If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Duloxetine, a serotonin-norepinephrine dual reuptake inhibitor, may improve analgesia after total knee arthroplasty (TKA). Previous studies had one primary outcome, did not consistently use multimodal analgesia, and used patient-controlled analgesia devices, potentially delaying discharge. We investigated whether duloxetine would reduce opioid consumption or pain with ambulation.
A total of 160 patients received 60 mg duloxetine or placebo daily, starting from the day of surgery and continuing 14 days postoperatively. Patients received neuraxial anesthesia, peripheral nerve blocks, acetaminophen, nonsteroidal anti-inflammatory drugs, and oral opioids as needed. The dual primary outcomes were Numeric Rating Scale (NRS) scores with movement on postoperative days 1, 2, and 14, and cumulative opioid consumption surgery through postoperative day 14.
Duloxetine was noninferior to placebo for both primary outcomes and was superior to placebo for opioid consumption. Opioid consumption (mean ± SD) was 288 ± 226 mg OME [94, 385] vs 432 ± 374 [210, 540] (duloxetine vs placebo) P = .0039. Pain scores on POD14 were 4.2 ± 2.0 vs 4.8 ± 2.2 (duloxetine vs placebo) P = .018. Median satisfaction with pain management was 10 (8, 10) and 8 (7, 10) (duloxetine vs placebo) P = .046. Duloxetine reduced interference by pain with walking, normal work, and sleep.
The 29% reduction in opioid use corresponds to 17 fewer pills of oxycodone, 5 mg, and was achieved without increasing pain scores. Considering the ongoing opioid epidemic, duloxetine can be used to reduce opioid usage after knee arthroplasty in selected patients that can be appropriately monitored for potential side effects of the medication.
], it is desirable to find methods to reduce postoperative opioid use.
Duloxetine, a serotonin-norepinephrine dual reuptake inhibitor, is approved in the United States for depression, anxiety, diabetic peripheral neuropathy, fibromyalgia, and chronic musculoskeletal pain. A Cochrane review of duloxetine for chronic pain reported that 60 mg daily was effective for pain, 120 mg was no more effective, and 20 mg was ineffective [
]. There is strong justification for additional duloxetine studies with multiple hypothesis testing design. Moreover, previous duloxetine TKA studies used a variety of less-current analgesic regimens: intravenous patient-controlled analgesia (IV PCA) [
This study aimed to determine whether 60 mg of duloxetine daily provided analgesic benefits lasting up to 14 days for TKA patients receiving a modern analgesic regimen of nerve blockade and oral multimodal analgesia, without PCA. Secondary outcomes included pain scores at rest, pain interference, satisfaction, side effects, compliance with study drug administration, orthopedic outcomes using the Knee Injury and Osteoarthritis Outcome Score for Joint Replacement (KOOS, JR), and blinding success.
This triple-blinded randomized controlled trial was approved by the Institutional Review Board in June 2017 (IRB# 2017-0655) and registered with clinicaltrials.gov (NCT03271151). Written informed consent was obtained from all study participants.
Eligible participants included English-speaking patients, age 25 to 75 years, scheduled to receive regional anesthesia for primary TKA who were planning to be discharged home or to a participating rehabilitation facility. Exclusion criteria included inability to understand the study protocol, opioid consumption for longer than 3 months, planned use of general anesthesia, prior major ipsilateral open knee surgery, allergy or intolerance to study medication, hepatic insufficiency, estimated creatinine clearance <50 mL/min, American Society of Anesthesiologist physical status of IV, and chronic gabapentin or pregabalin use (regular use for more than 3 months). Patients were also ineligible for participation if they were using duloxetine or other SNRIs, SSRIs, MAOIs, tricyclic antidepressants, triptans, lithium, buspirone, or St. John’s Wort.
Prior to study initiation, a statistician not otherwise involved with the study created a randomization table. Group assignments were provided to the investigational pharmacist. After consent was obtained, the pharmacist released the study medication: either 60 mg duloxetine (Cymbalta) or a similar-appearing placebo. Group assignment was concealed from patients, physicians, and research assistants collecting data. The statisticians remained blinded until after completing the analysis of the primary outcomes. Patients were assigned in a 1:1 ratio to take 15 oral doses of duloxetine or a placebo pill, using block randomization. The first dose was administered before transfer to the operating room (OR) and then once daily until the postoperative day (POD) 14.
On the day of surgery, patients received preoperative study medication, 10 mg sustained-release oxycodone, and 15 mg meloxicam. Patients received a standard intraoperative and postoperative multimodal anesthetic protocol: a spinal-epidural (subarachnoid mepivacaine, 45-60mg); adductor canal block (ultrasound-guided; 15cc bupivacaine, 0.25%, with 2mg preservative-free dexamethasone); Interspace between the popliteal artery and capsule of the posterior knee (IPACK) block (ultrasound-guided, 25 cc bupivacaine, 0.25%); periarticular injection by surgical staff (30 cc bupivacaine, 0.25% with epinephrine, methylprednisolone 40 mg/mL, 1 mL, 500 mg cefazolin in 10 mL, 22cc normal saline). Intravenous (IV) medications included 30-50 mg ketamine, 4 mg dexamethasone, 4 mg ondansetron, 20 mg famotidine, and 15 mg ketorolac. The IV sedation included up to 5 mg midazolam and propofol titrated to effect. For postoperative pain management, patients were scheduled to receive the study medication once daily for 14 days; 4 doses of 1000 mg IV acetaminophen every 6 hours followed by 1000 mg oral acetaminophen every 8 hours; 4 doses of 15 mg IV ketorolac followed by 15 mg meloxicam every 24 hours; and 5-10 mg oral oxycodone was given as needed for pain. Patients could have pain medications adjusted as indicated.
Data were collected through review of medical records, patient interviews and by telephone after discharge. Preoperative characteristics: age, sex, race, ethnicity, and body mass index. In the holding area, POD0: Numeric Rating Scale (NRS) for pain (at rest, movement, average, and worst in 24 hours prior to surgery), preoperative medication usage, fibromyalgia symptom severity index, Michigan Body Map, questions based on the Brief Pain Inventory (BPI), Patient-Reported Outcomes Measurement Information System (PROMIS) anxiety and depression, painDETECT, Quality of Recovery-9 (QoR-9), and the Pain Catastrophizing Scale. POD1: NRS pain at rest and with activity, and the Opioid-Related Symptom Distress Scale (ORSDS). POD2: NRS pain at rest and with activity, and QoR-9. Patients were given a journal to document their pain medication usage. POD7: compliance with taking the study medication, and pain medication usage. POD14: pain medication usage, NRS pain scores, BPI pain intensity questions, PROMIS anxiety and depression, ORSDS, satisfaction with pain management, study medication compliance, KOOS, JR, and blinding assessment. At 6 weeks: medication usage in the past 24 hours. POD90: BPI pain intensity questions, painDETECT, KOOS, JR and medication usage in the past 24 hours.
Study data were collected and managed using REDCap (Research Electronic Data Capture) electronic data capture tools (version 10.0.28) hosted at Weill Cornell Clinical and Translational Science Center [
Sample size calculation was driven by the opioid consumption noninferiority test because this outcome required more patients than the NRS noninferiority test in order to achieve adequate statistical power. For NRS at rest, with a noninferiority margin of 1.6 points and standard deviations of 2.7 points for treatment and control groups, a sample size of 49 patients per treatment group would provide 80% power. For cumulative opioid consumption, with a noninferiority margin of 20% (159 mg based on previously published data), a duloxetine standard deviation of 279 mg and a standard deviation control of 434 mg, a sample of 84 patients per group would provide 80% power at a one-sided alpha level of 0.025.
Balance comparisons for patient characteristics were made using standardized differences. If the standardized difference had an absolute value greater than 1.96 × (2/80)1/2 = 0.31, this was evidence of imbalance between the groups for that factor. An intention-to-treat basis was used.
Summary statistics for continuous variables were presented as mean (standard deviation) or median (interquartile range), depending on the distribution of the data. Shapiro-Wilk tests were used to assess the normality of continuous variables. A one-sided two-sample t-test with a noninferiority margin of 20% (86.5 mg) of the placebo group’s mean cumulative opioid consumption (from day of surgery to POD14) was run to assess the noninferiority of the duloxetine group to the placebo group at the alpha level of 0.025. For the superiority test, a one-sided two-sample t-test was performed at the alpha level of 0.017, which used a Bonferroni adjustment to account for multiple outcomes. In the superiority test, a 25% (108.1 mg) reduction in cumulative opioid consumption was considered to be clinically meaningful. The median cumulative opioid consumption value was used for imputation of missing values.
If NRS pain with movement and cumulative opioid consumption were both noninferior in the duloxetine group, and if either was superior in the duloxetine group, then duloxetine would be recommended for future procedures. One-sided two-sample t-tests with a noninferiority margin of 1.6 points were run to determine if NRS pain with movement (POD 1, 2, 14) in the duloxetine group was noninferior to that in the placebo group, with an alpha level of 0.025.
For secondary outcomes, continuous outcomes measured at one time point were analyzed using two-sample t-tests or Wilcoxon rank-sum tests for non-normal variables. χ2 or Fisher’s Exact tests were used to analyze categorical variables. Continuous outcomes measured at more than one time point were analyzed using linear mixed models that included a group by time interaction. Bang’s Blinding Index was used to assess the blinding success in each treatment group [
Statistical procedures were performed using SAS version 9.4 (SAS Institute, USA).
From August 2017 to November 2020, 160 patients enrolled (Fig. 1). The most common cause for not enrolling was patient refusal (n = 310/806, 38%). Patients who declined commonly either did not want to take an additional medication or were concerned about side effects. An additional 33% of screened patients did not meet inclusion criteria and 21% were excluded for logistical reasons, primarily relating to lack of research staff for data collection on some weekends and evenings. Patient characteristics and tourniquet time are presented in Table 1. Characteristics such as anxiety, depression, quality of recovery, and neuropathic pain are included in Table 3 so as to facilitate comparison with postoperative values.
For relationship status; ‘single’ included ‘single’ ‘separated’, ‘divorced’ and ‘widowed’. ‘In a relationship’ included ‘married’, ‘living with partner in committed relationship’ and ‘in a relationship’.
For relationship status; ‘single’ included ‘single’ ‘separated’, ‘divorced’ and ‘widowed’. ‘In a relationship’ included ‘married’, ‘living with partner in committed relationship’ and ‘in a relationship’.
For neuropathic pain, score of ≥ 19 on painDETECT scored as ‘yes’.
Pain interference with sleep (BPI). median (Q1,Q3)
Pain interference with enjoyment of life (BPI) median (Q1,Q3)
Fibromyalgia Symptom Severity Score median (Q1, Q3)
Michigan Body MapNumber of marks median (IQR)
Tourniquet time, minutes median [Q1,Q3]
48.5 [39, 55]
45 [35, 55]
SD, standard difference; BMI, body mass index; ASA, American Society of Anesthesiologists; Q1, Q3, 25th percentile, 75th percentile.
a For relationship status; ‘single’ included ‘single’ ‘separated’, ‘divorced’ and ‘widowed’. ‘In a relationship’ included ‘married’, ‘living with partner in committed relationship’ and ‘in a relationship’.
b For employment, includes full or part-time employment.
c For disability payments, includes pending payments.
d For neuropathic pain, score of ≥ 19 on painDETECT scored as ‘yes’.
Duloxetine was noninferior to placebo for both primary outcomes (pain and opioid consumption) and was superior to placebo for opioid consumption (Table 2, Fig. 2). Opioid consumption (mean (SD)) was 288 (226) mg OME vs 432 (374) (duloxetine vs placebo), P = .0039. The effect size for cumulative opioid consumption was based on data postimputation (for missing/incomplete values).
Table 2Dual Primary Outcomes.
Effect Size: Difference in Means (95% CI)
P Value (Noninferiority)
P Value (Superiority)
Cumulative Opioid Consumption (mg OME, POD 0-14) (Mean ± SD)
] (duloxetine vs placebo), P = .046 (Table 3). Depression was uncommon among study participants and did not differ either at baseline or POD14. Anxiety, length of stay, pain at rest, neuropathic pain, QoR9, and compliance with administration of study drug were not different among groups. On POD 0 (preoperatively), 80% of duloxetine patients and 72% of placebo patients had ‘none to slight anxiety’. The standardized difference for anxiety was 0.296. The standardized difference threshold that was calculated for this study was 0.31 (using the Austin 2009 equation: 1.96∗sqrt(2/80) = 0.31. If the standardized difference is under that threshold, then balance is considered to have been achieved. So according to this analysis, the two groups can be considered balanced concerning preoperative anxiety. Three placebo and two duloxetine patients did not receive Ketorolac IV intraoperatively. This difference was not statistically significant at P = .32.
Table 3Secondary Outcomes.
NRS pain at rest (mean ± SD)
3.4 ± 2.3
2.7 ± 2.4
4.0 ± 2.5
4.4 ± 2.8
2.5 ± 2.0
3.0 ± 2.1
Satisfaction with Pain Management (0-10)Median [Q1, Q3]
Median BPI summary interference scores (on a 0-10 scale) on POD14 were 2.3 (1.1, 4.7) vs 4.1 (2.3, 5.9) (duloxetine vs placebo), P = .0015 (Table 4). Statistical significance was achieved for all 7 categories. An effect size of 2 or greater favoring duloxetine was noted for general activity, mood, normal work and enjoyment of life. On POD90, summary interference scores did not differ, but pain interference with general activity and mood favored duloxetine; the effect size was 2 for general activity.
Table 4Pain Interference and Opioid Side Effects.
Brief Pain Inventory: Median (IQR)
5.1 (3.4, 6.6)
4.7 (2.4, 6.7)
2.3 (1.1, 4.7)
4.1 (2.3, 5.9)
1,4 (0.4, 2.9)
Relations with Other People
Enjoyment of Life
Relations with Other People
Enjoyment of Life
ORSDS Clinically Meaningful Events
POD1; mean (SD)
POD14; mean (SD)
Nausea n ‘yes’ (%) POD1
Nausea n ‘yes’ (%) POD14
Vomiting n ‘yes’ (%) POD1
Vomiting n ‘yes’ (%) POD14
Drowsiness n ‘yes’ (%) POD1
Drowsiness n ‘yes’ (%) POD14
Confusion n ‘yes’ (%) POD1
Confusion n ‘yes’ (%) POD14
Itchiness n ‘yes’ (%) POD1
Itchiness n ‘yes’ (%) POD14
Headache n ‘yes’ (%) POD1
Headache n ‘yes’ (%) POD14
BPI summary: average score for 7 interference questions.
ORSDS: summary statistic represents data from 12 symptoms. Selected data are additionally presented for 6 symptoms.
Opioid-Related Symptom Distress Scale summary statistics were not different on POD 1 and POD 14 (Table 4). Clinically Meaningful Events (CME) were defined as side effects occurring with moderate to very severe strength. CMEs related to vomiting (POD1) were more common for the placebo group: 2 (1%) vs 8 (5%) (duloxetine vs placebo), P = .0498. Headache CMEs (POD1) were more common for duloxetine group: 12 (15%) vs 4 (5%), (duloxetine vs placebo), P = .03. The numbers of patients with headaches occurring ‘frequently’ or ‘almost constantly’ was 2 (duloxetine) vs 1 (placebo). The number of patients who rated headaches as ‘severe’ or ‘very severe’ was 1 (duloxetine) vs 2 (placebo). The number of patients who rated headache ‘bothersomeness’ as ‘quite a bit’ or ‘very much’ was 1 (duloxetine) vs 2 (placebo).
The difference in means between KOOS Jr. scores at 3 months was 6 (CI 1,10; P = .02) (Table 5). Worst pain was lower at 3 months for duloxetine patients; 1.7 (1.5) vs 2.7 (2.1), P = .02.
Table 5Knee Injury and Osteoarthritis Outcome Score for Joint Replacement (KOOS, JR).
Adverse events were reported to the IRB for four duloxetine patients and 10 placebo patients (some patients had multiple events). For duloxetine patients, adverse events consisted of hallucinations, fall during physical therapy, inability to ejaculate, fever with rash at the surgical site and increased pain. For placebo patients, adverse events consisted of hyponatremia, opioid overdose and influenza, oral H. simplex, surgical site blisters, 3rd cranial nerve palsy, kidney stone, fall on POD 90, hallucinations and confusion, diaphoresis, hypertension, diarrhea, decreased urine output and tooth pain, and acne.
] was −0.04 (P = .68), indicating that 4% fewer patients correctly guessed that they received duloxetine than expected by chance. The placebo Bang Blinding Index was 0 (P = .5), indicating the frequency of correct guesses equal to what would be expected by chance. These numbers support adequate blinding.
When using a comprehensive multimodal analgesic regimen for TKA, duloxetine was not worse than placebo for both pain and opioid use and was superior for opioid use. The effect side of cumulative opioid use (from POD 0 to 14) was reduced by −126.5 (−221.5, −31.4) mg or 29% in the duloxetine group postimputation (missing/incomplete values), which corresponds to 17 fewer pills of oxycodone, 5 mg. Among patient-oriented outcomes, patients receiving duloxetine reported higher satisfaction with pain management, as well as less interference by pain with mood, walking, normal work, relations with other people, sleep, and enjoyment of life.
] by using more generalizable pain management and by using a dual-hypothesis testing statistical method (the previous study had pain as a single primary outcome). This protocol did not use postoperative PCA, unlike previous protocols based on epidural PCA [
] demonstrated benefits for duloxetine among TKA patients with central sensitization (in that study, 78% of screened patients did not meet inclusion criteria)—the current study shows the reduction of opioid use among patients not selected for central pain sensitization.
One limitation of this study was the high rates of patient exclusions and refusal to enter the study. It is possible that patients who chose not to enter the study were different from those who did enter the study and this may affect the generalizability of these results. For avoiding drug interactions, patients taking most antidepressant medications were excluded. Few patients had depression or anxiety, both at baseline and POD14, and there were no differences between groups, indicating that duloxetine did not improve patient outcomes by reducing anxiety or depression.
Most patients took the 15 doses of study medication as prescribed, and additionally, some others were partially compliant. Because both groups had similar rates of noncompliance, it is unlikely that discontinuation was caused by drug-related symptoms. It is possible that compliance with self-administration of duloxetine would be lower if duloxetine were prescribed as part of a protocol instead of as part of a research trial, given that patients who entered the trial had by definition agreed to take study medication. Future research could focus on ways to encourage patients to follow analgesic protocols at home. Effect sizes may have been greater given perfect compliance, but this is hard to achieve after discharge from the hospital.
This was a single-center study. The sample size of 160 patients does not rule out rare duloxetine side effects, but possible side effects of increased opioid doses taken in the placebo group should also be considered. The optimal duration of therapy is unclear–study drug was given for 15 days in order to increase the chance of reducing pain at POD14. The 60 mg dose is commonly used for analgesia [
] and higher doses may have more side effects. It is possible that patients who were not enrolled would have different outcomes than those who enrolled. There is not a universally accepted definition of MCID for opioid use, but the 29% reduction exceeded the predetermined cutoff of 25%. The difference in means for the 3-month KOOS, JR result was 6, favoring duloxetine, but the proposed MCID is 7 [
]. MCIDs are not available for most of the other outcomes. Of note, we did not observe a difference in opioid consumption between the study group and the control group during the first 48 hours postoperatively. Duloxetine reduces opioid consumption after TKA, for the patients studied in this protocol. It is not clear whether duloxetine would safely reduce opioid consumption if given to patients not similar to those entered in the study. This protocol excluded patients potentially at risk for duloxetine-related side effects and we would caution against the administration of duloxetine to unselected or unmonitored cohorts at this time. Future pragmatic research is needed to evaluate the safety and efficacy of perioperative duloxetine in less selected and less monitored groups of patients. We hypothesize that duloxetine takes time to come into effect so this may explain the delayed onset of the benefit. In addition, patients were in the hospital for the first 48 hours, and opioid consumption in the hospital may vary compared to opioid consumption at home. Postoperative headaches are common and may be multifactorial in origin. The intraoperative spinal-epidural anesthetic can cause postdural puncture headaches. Unfortunately, we did not track the duration of the headaches. Fortunately, severe, or very severe, headaches were uncommon (severe or very severe headaches occurred in one duloxetine patient and two placebo patients).
The comprehensive multimodal analgesic pathway included neuraxial analgesia, intraoperative ketamine, two peripheral nerve blocks, periarticular injection, NSAIDs and oral opioids. Different results may be observed with different pain therapies. The study goal was to determine whether duloxetine would benefit patients receiving a version of what is regarded by our institution as the current best analgesic practice. This study administered 15 daily doses of duloxetine, starting on the day of surgery. The optimal time to initiate duloxetine therapy and the optimal duration of therapy is not known, and these questions would require further research.
In conclusion, patients receiving duloxetine needed lesser opioids to reach similar pain scores, compared to placebo. Moreover, duloxetine patients reported higher satisfaction with pain management and found that pain was less likely to interfere with mood, walking, work, relations with other people, sleep, and enjoyment of life. Major side effects were not noted. More adverse events were reported among the placebo group than the duloxetine group. Duloxetine can be used to reduce opioid usage after knee arthroplasty in selected patients that can be appropriately monitored for potential side effects of the medication. Additional studies are needed to define the optimal duration of therapy, determine rates of rare side effects, and examine the applicability of postoperative duloxetine for other surgeries.
We would like to thank Kara Fields for developing the statistical analysis plan and the power analysis, Mylinh Duong for pharmacy assistance, Julia Reichel for assistance with the manuscript, and Nicole Brunetti, Avery Schnell and Carrie Freeman for helping with enrollment and data collection.
One or more of the authors of this paper have disclosed potential or pertinent conflicts of interest, which may include receipt of payment, either direct or indirect, institutional support, or association with an entity in the biomedical field which may be perceived to have potential conflict of interest with this work. For full disclosure statements refer to https://doi.org/10.1016/j.arth.2022.02.022.
Conflicts of Interest: JY, DD, EG, HZ, KJE, YL, EAG, DHK, RLK, SCH declare no competing interests. DEP has financial interests in DJO Global (Vista, CA) and P.S.I, LLC (Colorado Spring, CO) unrelated to this study. TPS has financial interest in Exactech Corporation (Gainsville, FL) and is a consultant for Lima Corporate (Arlington, TX), both unrelated to this study. GW has financial interest in Stryker (Mahwah, NJ) and Exactech Corporation (Gainsville, FL) and is a consultant for Ethicon, all unrelated to this study.
Funding: This study was supported by Hospital for Special Surgery Anesthesiology Department Research and Education Fund, New York, NY. Research reported in this publication was supported by the National Center for Advancing Translational Science of the National Institute of Health Under Award Number (UL1TR002384), National Institutes of Health , Bethesda, MD. The use of REDCAP in this study was funded by resources supported by the CTSC grant that was awarded to Weill Cornell Medicine (NIH/NCATS Grant # UL1-TR-002384).