Although no difference occurred between GA and RA in the three trials that compared the incidence of PE, the method of randomization was inadequate in one trial and the sample size was inadequate in the other two.

Although we found no difference in blood loss or transfusion requirements with RA compared with GA, all studies were of poor methodologic quality or had inadequate sample sizes. Furthermore, intraoperative blood loss in TKA is minimal because of the use of a tourniquet and only two studies actually extended measurements into the postoperative period [16, 41]. Blood loss in TKA is multifactorial and it is not clear therefore if RA can offer any additional benefit [43].

Our systematic review revealed RA reduced postoperative pain, particularly on movement, in TKA. Even when no differences in pain scores were reported between the systemic and regional analgesia, patient satisfaction scores still favored CFNB compared with intravenous patient-controlled analgesia [69]. Choi et al. [15] published a meta-analysis comparing postoperative epidural analgesia with systemic analgesia after THA or TKA. They concluded epidural analgesia provided better pain relief only for up to 6 hours postoperatively compared with systemic analgesia. All patients, however, whether they had THA or TKA, were analyzed in aggregate as one group despite important differences in the severity of postoperative pain between these two surgical procedures. In contrast, we found when TKA is examined independently, pain scores frequently were reduced for up to 48 hours. As may be expected, this analgesic benefit was most evident when continuous catheter techniques were used, whether epidural or peripheral perineural techniques.

We did not aim to determine the ideal choice of RA technique for TKA. A recently published meta-analysis concluded CPNB is superior to epidural anesthesia for TKA [23]. Another systematic review also addressed this issue [22]. Furthermore, our search criteria did not extend to examining whether addition of a sciatic or other nerve block to FNB is beneficial; this subject remains a matter of controversy in the anesthetic literature [60].

Like postoperative pain, opioid-related adverse effects, especially nausea and vomiting, are a major concern to patients and can delay discharge from the hospital [51, 66]. We found Level I evidence that FNB, either alone or in combination with obturator or sciatic block, reduced opioid-related side effects. In general, a reduction in morphine consumption with RA was not always associated with a reduction in adverse effects, but this may be the result of inadequate powering. The 10 RCTs in which no benefit was observed were graded as Level II. These were either inadequately powered or had poor methodologic quality.

Although no difference in cognitive defects was observed between GA and RA, intravenous sedation was administered to the RA groups in both studies and this could have influenced results. In a separate well-conducted trial that examined patients who had THAs and patients who had TKAs as one population (and therefore was not considered in our review), there was no long-term cognitive difference between the RA and GA groups [40].

RA and regional analgesia may shorten length of stay in the hospital and hasten postoperative rehabilitation, which potentially could have important economic benefits. Although epidural analgesia and CFNB for TKA each can facilitate rehabilitation, CFNB generally is preferred, because bilateral blockade, hypotension, pruritis, bradycardia, and nausea and vomiting are avoided whereas the patient’s anticoagulation status is arguably less of a concern [6, 8, 14, 17, 32, 36, 74, 85]. However, in two of the three trials showing a reduced length of stay with RA compared with systemic analgesia, the hospital and rehabilitation center stays were longer than current practice [14, 74]. As managed clinical pathways and shorter hospital stays become increasingly prominent after TKA, so too may the role of RA. However, with short, protocolized inpatient visits and multiple confounding factors influencing discharge time, it is becoming more difficult to show reduced lengths of stay with RA [38, 68]. Reports of TKAs facilitated by ambulatory CFNB catheters at home have been published and may populate the literature in the near future [37, 38].

In conclusion, we found insufficient evidence from RCTs alone to conclude if anesthetic technique influenced mortality, cardiovascular morbidity other than postoperative hypotension, or the incidence of DVT and PE in the setting of routine thromboprophylaxis. Our systematic review does not suggest a difference in blood loss or duration of surgery in patients receiving GA and/or systemic analgesia versus RA and/or RA for TKA. However, RA does reduce postoperative pain and opioid-related adverse effects for TKA. Length of stay also may be reduced and rehabilitation facilitated by RA compared with GA.