Anesthesia Method, Tourniquet Use, and Persistent Postsurgical Pain after Total Knee Arthroplasty: A Prespecified Secondary Analysis of a Randomized Trial

This study was a secondary analysis of a trial that was approved by the ethics committee of HUS Helsinki University Hospital, Surgery (Helsinki, Finland) (June 8, 2016; reference No. HUS1703/2016) and the Finnish Medicines Agency Fimea (Kuopio, Finland) (May 20, 2016; reference No. KL72/2016). The study was registered to EudraCT (reference No. 2016-002035-15, principal investigator: Anne Vakkuri) on May 12, 2016. We obtained written informed consent from every participant.

This study was a secondary analysis of an open-label, parallel, longitudinal, single-center, randomized controlled trial conducted at the Arthroplasty Center of HUS Helsinki University Hospital, Vantaa, Finland. A separate study description and in-hospital results, including the primary outcome concerning the use of intravenous (IV) oxycodone with a patient-controlled analgesia device, have been published.^25,26 

Patients aged 18 to 75 yr with a body mass index of 40 kg/m² or less and who were referred for primary total knee arthroplasty due to Kellgren–Lawrence grade III to IV knee arthritis were included in the study. We excluded patients with severe malalignment, extension or flexion deficit, or previous major surgery of the same knee. In addition, we excluded patients with American Society of Anesthesiologists (Schaumburg, Illinois) physical status class IV or higher, those with contraindications to the study medications or to either anesthesia method, those with a need for bridging anticoagulation, those who underwent a bilateral operation, or those with ongoing usage of strong opioids.²⁶ 

Sealed and opaque randomization envelopes were created in blocks of 20 by a nonparticipating physician. The patients were randomized into the following four groups in a 1:1:1:1 allocation ratio: spinal anesthesia + no tourniquet, spinal anesthesia + tourniquet, general anesthesia + no tourniquet, and general anesthesia + tourniquet. The envelopes were opened at the earliest 2 h before the operation by a nurse not associated with the study. Blinding the medical staff or patients was infeasible.

Premedication included 5 mg oral diazepam, 1 g acetaminophen, and 400 to 800 mg ibuprofen according to the ideal body weight.²⁶  Induction of spinal anesthesia was performed with 15 mg isobaric bupivacaine, and patients received propofol infusion for light sedation. Target-controlled infusions of propofol and remifentanil were used for general anesthesia. At the end of the surgery, the patients under general anesthesia received 0.1 mg/kg IV oxycodone based on their ideal body weight. When used, tourniquets were maintained at 250 mmHg and applied for no more than 2 h. Surgeons injected local infiltration analgesia with 30 mg ketorolac, 300 mg ropivacaine, and 0.5 mg epinephrine to every patient with an organized multipuncture method and 100 mg of ropivacaine to subcutaneous wound edges. Surgeons used a single type of cemented implant with patellar resurfacing and operated through a midline incision and with a medial parapatellar approach.²⁶ 

Postoperatively, acetaminophen and ibuprofen were administered three times daily with premedication doses. A patient-controlled analgesia device was used with no baseline infusion and a preset dose of 0.04 mg/kg IV oxycodone for a maximum of four doses per hour for the first 24 h. Subsequently, the patients received one extended-release oxycodone tablet of 5 to 15 mg. Repeated immediate-release oxycodone (5 to 15 mg orally or 4 to 12 mg intramuscularly) was allowed upon request. All oxycodone doses were predefined according to the ideal body weight.²⁶  From the second postoperative morning, the patients received 50 mg oral tramadol or a combination of 500 mg acetaminophen and 30 mg codeine (one to two tablets up to three times daily). If the immediate-release oxycodone was insufficient, 75 to 300 mg pregabalin orally twice daily was allowed as a rescue analgesic. Peripheral nerve block was allowed if treatment with analgesics proved insufficient.

Patients received prescriptions for acetaminophen, a nonsteroidal anti-inflammatory drug, and either tramadol or a combination of acetaminophen and codeine. Possible contraindications, such as the permanent use of anticoagulants, were noted. Strong opioids or gabapentinoids were prescribed only if the surgeon or anesthesiologist assessed that routine medication was insufficient. Once discharged, patients were allowed to obtain prescriptions from other physicians and healthcare organizations.

The patients completed the Brief Pain Inventory–short form²⁷  and Oxford Knee Score^28,29  questionnaires median 7 days preoperatively and 3 and 12 months postoperatively. The Brief Pain Inventory–short form is a self-administered, validated, and widely used questionnaire in clinical studies for assessing pain. Patients evaluate four pain severity variables (average pain, worst and least pain in the last 24 h, and current pain) and seven pain interference variables with a numerical rating scale (0 = no pain/pain interference, and 10 = worst imaginable pain/pain interference). The Oxford Knee Score is a self-administered and validated questionnaire designed to measure knee pain and function after total knee arthroplasty. From the Oxford Knee Score questionnaire, we extracted responses to the question concerning the description of usual knee pain during the past 4 weeks. The responses were rated on a 5-point scale (none, very mild, mild, moderate, and severe).²⁸ 

We obtained information on opioid and gabapentinoid prescriptions up to 12 months after the operation from the National Prescription Center. This database includes all prescriptions concerning these analgesics. If additional information was necessary to check the indications on prescriptions, we investigated the electronic patient records.

The prespecified main outcome of this secondary analysis was the change in the “average pain” 12 months postoperatively, measured using the Brief Pain Inventory–short form (numerical rating scale).²⁶  Other prespecified secondary outcomes included the change in the average pain 3 months postoperatively and the change in the other three (worst and least pain in the last 24 h and current pain) Brief Pain Inventory–short form pain severity variables, in the arithmetic mean of the four pain severity variables, and in the arithmetic mean of the seven pain interference variables 3 and 12 months postoperatively. The scores of the respective Brief Pain Inventory–short form pain variables at 3 and 12 months were also explored in a post hoc sensitivity analysis.

The prevalence of moderate to severe knee pain at 3 and 12 months postoperation, as defined by the Oxford Knee Score question, was added as a secondary outcome post hoc.²⁸  Given the wide variations in the presented cutoff values for moderate to severe pain, we further included the prevalence of average pain (Brief Pain Inventory–short form) with three different cutoffs (numerical rating scale of 3 or higher, 4 or higher, and 5 or higher) at 3 and 12 months as secondary outcomes post hoc.^4,9,30  In addition, the secondary post hoc outcomes included the number of patients who were prescribed gabapentinoids or opioids (except tramadol and codeine) because of the study operation during the 12-month follow-up.

We conducted sample size calculations of the primary trial with parametric methods to address the prespecified main outcome of this secondary analysis, “average pain” of the Brief Pain Inventory–short form, with the estimated mean numerical rating scale of 5.5 (SD 2.2) and used two-tailed tests with an α level of 0.05 and a power of 80%.²⁶  Subsequently, we increased sample sizes by 16% to adjust for possible nonparametric analyses. We set numerical rating scale 1.0 as the minimal clinically important difference between groups, for which a sample size of at least 90 patients/group was required for nonparametric comparisons.²⁶ 

Categorical data were expressed as frequencies with percentages, normally distributed data as means with standard deviations, and nonnormally distributed data as medians with interquartile ranges. The analysis plan was finalized after the completion of data collection. Comparisons of patient characteristics between the four randomization groups were conducted using the chi-square test or Fisher’s exact test and Bonferroni adjustments in further pairwise comparisons for categorical data and one-way ANOVA for continuous data. The prespecified main outcome of this secondary analysis, the change in average pain 12 months after the operation, and changes in other continuous pain outcome variables at 3 and 12 months after the operation were analyzed using the two-way ANOVA, with the main effects for anesthesia (spinal vs. general) and tourniquet (no tourniquet vs. tourniquet) and an interaction effect between anesthesia and tourniquet. Post hoc sensitivity analyses concerning pain scores at 3 and 12 months after surgery were conducted using the analysis of covariance, with the main effects for anesthesia and tourniquet, an interaction effect between anesthesia and tourniquet, and the preoperative pain score of the respective postoperative score as a covariate. The results from the two-way ANOVA and analysis of covariance were reported as estimated marginal mean differences (95% CI). The comparisons concerning dichotomous secondary outcome data were conducted using binary logistic regression, and the results were presented as odds ratios (95% CI). All statistical tests were two-sided, and P values less than 0.05 were considered statistically significant.

We reported mean pain interference scores if data were available on at least four of the seven items.²⁷  We imputed the scores concerning the mean of four pain severity variables if at least three of the four items were reported. We did not analyze data on patients with randomization deviations because of their low number. Statistical analyses were performed using IBM SPSS Statistics 26 (IBM Corp., USA).

Patient recruitment began in October 2016. A total of 2,783 patients referred for knee arthroplasty were evaluated, 413 patients signed informed consent forms, and 404 were randomized. Preoperative data were eventually analyzed from 395 patients (table 1). The 3- and 12-month follow-ups of this secondary analysis ended in March and December 2019 with data from 391 and 387 patients, respectively (fig. 1). Pain scores of the randomization groups at different time points are presented in table 2. We derived information on prescriptions from 390 patients.

The change in average pain scores (Brief Pain Inventory–short form) 1 yr after the operation did not differ significantly between the anesthesia groups (tables 3 and 4; fig. 2). At 12 months, the spinal anesthesia group reported greater decreases in scores concerning least pain in the last 24 h and arithmetic means of the four pain severity variables and seven pain interference variables, compared with the general anesthesia group (tables 3 and 4). These differences, however, did not reach the predefined threshold (numerical rating scale of 1.0 or more) for clinical importance. The interaction effect between anesthesia and tourniquet was not significant in any pain variable at 3 or 12 months after total knee arthroplasty, indicating that the effect of anesthesia on pain variables was not different in the no-tourniquet and the tourniquet groups (table 4).

In the post hoc sensitivity analysis, the differences between the anesthesia groups in pain scores at 3 and 12 months were consistent with the main analysis, except for the scores concerning the mean of four pain severity variables, which did not differ significantly between the groups (Supplemental Digital Content 1, https://links.lww.com/ALN/C663).

Regardless of the cutoff value (numerical rating scale of 3 or higher, 4 or higher, and 5 or higher), no significant differences in the prevalence of moderate to severe average pain at 3 and 12 months emerged between the groups (Supplemental Digital Content 2, https://links.lww.com/ALN/C664). However, the prevalence of Oxford Knee Score–derived moderate to severe knee pain at 12 months appeared lower in the spinal than in the general anesthesia group (Supplemental Digital Content 2, https://links.lww.com/ALN/C664). The number of patients who received prescriptions for gabapentinoids or oxycodone did not differ significantly between the two groups (Supplemental Digital Content 3, https://links.lww.com/ALN/C665).

The average pain scores decreased less in the no-tourniquet than in the tourniquet group during the 12-month follow-up (tables 4 and 5; fig. 2). In addition, all other pain severity scores and the scores concerning the arithmetic mean of the four pain severity variables in 12 months decreased less in the no-tourniquet than in the tourniquet group (tables 4 and 5). Nevertheless, these differences did not reach the level of minimal clinical importance. The interaction effect between anesthesia and tourniquet was not significant in any pain variable at 3 or 12 months after surgery, indicating that the effect of tourniquet on pain variables was not different in the spinal and the general anesthesia groups (table 4). In the post hoc sensitivity analysis, none of the differences in pain scores between the tourniquet groups at 3 and 12 months were statistically significant (Supplemental Digital Content 1, https://links.lww.com/ALN/C663).

The number of patients with moderate to severe average pain with different cutoff values or knee pain at 3 or 12 months did not differ significantly between the groups (Supplemental Digital Content 4, https://links.lww.com/ALN/C666). Furthermore, the groups did not differ regarding the number of patients receiving prescriptions for oxycodone and gabapentinoids (Supplemental Digital Content 3, https://links.lww.com/ALN/C665).

The total prevalence of moderate to severe knee pain, as defined by the Oxford Knee Score question, was 77% (303 of 395 patients) preoperatively, 24% (93 of 389 patients) 3 months after surgery, and 7% (27 of 387 patients) 12 months after surgery. At the cutoff values (numerical rating scale) of 3 or higher, 4 or higher, and 5 or higher, the total prevalence of moderate to severe pain was 45% (175 of 388 patients), 28% (108 of 388 patients), and 18% (71 of 388 patients), as defined by the Brief Pain Inventory–short form 3 months after surgery, and 37% (144 of 386 patients), 24% (94 of 386 patients), and 15% (58 of 386 patients) 12 months after surgery, respectively.

Only one patient was prescribed buprenorphine, and eight patients (2%) were prescribed oxycodone for postoperative knee pain. No other strong opioids were prescribed. In addition, because of knee pain after the study operation, 40 patients (10%) were prescribed gabapentinoids.

In this secondary analysis of a previously presented randomized trial, the change in average pain scores 1 yr after total knee arthroplasty did not differ between the spinal and general anesthesia groups. The no-tourniquet group reported a smaller decrease in the average pain scores 1 yr after surgery than the tourniquet group; however, this difference was not clinically important. The scores concerning the arithmetic means of the four pain severity variables and seven pain interference variables decreased more in the spinal anesthesia group during the 12-month follow-up compared with the general anesthesia group, although the differences remained below the borderline of clinical importance. In addition to average pain, the no-tourniquet group had smaller decreases in the scores of the other three pain severity variables (worst and least pain in the last 24 h and current pain) and in the arithmetic mean of all four severity variables 12 months after total knee arthroplasty compared with the tourniquet group, but without clinical importance. No interaction effects between anesthesia and tourniquet methods were detected in the pain variables.

In the post hoc sensitivity analysis with the analysis of covariance, the differences in pain scores between the anesthesia groups were mostly in line with the main analysis. However, in the comparison of the no-tourniquet and tourniquet groups, no differences in pain scores at 3 and 12 months were observed. In the other post hoc analyses, the spinal anesthesia group had a lower prevalence of Oxford Knee Score–based moderate to severe knee pain 12 months after total knee arthroplasty compared with the general anesthesia group. However, the prevalence of Brief Pain Inventory–short form–based moderate to severe average pain did not differ between these groups at any time point. Comparing the no-tourniquet and tourniquet groups, the prevalence rates of moderate to severe pain did not differ significantly at 3 or 12 months. The number of patients receiving prescriptions for oxycodone or gabapentinoids for postoperative knee pain did not differ in either comparison.

This secondary analysis of a previously presented trial included both prespecified and post hoc outcomes. Our previously published primary in-hospital results from the same patient cohort revealed that during the hospital stay (median of 2.2 days after surgery), pain management did not differ between the study groups.²⁵  In addition, even though general anesthesia patients reported more pain during the immediate postoperative phase in the recovery room, the differences in pain scores were not clinically important 24 h after surgery.²⁵ 

Previous cohort studies have suggested no differences in the prevalence of persistent postsurgical pain after total knee arthroplasty between regional and general anesthesia groups.^8,13  The main results of this secondary analysis are in line with these studies. Thus, the recommendations to use spinal anesthesia as the primary method in total knee arthroplasty seem to remain unaffected by the outcomes concerning persistent pain.^20–22 

Our results seem consistent with two smaller randomized trials showing no significant differences in persistent postsurgical pain after total knee arthroplasty between tourniquet and no-tourniquet groups.^14,15  Compared with the Brief Pain Inventory–short form–based results of the current trial, some prospective studies have presented lower prevalence rates for moderate to severe or for “significant” persistent postsurgical pain 1 yr after total knee arthroplasty.^4,6,8  Studies of 63 and 116 patients reported prevalence rates of 27% (numerical rating scale of 3 or higher) and 13% (visual analog scale greater than 40 of 100).^4,6  Furthermore, a recent study of 288 patients reported that 16% suffered from moderate to severe persistent postsurgical pain (Western Ontario McMaster Universities Osteoarthritis Index pain score of 30 of 100 or higher).⁸  However, based on the Oxford Knee Score question, the prevalence of persistent postsurgical pain appeared much lower in our analysis. These differences may have arisen from differences in the definitions of persistent postsurgical pain, surgery and anesthesia protocols, sample sizes, and patient characteristics. Questionnaires also play a crucial role, as observed in our results. Oxford Knee Score and Brief Pain Inventory–short form questionnaires involve different time frames. Furthermore, the Oxford Knee Score focuses on the knee, whereas the Brief Pain Inventory–short form is a universal instrument for assessing pain.

Our findings on prescriptions are in line with a recent study reporting a very low proportion of patients receiving postoperative prescriptions for strong opioids.³¹  This result suggests that prescribing strong opioids routinely for pain after total knee arthroplasty may be unnecessary, at least in some populations.

The strengths of this study include its randomized design, adequate sample size, and very low (2%) dropout rate during the follow-up. In addition, this study was conducted in a publicly funded, high volume, tertiary hospital, and numerous arthroplasty surgeons and anesthesiologists treated patients who underwent total knee arthroplasty with modern fast-track protocols.

Among the limitations of this study, its open-label design is the most obvious. Blinding, however, was not feasible. Personnel had either first-hand knowledge or free access to information on operating room events, and patients were undoubtedly aware of the anesthesia method. In addition, we regarded blinding the tourniquet use in the spinal anesthesia group as unreliable. The single-center design and high exclusion rate are also noteworthy limitations. Of 1,583 patients who failed to meet the inclusion criteria, 51% were either over 75 yr, referred for a unicompartmental knee replacement, underwent a bilateral operation, or received a nonprotocol prosthesis, such as a hinge prosthesis.²⁵  In addition, due to randomization deviations, five patients were excluded from the analyses, which might have biased the results.

A further limitation concerns the type of pain. A recent definition for chronic postsurgical pain refers to pain that develops after surgery and persists longer than 3 months.³²  In this study, we used the term “persistent postsurgical pain” to describe the combination of residual pain from the preoperative period and possible chronic pain caused by the surgery. Estimating the proportions of these two pain modalities or their precise nature was not possible with the questionnaires used in this study, and this should be noted in interpreting the results.

In the current analysis, we used numerical rating scale 1.0 as the threshold for minimal clinically important differences between the groups. This was consistent with previous studies assessing the minimal clinically significant changes in pain scores.^33,34  However, for persistent pain after total knee arthroplasty, the optimal threshold for clinically important differences between groups remains to be established. In addition, for data comparability, a consensus should be reached on time points for measuring persistent pain after total knee arthroplasty. We suggest 1 yr as the primary timepoint for future studies, given the current and previous results, which strongly indicate that pain continues to decrease up to 12 months after the operation.^4,6,8  Furthermore, studies with longer follow-up times have not presented lower prevalence rates for persistent postsurgical pain.^3,7,35 

In conclusion, the results from this secondary analysis of a randomized trial concerning total knee arthroplasty suggest that spinal and general anesthesia do not lead to clinically important differences in persistent postsurgical pain. Similarly, operating with or without a tourniquet has no clinically important impact on persistent pain. Prescriptions for strong opioids may be rarely necessary after total knee arthroplasty.

The authors thank Arja Mäkelä, R.N., and Katarina Lahtinen, M.D. (Department of Anesthesiology, Intensive Care and Pain Medicine, Peijas Hospital, University of Helsinki and HUS Helsinki University Hospital, Vantaa, Finland), for their help in data collection. They thank Eero Pesonen, M.D., Ph.D. (Department of Anesthesiology, Intensive Care and Pain Medicine, University of Helsinki and HUS Helsinki University Hospital, Helsinki, Finland), Tatu Mäkinen, M.D., Ph.D. (University of Helsinki and Pihlajalinna, Helsinki, Finland), and Vesa Kontinen, M.D., Ph.D. (Department of Anesthesiology, Intensive Care and Pain Medicine, University of Helsinki and HUS Helsinki University Hospital, Espoo, Finland), for their advice concerning the study design. The authors also thank Pekka Kairaluoma, M.D., Ph.D. (Department of Anesthesiology, Intensive Care and Pain Medicine, Peijas Hospital, University of Helsinki and HUS Helsinki University Hospital, Vantaa, Finland), for helping in patient recruitment; Eliisa Nissilä, M.D. (Department of Anesthesiology, Intensive Care and Pain Medicine, University of Helsinki and HUS Helsinki University Hospital, Helsinki, Finland), for creating the randomization envelopes; and all of the anesthesiologists, surgeons, other participating personnel, and patients who made this study possible.

Supported by HUS Helsinki University Hospital (Helsinki, Finland) grant No. Y102011095 and HUS Helsinki University Hospital Finnish government science grants Nos. TYH 2017239 and TYH 2019113. Personal grants for dissertation research, including the preparation of this article, were granted to Dr. Palanne by the Paulo Foundation (Espoo, Finland), Finnish Medical Foundation (Helsinki, Finland), Finnish Arthroplasty Society (Helsinki, Finland), Research Foundation for Orthopaedics and Traumatology (Finland), and Finnish Society of Anaesthesiologists (Helsinki, Finland). In addition, Dr. Palanne received a travel grant from the University of Helsinki (Helsinki, Finland).

Dr. Madanat has received fees for consultancy and lectures from Pfizer (Helsinki, Finland) and Stryker (Vantaa, Finland) and holds stock options and is a medical advisor for Osgenic Ltd. (Helsinki, Finland). Dr. Reponen has received personal grants from the Foundation for Economic Education, Pulsus Foundation, Finnish Medical Association, and Finnish Society of Anaesthesiologists (all in Helsinki, Finland). Dr. Linko has received a fee from Fisher&Paykel (Helsinki, Finland) for a lecture concerning nasal high flow therapy. The other authors declare no competing interests.

Full protocol available at: riku.palanne@ksshp.fi. Raw data available at: riku.palanne@ksshp.fi.