The use of compression bandages is common after TKA. These bandages apply external pressure on the knee, thereby reducing intraarticular and extraarticular accumulation of blood and limb swelling [7, 9]. Less leg swelling may enhance recovery by reducing pain and increasing ROM . However, a prospective randomized study has shown no difference in blood loss, blood transfusion requirements, pain, or swelling when comparing the use of a modified Robert Jones compression dressing with a standard dressing . Another randomized controlled trial has shown that compression bandages reduced pain when high-volume local infiltration analgesia was used . Although the potential advantages of compression bandages are reduced swelling as well as subsequent enhanced recovery and a reduced complication rate, the evidence has been inconclusive. The potential disadvantages of these bandages are additional cost and discomfort in knee flexion > 45° as a result of the nonelastic properties of the dressing .
Currently, at our institution, it is up to the discretion of the surgeon whether to utilize a compression bandage after TKA. The compression bandage under investigation in this study is commonly referred to as an ace bandage and can be described as a large, 6-inch elastic bandage used to create localized pressure. This is the standard compression bandage used at our institution. The literature does not adequately address the use of a compression bandage of this type compared with that of no bandage when releasing the tourniquet after wound closure as a result of variation in bandage type and methodology of application.
We therefore sought to determine whether use of a compression bandage after TKA was associated with (1) less leg swelling (our primary endpoint); or (2) secondary study endpoints, including improved ROM of flexion and extension, lower visual analog scale (VAS) pain scores for worst pain and pain during physical therapy just before surgery, postoperative day (POD) 1, POD 2, and POD 28, or fewer wound complications within 90 days of surgery. The null hypothesis is that there will be no difference between the groups.
Patients and Methods
We conducted a prospective, single-center, two-arm, parallel-group randomized controlled trial (RCT), in which we enrolled 57 patients undergoing simultaneous bilateral primary TKA between February 2015 and August 2016. The specific flow of patients from enrollment to analysis has been outlined (Fig. 1). The study proposal was approved by our institutional review board and registered at clinicaltrials.gov (ID number: NCT03521869). Each patient served as his or her own control with one limb receiving an elastic compression bandage (Sterile Matrix Elastic Bandage; Medline Industries, Inc, Mundelein, IL, USA) over an Aquacel Surgical dressing (AQUACEL® Ag Surgical; ConvaTec Inc, Greensboro, NC, USA) and the other limb only receiving an Aquacel Surgical dressing.
Ages of eligible participants ranged between 40 and 83 years with a mean age of 62 years. Patients were excluded if they had a body mass index > 40 kg/m2, a previous venous thromboembolic event, or lymphedema in one or both legs. In addition, patients undergoing unicondylar arthroplasty, a surgical approach other than a medial parapatellar approach, and TKA resulting from posttraumatic arthritis were not included. Patients allergic to the Aquacel dressing or compression bandages were not included. Three adult reconstruction surgeons performed the surgeries (AFC, RHR, WJH) at a single academic institution. The left knee was always operated on first, and the same team did one knee followed by the subsequent knee. The closure technique and sequence were the same for all cases, ending with dressing with Aquacel for 1 week followed by application of a compression dressing unilaterally as the study variable.
We completed randomization on the day of surgery to select which limb would get the compression bandage and which one would not. Allocation order was created using an Excel random number generator (Excel 2013; Microsoft, Redmond, WA, USA) following simple randomization procedures. Sequentially numbered sealed envelopes were opened consecutively irrespective of the surgeon just before compression bandage application. Separate individuals performed the random allocation sequence (CR), patient recruitment and enrollment (TD), opening of randomization envelopes, and assessment of outcomes (CNM). The following individuals were blinded: research recruiter (TD), outcome assessor (CNM), and statistician (MGM). However, it was impossible to blind the patient, operating room staff, and surgeons as a result of the obvious appearance and physical sensation of the compression bandage.
Aquacel dressing was applied to both limbs after closure with the compression bandage then applied to the randomized limb just around the knee according to a standard protocol. Both bandaging techniques are approved and used routinely at our institution. The compression bandage was removed the morning of POD 1.
Components of perioperative care such as prophylactic antibiotics, surgical technique, knee implant fixation, thromboembolism prophylaxis, pain management protocol, postoperative rehabilitation, and followup intervals were similar for all patients. Preoperatively, patients received oral acetaminophen (975 mg), celecoxib (400 mg), and pregabalin (75 mg) within 2 hours of surgery. Surgeon discretion was maintained regarding implant model and manufacturer used; however, all patients received a cemented, posterior-stabilized prosthesis with patellar resurfacing. Surgical drains were not used. Of critical importance to this study, the tourniquet was not released before full wound closure. Postoperatively, standing doses of oral acetaminophen (650 mg) every 6 hours, pregabalin (75 mg) every 12 hours, and intravenous ketorolac (30 mg) were administered. Oral narcotics for residual breakthrough pain included oxycodone (10 mg) and tramadol (50 mg). Thromboembolism prophylaxis of aspirin (81 mg or 325 mg) or warfarin (target international normalized ratio, 1.5-2.0) was administered. Thirty-eight patients (75%) received tranexamic acid (20 mg/kg); however, 13 patients (25%) were not cleared to receive it by anesthesia. All patients were mobilized on the day of surgery.
Just before surgery, as well as 24 ± 4 hours after surgery, 48 ± 4 hours after surgery, and on POD 28 ± 7 days, an individual not involved with care (CNM) measured all patients for the following: circumference of each leg, ROM on each leg, and VAS pain scores for each leg. With a tape measure, we measured thigh circumference (10 cm above the proximal pole of the patella); knee circumference (at the center of the patella); and shin circumference (10 cm below the distal pole of the patella). We used a goniometer to measure the ROM of both knees (the superior arm aligned with the greater trochanter of the femur, the axis aligned with the middle of the patella, and the inferior arm aligned with the lateral malleolus) (see Appendix, Supplemental Digital Content 1). VAS pain scores included the worst pain in the past 24 hours and pain during physical therapy using a VAS scale of 0 (no pain) to 100 mm (worst pain possible). Patients were instructed to indicate their pain levels by marking an X on the scale, which we then measured with a ruler (for the VAS pain questionnaire, see Appendix, Supplemental Digital Content 2). Wound healing was assessed by retrospectively reviewing patient medical charts, phone records, and visits within 90 days after surgery. Wound complications recorded were erythema, drainage, and the type of drainage classified as serous or bloody.
Review of the literature [1, 3, 8, 10] indicated that the minimal clinically important difference (MCID) for circumference was 2 cm with a SD of 2 cm in the circumference. For VAS, it was 2 points with a SD of 2. For ROM, it was 10° with a SD of 15. We conservatively picked an effect size of 0.5 SD and assumed a correlation between limbs of 0.3. This set the power level at 0.80 with an α error of 0.05; thus, a power analysis for paired t-tests indicated that 45 patients would be an appropriate sample size.
For the statistical analysis, we used R software 3.5 (R Foundation for Statistical Computing, Vienna, Austria) on 51 patients. We utilized a paired t-test to compare the measurements for the limb that received the compression bandage and the limb that did not at each time interval separately for circumference, ROM, VAS worst pain, and VAS pain during physical therapy. We used a linear mixed model to compare the measurements for the limb that received the compression bandage and the limb that did not for each participant across time. Each outcome (circumference, ROM, VAS worst pain, and VAS pain during physical therapy) had time in days, randomization of the compression bandage, and preoperative measurement as predictors. We performed a Fisher’s exact test to determine the difference between wound complications among the groups. Probability values < 0.05 were considered statistically significant.
Although 57 patients undergoing bilateral TKA consented to participate, 51 patients received the intervention and were included in the statistical analyses. This included 29 patients randomized to receive the compression bandage on the right leg and 22 patients randomized to receive the compression bandage on the left leg. Patients were excluded for a variety of reasons ranging from voluntary drop, allergy to the dressing, to surgery during their postoperative stay (Table 1).
There was no difference between the limb with and without the compression bandage regarding demographics, comorbidities, and operative time, because each patient served as his or her own matched pair (Table 2). In addition, there were no differences preoperatively between the limb with and without the compression bandage in terms of circumference, ROM, or pain (Table 3).
We found no difference at any time point in circumference measurements between the limbs treated with a compression bandage and those not so treated. Specifically, there were no differences between the groups in terms of leg swelling at the thigh (POD 1: mean ± SD = 51 ± 6 with compression bandage versus mean ± SD = 51 ± 6 without compression bandage, mean Δ = -0.14, 95% confidence interval [CI], -0.65 to 0.37, p = 0.586; POD 2: mean ± SD = 53 ± 6 with compression bandage versus mean ± SD = 53 ± 7 without compression bandage, mean Δ = -0.22, 95% CI, -0.95 to 0.51, p = 0.548; POD 28: mean ± SD = 47 ± 6 with compression bandage versus mean ± SD = 47 ± 6 without compression bandage, mean Δ = -0.01, 95% CI, -0.39 to 0.38, p = 0.975; adjusted mean Δ = -0.31, 95% CI, -0.81 to 0.18, p = 0.219), at the knee (POD 1: mean ± SD = 45 ± 4 with compression bandage versus mean ± SD = 45 ± 5 without compression bandage, mean Δ = -0.44, 95% CI, -1.16 to 0.28, p = 0.223; POD 2: mean ± SD = 46 ± 4 with compression bandage versus mean ± SD = 46 ± 4 without compression bandage, mean Δ = -0.30, 95% CI, -0.69 to 0.10, p = 0.137; POD 28: mean ± SD = 42 ± 5 with compression bandage versus mean ± SD = 42 ± 5 without compression bandage, mean Δ = 0.21, 95% CI, -0.34 to 0.76, p = 0.446; adjusted mean Δ = -0.14, 95% CI, -0.55 to 0.27, p = 0.509), and at the shin (POD 1: mean ± SD = 40 ± 4 with compression bandage versus mean ± SD = 40 ± 4 without compression bandage, mean Δ = -0.22, 95% CI, -1.23 to 0.79, p = 0.659; POD 2: mean ± SD = 41 ± 4 with compression bandage versus mean ± SD = 41 ± 4 without compression bandage, mean Δ = -0.31, 95% CI, -0.72 to 0.09, p = 0.126; POD 28: mean ± SD = 37 ± 4 with compression bandage versus mean ± SD = 37 ± 4 without compression bandage, mean Δ = -0.34, 95% CI, -0.92 to 0.24, p = 0.246). However, there was a difference found in the mixed model showing less swelling in the limb with the compression bandage compared with that without the compression bandage across all time points (adjusted mean Δ = -0.49, 95% CI, -0.96 to -0.02, p = 0.045). Although the p value was significant, the mean difference did not reach the MCID (Table 4). We observed similar trends for circumference measurements of the thigh, knee, and shin in the limb with the compression bandage and the limb without the compression bandage when comparing preoperative measurements with POD 1, POD 2, and POD 28, respectively. The thigh, knee, and shin circumference increased from preoperative values to POD 1; however, we observed the most noticeable increase in all circumference measurements from preoperative values on POD 2. At POD 28 the thigh, knee, and shin circumference measurements were the smallest in the postoperative period and were similar to the preoperative values (Figs. 2, 3, and 4, respectively).
There were no differences between the limb that received the compression bandage and the limb that did not in terms of flexion ROM (POD 1: mean ± SD = 56 ± 25 with compression bandage versus mean ± SD = 58 ± 22 without compression bandage, mean Δ = -2.63, 95% CI, -7.01 to 1.76, p = 0.234; POD 2: mean ± SD = 64 ± 20 with compression bandage versus mean ± SD = 63 ± 23 without compression bandage, mean Δ = 1.22, 95% CI, -2.69 to 5.13, p = 0.534; POD 28: mean ± SD = 101 ± 20 with compression bandage versus mean ± SD = 102 ± 20 without compression bandage, mean Δ = -1.64, 95% CI, -3.63 to 0.35, p = 0.103; adjusted mean Δ = -1.17, 95% CI, -5.18 to 2.83, p = 0.568) and extension ROM (POD 1: mean ± SD = 12 ± 7 with compression bandage versus mean ± SD = 12 ± 7 without compression bandage, mean Δ = 0.51, 95% CI, -0.55 to 1.57, p = 0.328; POD 2: mean ± SD = 9 ± 5 with compression bandage versus mean ± SD = 10 ± 6 without compression bandage, mean Δ = -1.28, 95% CI, -2.63 to 0.06, p = 0.061; POD 28: mean ± SD = 6 ± 14 with compression bandage versus mean ± SD = 4 ± 4 without compression bandage, mean Δ = 2.19, 95% CI, -1.60 to 5.97, p = 0.252; adjusted mean Δ = 0.40, 95% CI, -1.20 to 1.99, p = 0.629). With the numbers available, we observed greater maximal postoperative pain for the limb with the compression bandage than the control limb on POD 1 and POD 2, but not on POD 28 or across time (POD 1: mean ± SD = 8 ± 3 with compression bandage versus mean ± SD = 7 ± 3 without compression bandage, mean Δ = 0.66, 95% CI, 0.07-1.26, p = 0.030; POD 2: mean ± SD = 7 ± 2 with compression bandage versus mean ± SD = 7 ± 3 without compression bandage, mean Δ = 0.80, 95% CI, 0.22-1.39, p = 0.008; POD 28: mean ± SD = 4 ± 3 with compression bandage versus mean ± SD = 3 ± 3 without compression bandage, mean Δ = 0.14, 95% CI, -0.30 to 0.58, p = 0.526; adjusted mean Δ = 0.50, 95% CI, -0.02 to 1.02, p = 0.061) (Fig. 5). Similarly, there was greater pain during physical therapy for the limb with the compression bandage than the limb without on POD 2, but not on POD 1, POD 28, or across time (POD 1: mean ± SD = 7 ± 3 with compression bandage versus mean ± SD = 6 ± 3 without compression bandage, mean Δ = 0.29, 95% CI, -0.50 to 1.08, p = 0.460; POD 2: mean ± SD = 8 ± 2 with compression bandage versus mean ± SD = 7 ± 3 without compression bandage, mean Δ = 0.67, 95% CI, 0.12-1.22, p = 0.018; POD 28: mean ± SD = 5 ± 2 with compression bandage versus mean ± SD = 5 ± 3 without compression bandage, mean Δ = 0.14, 95% CI, -0.40 to 0.68, p = 0.600; adjusted mean Δ = 0.35, 95% CI, -0.16 to 0.86, p = 0.183). With the numbers available, we observed no difference in 90-day wound healing complications between the limb with and the limb without the compression dressing; however, the sample size was too small to analyze this in a meaningful statistical way. Overall, there were 6% (three of 51) total wound complications in the limb with the compression bandage and 12% (six of 51) total wound complications in the limb without the compression bandage (odds ratio [OR], 0.47; 95% CI, 0.07-2.37; p = 0.487). Drainage was not observed in the group with the compression bandage, whereas the group without the compression bandage had 6% (three of 51) drainage (OR, 0; 95% CI, 0-2.39; p = 0.243) (Table 5). There were no deep infections or reoperations within 90 days postoperatively.
The appropriate management of swelling has an essential role in the outcome of any surgical procedure and postoperative recuperation; compression bandages are one method that has been used to reduce swelling. According to Silver et al. , a limb will swell by approximately 10% of its original volume after release of a tourniquet as a result of return of exsanguinated blood. According to a meta-analysis by Rama et al. , the evidence supports early tourniquet release before closure because it allows the surgeon to secure homeostasis and to more easily recognize major vascular injuries. Therefore, if the tourniquet was released before wound closure, eschewing the compression bandage was less controversial because the surgeon would have been able to identify and coagulate bleeding vessels before closure. As a result, the effects of a compression bandage have not been considered instrumental. In 2015, Heller et al.  illustrated that releasing the tourniquet after wound closure but before applying a compression bandage reduced blistering without increasing postoperative blood loss when using calculated blood loss to account for blood extravasation into soft tissues, bleeding into the joint, and hemolysis. Therefore, we feel it is reasonable to question the value of a compression dressing even if the tourniquet is not released until full wound closure has been obtained. This study sought to determine whether use of a compression bandage after TKA was associated with less leg swelling, improved ROM, lower VAS pain scores, or fewer wound complications.
There are several limitations to this study. First, we were not able to blind the patients, operating room staff, and surgeons. However, any study evaluating the use of a compression bandage after TKA will be subject to the same limitations as a result of its appearance and physical presence. Second, blood loss was not recorded for each knee. No drains were used; therefore, we are only able to measure total blood loss. Third, the sample size was underpowered to adequately analyze wound complications and identify less frequent TKA-related complications. Fourth, we want to acknowledge that there was an unfortunate vascular injury, which represented 0.8% (one of 114) of the 114 TKAs performed on the 57 patients undergoing bilateral TKA who were consented in this study. This type of injury is a potential complication of any TKA and is unlikely related to the procedures used in this study. This type of injury is usually caused by manipulation of the vessels during surgery and manifests as loss of pulse distal to the injury. This loss of pulse is generally discovered in the recovery room. Finally, the pressure of the compression bandage was not recorded. Consequently, we are unsure if the same effects from pressure were standardized across all study participants. However, it should be noted that in standard practice, compression bandages are applied without pressure recordings. Therefore, the same variation or lack thereof in pressure would be found in our study as in regular practice.
We found no clinical difference in postoperative lower extremity swelling whether or not a compression bandage was used. Recently, Brock et al.  compared the use of an inelastic compression bandage from toe to groin and a wool and crepe bandage of the knee applied for 24 hours after TKA. They found no difference in knee swelling between the two groups, which neither supports nor refutes the use of an elastic compression bandage. However, it should be noted that the elastic compression bandages utilized in this study and the current study are different. Kayamori et al.  conducted a RCT comparing use of a polyethylene foam pad wrapped with an elastic bandage for 7 days after TKA and an elastic bandage with no foam pad until 10 days after TKA with no difference in swelling. The time period of application was much longer in the study by Kayamori et al. . The current study further investigated use of an elastic compression bandage versus no bandage, which not only agrees with previous studies, but further extends the evidence that there is no benefit of an elastic compression bandage in terms of swelling. Because the clinical studies available on bandages after TKA assess differing types of bandages, positive and negative controls, size of application, and length of use, it would be of benefit to further investigate this topic with multiple groups for each of these factors in one study design.
We found no difference in postoperative knee flexion ROM and extension ROM whether or not a compression bandage was used. Brock et al.  and Kayamori et al.  also found there to be no difference in ROM between their respective groups assessed. However, Cheung et al.  found improved flexion ROM when comparing an inelastic compression bandage from toes to groin and a crepe bandage around the knee when applied 24 hours postoperatively in an enhanced recovery setting. In addition, we found differences in postoperative maximal pain and pain during physical therapy between the bandaged limb and the unbandaged limb. According to Brock et al. , they found no difference in pain, which was recorded before and after physical therapy until discharge and at 6 weeks. These findings still align with the current study because the MCID was not reached. Lastly, we found no difference in postoperative wound complications between the limb with the compression bandage and the limb without the compression bandage. Brock et al.  and Kayamori et al.  found no wound complications either. However, as stated in the previous paragraph, these other studies mentioned did not investigate an elastic compression bandage versus no bandage as the current study does. In addition, methodology used and type of bandage utilized differ in the literature, which is why a comprehensive prospective study should be used.
Applying a compression bandage after TKA did not result in any clinical improvement of the measured outcomes when analyzed with each patient undergoing simultaneous bilateral primary TKA serving as their own matched pair, even when the tourniquet was not released before wound closure. Therefore, we believe applying compression bandages after TKA is not necessary, and we no longer use compression dressings for routine primary TKA.
We thank the operating room staff (for consultancy and assistance with the logistics), Tiffany Morrison (for consultancy regarding institutional review board approval), Snir Heller MD (for consultancy regarding study design), and Camilo Restrepo MD (for consultancy regarding study design, logistics, and statistical analysis).
1. Andersen LØ, Husted H, Otte KS, Kristensen BB, Kehlet H. A compression bandage improves local infiltration analgesia in total knee arthroplasty. Acta Orthop. 2008;6:806–811.
2. Brock TM, Sprowson AP, Muller S, Reed MR. STICKS study–Short-sTretch Inelastic Compression bandage in Knee Swelling following total knee arthroplasty–a feasibility study. Trials. 2017;1:6.
3. Charalambides C, Beer M, Melhuish J, Williams RJ, Cobb AG. Bandaging technique after knee replacement. Acta Orthop. 2005;1:89–94.
4. Cheung A, Lykostratis H, Holloway I. Compression bandaging improves mobility following total knee replacement in an enhanced recovery setting. J Perioper Pract. 2014;4:84–86.
5. Heller S, Chen A, Restrepo C, Albert E, Hozack WJ. Tourniquet release prior to dressing application reduces blistering following total knee arthroplasty. J Arthroplasty. 2015;7:1207–1210.
6. Kayamori S, Tsukada S, Sato M, Komata K, Isida Y, Wakui M. Impact of postoperative compression dressing using polyethylene foam pad on the multimodal protocol for swelling control following total knee arthroplasty: a randomized controlled trial. Arthroplast Today. 2016;4:199–204.
7. Matsen FA III, Krugmire RB Jr. The effect of externally applied pressure on post-fracture swelling. J Bone Joint Surg Am. 1974;8:1586–1591.
8. Munk S, Jensen NJ, Andersen I, Kehlet H, Hansen T. Effect of compression therapy on knee swelling and pain after total knee arthroplasty. Knee Surg Sports Traumatol Arthrosc. 2013;2:388–392.
9. Ogata K, Whiteside LA. Effects of external compression on blood flow to muscle and skin. Clin Orthop Relat Res. 1982;168:105–107.
10. Pinsornsak P, Chumchuen S. Can a modified Robert Jones bandage after knee arthroplasty reduce blood loss? A prospective randomized controlled trial. Clin Orthop Relat Res. 2013;471:1677–1681.
11. Rama KR, Apsingi S, Poovali S, Jetti A. Timing of tourniquet release in knee arthroplasty: meta-analysis of randomized, controlled trials. J Bone Joint Surg Am. 2007;89:699–705.
12. Silver R, de la Garza J, Rang M, Koreska J. Limb swelling after release of a tourniquet. Clin Orthop Relat Res. 1986;206:86–89.