Comparison of efficacy and safety of different tourniquet applications in total knee arthroplasty: a network meta-analysis of randomized controlled trials

Abstract Objective Total knee replacement (TKA) is an effective way to treat teratogenic and disabling knee diseases such as advanced osteoarthritis. Tourniquets are often used in TKA to reduce bleeding and to get a better visualization of the surgical field, while it is related to safety concerns. We did this network meta-analysis to comprehensively compare the efficacy and safety of various tourniquet application strategies. Method PubMed, Embase, Cochrane Library, CNKI, and WanFang Database were systematically searched from January 1990 to May 2020. A network meta-analysis with a frequentist framework was done to assess the relative efficacy and safety by comparing seven clinical important endpoints. Results 38 eligible studies that assessed 3007 participants who underwent TKA were included in this network meta-analysis. Tourniquet inflation before osteotomy then deflation after wound closure effectively reduce perioperative bleeding (WMD compared with control group −234.66, 95% CI [–409.19 to −60.13]), while shortening the operation time (WMD −8.98, 95%CI [–14.07 to −3.88]) and reducing postoperative complications, including DVT (OR −0.58, 95%CI [–1.19 to 0.03]) and minor wound complications (OR −1.38, 95%CI [–3.00 to 0.25]). No difference was found in the late postoperative knee pain and function outcomes. Conclusions Using tourniquets during the entire operation can effectively reduce blood loss, but it also can cause many safety problems, including DVTs, wound oozing, delayed healing, and serious wound complications. Tourniquet inflation before osteotomy then deflation after wound closure effectively can reduce perioperative bleeding while shortening the operation time and reducing postoperative complications, so it could be the ideal tourniquet application strategy in TKA. Key messages This is the first study that comprehensively compared different tourniquet application strategies to evaluate their impact on postoperative recovery following TKA, and five clinically important endpoints were assessed in this study: perioperative blood loss, operation time, postoperative pain and function, and complications. We conclude that tourniquet inflation before osteotomy then deflation after wound closure could be the ideal tourniquet application strategy in TKA.


Introduction
Knee osteoarthritis (KOA) and rheumatoid arthritis (RA) are common knee diseases that cause joint pain and loss of function. Pharmacological treatments may be used early to improve symptoms, but for end-stage patients, total knee replacement (TKA) is often the only effective intervention [1,2].
Some studies have found significantly more intraoperative bleeding [3], which can further lead to the deterioration of surgical visualization, the extension of surgical time, and can interfere with the bonding effect of bone cement. To solve these difficulties, tourniquets are widely used during TKA. However, it was associated with a higher risk of complications, including Thigh pain and swelling, paresthaesia, vascular injury, and venous thromboembolism, subcutaneous fat necrosis, postoperative stiffness, poor wound healing, and delayed recovery of quadriceps strength [4][5][6][7].
Previous studies have confirmed that the prolonged application of a tourniquet is a key factor in complications that may be associated with prolonged ischaemic time for tissues [7][8][9]. More and more researchers have begun to minimize tourniquet time to determine whether the limited application of tourniquets in TKA can reduce complications and promote postoperative recovery.
Kvederas G et al. [10] compared the efficacy and safety of three tourniquet application strategies (Tourniquet inflation before incision then deflation after wound closure; inflation before incision then deflation after cement hardening; and inflation just during cementing). Their study demonstrated that tourniquet inflation before incision then deflation after cement hardening tends to give better outcomes. Inflation only during cementing was associated with the greatest total blood loss following TKA. Tai TW et al. [11] compared perioperative blood loss and postoperative pain between two groups with or without a tourniquet. They reported that the use of a tourniquet during TKA effectively reduces perioperative blood loss but was related to slightly more postoperative pain that did not affect recovery. A shorter duration of tourniquet application resulted in a faster recovery and less pain during the early postoperative period and a lower incidence of complications following TKA [12,13], but was associated with a significantly greater risk of blood loss and transfusion [14]. Therefore, the optimal timing of tourniquet application is still highly controversial and should be balanced with the increased blood loss and risk of transfusion with postoperative complications when using the tourniquet in TKA.
Until now, six different tourniquet applications have been proposed clinically, including non-use of a tourniquet (NTU); tourniquet inflation before incision than deflation after wound closure (throughout the operation, TTO); tourniquet inflation before incision and deflation after cement hardening (first half of operation, FHO); tourniquet inflation before osteotomy and deflation after wound closure (before osteotomy in operation, BOO); tourniquet inflation after osteotomy and deflation after wound closure (second half of operation, SHO); and tourniquet inflation only during cementing (middle of operation, MO).
Previously, numerous meta-analyses in this field have been published. For instance, Alcelik I et al. [15] conducted a meta-analysis of outcomes with and without a tourniquet in TKA, whereas Wang C et al. [16] presented a meta-analysis comparing tourniquet application only during cementation and long-duration tourniquet application in TKA. However, there still has been no systematic review evaluating the relative efficacy and safety of the various tourniquet application strategies or comparisons of these strategies to cases without tourniquet application in TKA. Accordingly, to comprehensively quantify the clinical efficacy and safety of different tourniquet applications and select the best strategy, we designed and conducted this network meta-analysis.

Data Sources and searches
This network meta-analysis was designed and conducted based on the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines (PRISMA) [17].
PubMed, Embase, Cochrane Library, CNKI, and WanFang Database were systematically searched from January 1990 to May 2020, with the search terms (("tourniquet" OR "tourniquet time" OR "tourniquet application" OR "tourniquet release") OR ("ischaemic" OR "ischaemic injury" OR "tourniquet related ischaemic injury") AND ("knee arthroplasty" OR "knee replacement" OR "TKA")). Reference lists of relevant systematic reviews and identified articles also were reviewed to find additional eligible studies as completely as possible. No restriction was placed on the language of publication.

Study selection
Any article which met the following inclusion criteria were included: 1. The patients included in the study underwent total knee replacement or arthroplasty; 2. The study included a comparison among two or more different tourniquet release strategies; 3. The study used a randomized clinical trial design, whether blinded or not; 4. The study reported at least one desirable outcome: operation time, blood loss, functional rehabilitation, pain relief, the incidence of DVT, or other complications.
The exclusion criteria for this study were: 1. The study included another type of surgery, including unicompartmental knee arthroplasty or osteotomy; 2. Low-quality studies according to the Cochrane Handbook; 3. All animal studies and cadaver studies; 4. All reviews, systematic reviews and meta-analyses, conference abstracts, letters, and those without original study data.

Data extraction and quality assessment
Two authors reviewed the full manuscripts of all eligible studies and extracted relevant data from the studies, including author name, publication year, patient number, diagnosis, mean age, gender, surgical approach, whether to use tranexamic acid, whether to use drainage, whether to use anticoagulation and outcomes data. To avoid the influence of withdrawal bias, we preferred to extract data from the intention-totreat analysis if available.
Two other authors conducted the methodological quality and bias assessment of studies using the Cochrane risk of the bias assessment tool [18]. The following indices were evaluated and ranked the as low risk of bias, unclear risk of bias, or high risk of bias: Sequence generation, allocation concealment, blinding, incomplete outcome data, selection outcome reporting, and other sources of bias. All disputes were resolved through discussion.

Outcome measures
Operation time, blood loss, postoperative pain, and function scale and safety endpoints were chosen as outcome measurements considered important for decision making. Blood loss included intraoperative, postoperative, and total blood loss, which were compared among groups to identify the efficacy of each tourniquet application strategy. The weighted mean difference (WMD) with 95% confidence intervals (CI) was used to measure the operation time and blood loss.
Early postoperative pain/function status (within one week after surgery) and late status (at least one month after surgery) were compared. No restriction was placed on the types of questionnaire used in pain evaluation. The function subscales of the Western Ontario and McMaster Universities Arthritis Index (WOMAC) were used to evaluate postoperative functional status. If the WOMAC function score was not measured or reported, the Lequesne Index or other functional measurement scales were used. The standardized mean difference (SMD) was used because results from different scales were included in the same network. Additionally, the range of knee flexion was compared to reflect the postoperative functional recovery.
Safety endpoints comprised DVT, minor and major wound complications. Minor wound complications included wound oozing, erythema, cellulitis, minor dehiscence, and superficial infection. Major wound complications included any condition that led to surgical failure, required a second surgery, or caused disability or death. The odds ratio [OR] with 95% confidence intervals (CI) was used to measure relative safety.

Statistical analysis
Traditional pairwise meta-analysis was conducted to compare the efficacy of the target drugs with placebo using RevMan (Review Manager. Version 5.3, Copenhagen, The Nordic Cochrane Centre, The Cochrane Collaboration, 2014). The heterogeneity across studies was tested by the Q and I2 statistics, in which p < .05 or I 2 > 50% implies significant heterogeneity. A random-effects model was used if significant heterogeneity across studies was found, otherwise, a fixed-effects model was preferred.
The random-effects network meta-analysis was conducted using Stata/MP (version 14.0) in a frequentist method. The multivariate random-effects meta-regression was used to pool data with a proportional variance-covariance matrix, and a restricted maximumlikelihood method was used to assess model fit [19]. Global inconsistency tests and node-split tests were conducted, and the consistency model was adopted if no inconsistency was reported (p-value of Z-test >.05). If available, loop-specific inconsistency tests were also conducted to evaluate the inconsistency between direct and indirect comparisons within every closed triangle or quadratic loop. Inconsistency factors (IFs) and their 95% CI reported by loop-specific inconsistency tests reflect the consistency [20]. If inconsistency was reported in any network, the sensitivity analysis was used to identify the source of inconsistency and exclude these studies from this network. Funnel plots and Egger's tests were conducted in Stata/MP to detect publication bias for each endpoint. The trim and filling method was used for further analyses to confirm whether significant publication bias was present [21]. The surface under the cumulative ranking (SUCRA) was used to rank the efficacy and safety of different drugs [22]. To select the most effective and safest drug simultaneously, cluster-ranking plots were constructed. Significant differences were considered between treatments when the corresponding 95% CI did not contain 1 for OR. p < .05 was considered statistically significant.

Literature selection
Thirty-eight eligible studies were identified [10][11][12][13]. The PRISMA flow diagram is shown in Supplementary figure 1. Six different tourniquet strategies (NTU, TTO, FHO, MO, BOO, and SHO) were analyzed, and the NTU was selected as the standard control group. The network plot is presented in Figure 1.

Study characteristics
The mean age of the 3007 patients included in this study was 67.61 years (IQR 64.66 to 70.50 years) and the proportion of male patients was 31.79% (ranged from 10.91% to 53.23%) across studies. Most participants received primary TKA due to KOA or RA, except for five patients who underwent revision surgery. The medial parapatellar approach was the most common surgical approach (26 articles), followed by the midvastus approach (4 articles) and sub-vastus approach (1 article). Seven articles did not report the surgical approach used. Other baseline characteristics are presented in Supplementary Table 1. The details of quality and bias-risk assessments are presented in Supplementary Table 2. Based on these results, the main contributing factors to the risk of bias were performance bias, selection bias, and attrition bias.
Publication bias was not reported in any network according to the results of funnel plots and Egger's tests (Supplementary Figure 2). Results of cluster-rank analyses are presented in Supplementary Figure 3. The detailed results of conventional pair-wise metaanalyses are presented in Table 1. The forest plots of network meta-analyses are presented in Figure 2. The league plots which showed the relative effects between different groups are presented in Figure 3.
Detailed SURCA values were presented in Supplementary Tables 3-5.

Operation time
3.3.1. Conventional pair-wise meta-analysis Significant heterogeneity of the included studies and interventions was reported (I 2 ¼ 79.9%), so the random-effects model was adopted. The use of a tourniquet overall was associated with a significantly shorter operation time than non-use (WMD À4.00, 95% CI [-7.14 to À0.84]).

Network meta-analysis
A total of 33 articles were analyzed in this network. However, significant inconsistency was reported by the global inconsistency test (p < 0.05). In the nodesplit test and loop-specific inconsistency tests, inconsistency was found in the NTU -FHO -BOO loop ( 5,9,16,22). Following this, no inconsistency was reported in the reconstructed network and the consistency model was adopted.

Blood loss
3.4.1. Conventional pair-wise meta-analysis Significant heterogeneity was found in intraoperative blood loss (I 2 ¼ 88.2%), postoperative blood loss (I 2 ¼ 59.2%) and total blood loss (I 2 ¼ 92.2%) so the random-effects model was adopted.
The use of a tourniquet overall was associated with a significantly decreased intraoperative blood loss compared to non-use of a tourniquet (WMD À142.86, 95% CI [-175.05 to À110.67]), but also with increased postoperative blood loss (WMD 81. 28

Network meta-analysis
Thirty-three articles were assessed in this network. No significant inconsistency was reported by the global inconsistency tests, node-split tests, or loop-specific inconsistency tests for the three blood loss indicators.

Network meta-analysis
Seventeen articles were assessed in this network. The consistency model was adopted because no significant inconsistency was reported by global inconsistency, node-split, or loop-specific inconsistency tests.
The SHO group demonstrated the most obvious early postoperative pain compared with the NTU group (SMD 1.10, 95% CI [0.13 to 2.08]) and any of the other tourniquet application groups. The BOO group had the greatest probability of being the best option (SUCRA ¼ 96.5%), followed by SHO (SURCA ¼ 61.8%) and TTO (SURCA ¼ 53.9%). No significant differences were found in the late postoperative pain network. BOO had the largest probability of being the best option (SMD À0. 42

Network meta-analysis
Seventeen articles with postoperative functional status and 15 articles with postoperative knee flexion range were assessed in this network. The consistency model was adopted for the early/late postoperative functional status because no inconsistency was reported in the global inconsistency, node-split, or loop-specific inconsistency tests. However, significant inconsistency was found in the postoperative knee flexion range. Based on the loop-specific inconsistency test, inconsistency was found in the TTO -MO -SHO loop ( 28.71], p ¼ .000). Sensitivity analysis was conducted, and one study was excluded from this network (study 37). Following this, no inconsistency was reported in the reconstructed network and the consistency model was adopted.
Similar to the results obtained by pair-wise metaanalysis, network meta-analysis did not find any significant differences in functional outcomes. According to the SURCA rank, theoretically, the best strategies for positive functional outcomes were SHO (knee flexion range, WMD 1.

Network meta-analysis
Twenty-two articles were included in the safety network. The consistency model was adopted because no significant inconsistency was reported. Considering the actual clinical situation, the TTO group was chosen as the control group for this part of the analysis.
Consistent with expected results, the TTO group had a higher DVT incidence than any other group (the differences were all statistically significant

Discussion
This is the first study that comprehensively compared different tourniquet application strategies to evaluate their impact on postoperative recovery following TKA.
Previous systematic reviews have indicated that the use of tourniquets can reduce surgical bleeding to a certain extent, but can cause vascular endothelial injury or ischemia-reperfusion injury of the lower limbs, resulting in poor wound healing, DVT, swelling, increased postoperative pain, and loss of function.
To comprehensively weigh the pros and cons of tourniquet use to help clinicians make decisions, five clinically important endpoints were assessed in this study: perioperative blood loss, operation time, postoperative pain and function, and complications. All relevant RCTs were screened and 38 eligible RCTs were included in the study. A frequentist framework network meta-analysis was conducted. This method integrates all available direct or indirect evidence from RCTs comparing different tourniquet application strategies, thereby augmenting the number of studies within each comparison, which increases the power of the study. For instance, only five of the included studies compared SHO with other strategies, but through the network, there were 38 studies allowing six indirect comparisons.
The main results of this analysis revealed that: (1) BOO and TTO result in shorter operation times. Probably for this reason they also had less intraoperative and total blood loss; (2) TTO had more postoperative blood loss, while BOO had the least postoperative bleeding; (3) SHO may cause more early postoperative pain, but no difference was found in late postoperative pain; (4) BOO had the lowest early and late postoperative pain, but could negatively affect early functional recovery following TKA; (5) TTO was related to a greater incidence of DVT, as well as minor and major complications. BOO had the lowest risk of postoperative minor complications, whereas SHO had the lowest risk of postoperative DVT and major complications; (6) Regardless of whether a tourniquet is used or how it is used, the late knee pain and function following TKA will not be affected; (7) According to the results of the cluster-rank analysis, it is safer and more effective to use the BOO strategy to prevent DVT and to reduce bleeding and minor complications simultaneously, but SHO is more suitable for patients who are at high risk of major complications.
Alcelik I et al. [15] supposed that using a tourniquet to decrease blood loss may involve two main effects. First, bleeding at the surgical site might be reduced, consequently reducing blood loss during surgery.
Second, if there is little bleeding from the wound, the surgeon could spend less time controlling the bleeding, so the operation time should be shortened. According to our results, TTO and BOO were associated with shorter operation time and less intraoperative and total blood loss. This is in line with Alcelik I's speculation, but we believe that shortening the operation time as much as possible and closing the wound in advance also has a positive effect on reducing intraoperative blood loss. Theoretically, the use of a tourniquet may cause calcification of the vascular endothelium, blood flow stasis of the lower limbs, and ischemia-reperfusion injury to potentially increase the risk of postoperative complications. Previous studies have proven that a higher rate of complications was related to longer tourniquet times [57][58][59]. Olivecrona C et al. found that every additional 10 min of tourniquet time was associated with an increased risk for complications [60]. Because using a tourniquet throughout the surgery can reduce blood loss effectively, it is indeed related to many safety issues. Thus, it is crucial to find a balance between reducing surgical bleeding and shortening surgical time, i.e. the best tourniquet application strategy. According to our results, tourniquet inflation before osteotomy then deflation after wound closure may be the ideal choice, with the lowest perioperative blood loss and consequently the shortest operation time and the lowest risk of complications.
There are several limitations to this study. First, there are some inevitable confounding factors that may affect the stability of the results. For instance, the skills of the surgeons will undoubtedly affect the operation time and blood loss. It is an uncontrollable variable when comparing these studies and is impossible to control or eliminate by statistical means. Second, considering the unmanageable confounding factors in observational studies, only RCTs were included even though observational studies also can play an important role in evaluating the effectiveness and safety of tourniquet application following TKA. This may contribute to the small number of studies we could include. Although funnel plots and Egger's tests were performed in this study and no significant publication bias was found, bias can be a potential problem especially when the funnel plots showed a dubious asymmetry. Finally, the DVT and major complication networks inevitably involved rare event rates. Although the Cochrane Handbook recommends the omission of studies with no events in both arms, it is still controversial whether this exclusion reduces the coverage rates of confidence intervals and increases the bias estimates. Previous studies have shown that including such trials produces unbiased estimation with narrow confidence intervals, improving the accuracy of combined estimation in a meta-analysis. The results of this should be interpreted cautiously. More high-quality trials are needed.

Conclusion
Using tourniquets during the entire operation can effectively reduce blood loss, but it also can cause many safety problems, including DVTs and wound complications. Tourniquet inflation before osteotomy then deflation after wound closure effectively can reduce perioperative bleeding while shortening the operation time and reducing postoperative complications. We conclude that tourniquet inflation before osteotomy then deflation after wound closure could be the ideal tourniquet application strategy in TKA.

Author contributions
YJL and JHW conceived the study, participated in its design and coordination, and critically revised the manuscript. QXL and JHW searched the databases. JG and YJL reviewed the full manuscripts of all eligible studies and finished the data collection. YJL and ZQC conducted the quality assessment of studies. ZQC and YJL finished the data analysis. ZQC drafted the manuscript. YJL, JHW, and ZQC had full access to all the data collection, analysis, and interpretation. All authors read and approved the final manuscript.

Ethical approval
This article does not contain any studies with human participants or animals performed by any of the authors.

Disclosure statement
No potential conflict of interest was reported by the author(s).

Funding
This study was supported by the National Key R&D Program of China (No, 2019YFA0112100), the National Natural Science Foundation of China (No. 81472073), and the Natural Science Foundation of Hunan Province (2019JJ40518).