Article Text

Download PDFPDF

Acute postoperative pain is an independent predictor of chronic postsurgical pain following total knee arthroplasty at 6 months: a prospective cohort study
  1. Asokumar Buvanendran1,
  2. Craig J Della Valle2,
  3. Jeffrey S Kroin1,
  4. Mahendra Shah1,
  5. Mario Moric1,
  6. Kenneth J Tuman1 and
  7. Robert J McCarthy1
  1. 1 Department of Anesthesiology, Rush University Medical Center, Chicago, Illinois, USA
  2. 2 Department of Orthopedic Surgery, Rush University Medical Center, Chicago, Illinois, United States
  1. Correspondence to Robert J McCarthy, Department of Anesthesiology, Rush University Medical Center, Chicago, IL 60612, USA; Robert_j_mccarthy{at}


Background Approximately 15% of patients report persistent knee pain despite surgical success following total knee arthroplasty (TKA). The purpose of this study was to determine the association of acute-postsurgical pain (APSP) with chronic postsurgical pain (CPSP) 6 months after TKA controlling for patient, surgical and psychological confounding factors.

Methods Adult patients with osteoarthritis undergoing primary elective tricompartmental TKA, with the operated knee the primary source of preoperative pain, were studied between March 2011 and February 2017. Patients received standard operative management and a perioperative multimodal analgesia regimen. The primary outcome was CPSP at 6 months. The primary variable of interest was the APSP (weighted mean pain score) for 72 hours postoperatively. Patient, surgical and psychological confounders were assessed using binary logistic regression.

Results 245 cases were analyzed. The incidence of CPSP was 14% (95% CI 10% to 19%). Median APSP values were 4.2 (2.2–5.0) in the CPSP group and 2.8 (1.8–3.7) without CPSP, difference 1.4 (95% CI 0.1 to 1.8, p=0.005). The unadjusted odds for CPSP with an increase of 1 in APSP was 1.46 (95% CI 1.14 to 1.87, p=0.002)). After multivariable risk adjustment, the OR for CPSP for an increase of 1 in the APSP was 1.53 (95% CI 1.12 to 2.09, p=0.008).

Conclusions APSP is a risk factor for CPSP following TKA even after adjusting for confounding variables such as pain catastrophizing, anxiety, depression and functional status. Studies are needed to determine if APSP is a modifiable risk factor for the development of CPSP.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Most patients experience pain relief within 6–12 weeks (3 months) following total knee arthroplasty (TKA); however, 8%–34% of patients experience chronic postsurgical pain (CPSP), defined by the International Association for the Study of Pain as pain lasting more than 3 months after surgery, with limited improvement in functional outcomes often despite radiological and surgical success.1–4 More than 700 000 TKA surgeries were performed in the USA in 2012 and estimates of the expected increase by 2050 range between 143% and 565%.5 With one projected estimate of 3.48 million TKA surgeries per year in the USA by 2030, up to 500 000 patients annually could develop CPSP.6

Acute postsurgical pain (APSP), generally regarded as pain within the first 72 hours postoperatively, has been found to be independent risk factor of CPSP,7–10 although this effect has been inconsistently observed in other studies and may be less important than preoperative pain when assessed using multivariable statistical analysis.11 12 In addition, studies that have assessed APSP as a risk factor for CPSP have not uniformly controlled for confounding factors associated with increased APSP.11 Risk factors that have been shown to be associated with APSP include: preoperative pain, enhanced response to quantitative sensory testing, preoperative use of opioid analgesics, greater pain catastrophizing, depression and increased anxiety.9–20 Patient characteristics such as age, gender, ethnicity and disease severity have also been shown to be associated with CPSP.13

The purpose of this study was to prospectively evaluate the incidence of CPSP following a primary TKA in patients that received a standardized surgical, anesthesia and postoperative analgesia using a multimodal analgesic regimen. The primary variable of interest is APSP, which is defined as the weighted mean pain score for the first 72 hours postoperatively. Evaluation of risk factors for both CPSP and APSP were included in the study to control for confounding factors for both the primary variable of interest and the primary outcome. We hypothesized that APSP would be greater in patients with CPSP and that APSP would be an independent risk factor for the development of CPSP after controlling for risk factors associated with increased APSP.


The protocol was registered at (NCT01320150). This manuscript adheres to the Strengthening the Reporting of Observational Studies in Epidemiology statement guidelines. The study was a prospective, observational cohort conducted at Rush University Medical Center. Eligible patients were English-speaking adults ≤85 years of age undergoing a primary tricompartmental TKA with a diagnosis of osteoarthritis, with the knee to be replaced the primary source of the patient’s pain. Inclusion and exclusion criteria are shown in table 1.

Table 1

Inclusion and exclusion criteria

A convenience sample of eligible subjects were screened and approached 2–4 weeks prior to surgery by study personnel. Screening included an assessment of the patient’s medication history, plans for additional orthopedic surgery and availability for all study-related visits and assessments. All subjects attended an instructional class for preparedness prior to joint replacement surgery. Subjects meeting inclusion criteria were scheduled to meet with study personnel 1 week prior to surgery to undergo baseline testing. Subjects provided informed written consent for study participation prior to baseline testing. Subject characteristics recorded included age, height, weight, race, gender, a history of diabetes mellitus and the current use of an opioid analgesic for pain. The zip code of the subject’s primary residence was obtained so that median income and percentage of individuals below the poverty level within that zip code could be obtained as measures of socioeconomic status.

Baseline testing included physical, pain and psychological assessments as shown in table 2. Physical assessments included active and passive range of motion measured using a goniometer, pain assessment using the Numeric Rating Scale (NRS) of pain intensity, quantitative sensory testing and the Standardized Evaluation of Pain method.21 Psychological status was evaluated using the Pain Catastrophizing Scale (PCS), Becks Depression Inventory (BDI-II) and the State-Trait Anxiety Inventory (STAI). Functional status was assessed using the Western Ontario and McMaster Universities Osteoarthritis (WOMAC 3.1) index and the Optum SF health survey (SF-36v2).22 23

Table 2

Description of physical, pain and psychological assessments.

Patients received oral pregabalin 100 mg and celecoxib 400 mg 1 hour before surgery. Prior to surgery, a lumbar epidural catheter was placed at the L3–L4 or the L4–L5 vertebral interspace and threaded rostrally 3–5 cm into the epidural space. Anesthesia to the lower extremities was achieved with lidocaine 2% containing epinephrine 1:200 000, 10–20 mL titrated to surgical anesthesia. Intraoperative conscious sedation was titrated with propofol. If epidural catheter placement was unsuccessful, the patient had prior back surgery, or the patient refused epidural anesthesia, the patient received general anesthesia. These patients were offered a continuous adductor canal block at the anesthesiologist’s discretion.

The TKA was performed under tourniquet control, using an abbreviated medial parapatellar approach with arthrotomy extending into the quadriceps tendon for 2–4 cm above the superior pole of the patella and without patellar extension. All patients received a cemented tricompartmental TKA with patellar resurfacing. Local anesthetic of ropivacaine 0.5% 30 mL with or without ketorolac was infiltrated at closure of the surgical wound. The knee was closed in 90° flexion over a drain. Intraoperative complications were recorded.

Following surgery, the patient was transferred to the post-anesthesia care unit (PACU) and X-ray confirmation of joint alignment was performed. Patients with epidural catheters in place received patient controlled epidural anesthesia (PCEA) consisting of a continuous epidural infusion of bupivacaine 0.1% with fentanyl 5 µg/mL at a rate of 6 mL/hour with patient-activated bolus doses of 1 mL of the solution every 15 min with a 4-hour lockout of 40 mL. Subjects were instructed prior to surgery on the use of the PCEA. Breakthrough pain was initially treated by adjusting the PCEA parameters using a structured protocol previously described.24 Patients without an epidural catheter received IV morphine via patient-controlled analgesia. Patients with an adductor canal catheter receive ropivacaine 0.2% at 6 mL/hour. All patients received intravenous morphine 2 mg every 4 hours as needed for breakthrough pain. Patients were transitioned to a multimodal analgesia regimen consisting of oxycodone 10 mg every 12 hours, celecoxib 200 mg daily and pregabalin 50 mg every 12 hours. Breakthrough pain was treated with hydrocodone 10 mg plus acetaminophen 325 mg every 4 hour as needed.

Starting on arrival PACU, study personnel recorded NRS pain scores at intervals of 4–6 hours until hospital discharge. Following discharge, patients were assessed in person at 3 and 6 weeks and 3 and 6 months. NRS pain scores at rest and movement were assessed at all post-discharge assessments. At the 6-month follow-up, active and passive range of movements were measured, and the BDI-II, STAI, WOMAC and SF-36 questionnaires were repeated. In addition, the SF-McGill Pain Questionnaire (SF-MPQ-2) was administered to assess the sensory and affective qualities of pain.

CPSP was defined as NRS pain with movement ≥4 at 6 months.3 Recovery from TKA can take up to 3 months; therefore, we choose 6 months as a definition of CPSP for the current study. In addition, in an effort to minimize false positive assignment of CPSP classification based on a single NRS value, pain trajectories between 3 weeks and 6 months were assessed and a consistent pattern of NRS pain at rest >0 and at least one prior recoding of NRS pain with movement ≥4 was required to classify the patient positive for CPSP. The primary independent variable of interest, APSP, was calculated using trapezoidal integration as the weighted average NRS score during the first 72 hours postoperatively.

Statistical analysis

The primary outcome was the rate of CPSP at 6 months following surgery. CIs for the rate of CPSP were determined using the Pearson-Klopper method. The unadjusted association of APSP with CPSP was assessed using binary logistic regression and the Wilcoxon-Mann-Whitney (WMW)odds of a random pair of APSP values from a CPSP subject compared with a no-CPSP subject was calculated using the MWModds function in R. The difference in medians and 95% CI of the difference in APSP in the CPSP group compared with the no-CPSP group was calculated using a 10 000-sample bootstrap.

The univariable (unadjusted) analysis of patient characteristics, socioeconomic status, duration of surgery, primary anesthesia and postoperative analgesia method, functional and psychological testing of subjects with and without CPSP was determined by binary logistic regression. Standardized differences (95% CI) between the CPSP and no-CPSP groups were calculated as Hedge’s g for interval data and Cliff’s delta for ordinal and dichotomous data.

The univariable (unadjusted) association of gender, race, American Society of Anesthesiologists (ASA) physical status, diabetes, preoperative opioid, insurance type neuraxial anesthesia and epidural analgesia use with APSP was assessed by calculating using the Wilcoxon test and the WMWodds of a random pair of APSP values between patients that were positive or negative for the variable. The association of age, body mass index, length of surgery, median household income, percentage below poverty in the subjects zip code, knee circumference, preoperative range of motion, thermal pain sensitivity, baseline NRS pain at rest and with movement, the PCS, BDI-II, STAI scores, the WOMAC subscale scores and the SF-36 Physical Component and Mental Component Summary scores with APSP were assessed using graphical analysis and Spearman’s rho.

Multivariable (adjusted) modeling of the association of APSP with CPSP adjusting for confounding risk factors was performed by including risk factors in a logistic regression model (full model) that had a p<0.05 on univariable analysis for CPSP or for APSP. Missing values in the risk factors assessed were identified in 220 of 6640 (3%) values, with no single variable exceeding 10%. Multiple imputation of missing values was performed using the aregImpute function in the Hmisc package in R. Tolerance >0.1 and a variable inflation factor <10 were considered acceptable to enter the variable into the logistic regression model. A parsimonious model (final model) was created by backward and forward elimination from the full model using a stepwise elimination based on reduction in the Akaike information criteria (AIC). The final model was compared with the full model for goodness of fit using analysis of variance (ANOVA). Goodness of fit for the models were assessed using the Homer and Lemeshow test, Nagelkerke’s pseudo R2 and the area under the receiver operating characteristics curve (AUC). Validation of the final model using backward elimination based on the AIC and calculation of the optimism adjusted AUC was determined using a 1000-sample bootstrap. Measures of effect of variables in the model are reported as adjusted ORs and 95% CIs and as the percentage of the number of iterations in which the variable was retained.

Multivariate analysis of the risk factors in the full model were assessed using conditional tree analysis. The partitioning algorithm attempts to correctly classify members of population by splitting them into subpopulations based on dichotomous split points in the independent variables. All variables are available at each step in the process and groups may be re-split based on subgroups of the same variable. Stopping criteria for the analysis was based on multiplicity adjusted p-values with Bonferroni correction.

Secondary outcomes included the comparison of NRS scores across the post-discharge assessment periods and functional and psychological outcomes between the CPSP and no-CPSP groups at 6 months. An exploratory analysis of NRS pain scores and opioid consumption during the first 72 hours was performed. Weighted average NRS pain score were calculated for the intervals 0–24 hours, 25–48 hours and 49–72 hours. Oral and intravenous opioid analgesics (does not include epidural opioids) administered during the same intervals were converted to oral milligram morphine equivalents.25 Differences in NRS scores, morphine equivalents and functional outcomes between the CPSP and no-CPSP groups were assessed using the Mann-Whitney U test. A p <0.005 required to reject the null hypothesis to adjust for multiple comparisons of the secondary outcomes and exploratory analyses. The differences in medians and 99.5% CIs of the differences were calculate using a 10 000-sample bootstrap. All statistical tests were two tailed.

Based on our prior multicenter survey study, 46 of 259 (17.8%) were identified with CPSP at 1 year following surgery.3 The OR for CPSP was 1.24 (95% CI 1.10 to 1.39) for each one-point increase in reported acute postoperative pain. The average mean postoperative pain in that study was 5.5±2.9. Based on these values, the estimated probability of CPSP at the mean acute postoperative pain value was 15.5% and 30.3% at a mean plus 1 SD (8.5). A power analysis with 90% and alpha equal 0.05 with no other covariates gives us a required sample size of 130. Because we plan to use multiple covariates, we estimate multiple squared correlations at 0.5 increasing our estimate to 259. Assuming a 20% loss in sample due to patient withdrawals and loss to follow-up, we planned to enroll 311 subjects.

Statistical analysis was performed using RStudio V.1.1.453 (Integrated Development for R. RStudio, Boston, Massachusetts, USA; URL: and R V.3.5.1, release date 7 February 2018 (The R Foundation for Statistical Computing, Vienna, Austria). Sample size calculations were made using SAS V.9.2 using the macro program, POWERLOG.SAS (V.1.1).


Enrollment began in March 2011 and ended in August 2016. Patient recruitment and follow-up is shown in figure 1. CPSP was identified in 34 of the 245 (14%, 95% CI 10% to 19%) participants. The median difference in APSP values between the CPSP and no-CPSP groups was 1.42 (95% CI 0.03 to 1.89, p=0.005). The WMWodds for a random pair of APSP values from a CPSP subject compared with a no-CPSP subject was 1.85 (95% CI 1.20 to 2.80, p=0.005). The unadjusted OR for CPSP with an increase of 1 in APSP was 1.46 (95% CI 1.14 to 1.87).

Figure 1

Subject flow through study.

The univariate association of patient characteristics, duration of surgery, primary anesthesia and analgesia type, preoperative NRS scores, APSP and functional and psychological testing of subjects with and without CPSP are shown in table 3. Patients in the CPSP group were more likely to have a greater APSP, greater BDI-II scores, greater event (STAI-S) and trait (STAI-T) related anxiety, greater pain and decreased physical functioning as assessed by the WOMAC index and more impairment in the mental subcomponent of the SF-36 prior to surgery.

Table 3

Univariable (unadjusted) analysis of preoperative patient characteristics, type of insurance, socioeconomic status, study year, duration of surgery, primary anesthesia and analgesia type, preoperative functional and psychological testing of subjects with and without CPSP

Female gender and African American race were associated with an increase in APSP values, whereas ASA physical status II was associated with a decrease in APSP compared with an ASA physical status III (table 4). Active and passive range of motion, baseline NRS scores at rest and with movement, thermal pain sensitivity, the pain catastrophizing score, BDI-II scores, STAI-S, the WOMAC pain subscale score, stiffness subscale and physical function subscales and the SF-36 physical and mental health summary scales were correlated with APSP (table 5).

Table 4

Association of binomial variables with weighted mean pain scores

Table 5

Association of interval variables with weighted mean pain scores

Gender, race, ASA physical status, active and passive range of motion, thermal pain sensitivity values, preoperative NRS values at rest and with movement, the scores from the WOMAC pain, WOMAC stiffness, WOMAC physical functioning, STAI-S, STAI-T, PCS, BDI-II, SF-36 Physical Component Summary and SF-36 Mental Component Summary score and APSP were entered in a binary logistic regression model (full model) for CPSP. The full model had a Nagelkerke’s pseudo R2=0.29, a Homer and Lemeshow p=0.47 and an AUC of 0.81 (95% CI 0.74 to 0.89). The OR for CPSP for an increase of 1 in the APSP in the full model was 1.52 (95% CI 1.09 to 2.12, p=0.01).

Stepwise elimination identified the STAI-S score, WOMAC physical functioning and stiffness subscales, APSP, the SF-36 Physical Component Summary, female gender, ASA physical class II and preoperative passive range of motion as predictors in the parsimonious final model (table 6). The adjusted OR for CPSP with an increase of 1 in APSP in the final model was 1.53 (95% CI 1.12 to 2.09, p=0.008). Performance of the final model was not different than the full model (ANOVA; p=0.83). The final model had a Nagelkerke’s pseudo R2=0.26, a Homer and Lemeshow test p=0.33 and an AUC of 0.80 (95% CI 0.72 to 0.88). Bootstrap validation of the final model resulted in an optimism adjusted AUC of 0.74 (95% CI 0.66 to 0.82). Retention of variables in the validation bootstraps were 82.4%, 79.6%, 77.3%, 63.4%, 57.8%, 51.9%, 48.1% and 47.9% for the variables STAI-S, WOMAC physical functioning score, APSP, SF-36 Physical Component Summary, female gender, ASA physical status II, WOMAC stiffness subscale and passive range of motion, respectively (table 7) . Binary characteristics of the final model were sensitivity 18% (95% CI 7% to 36%), specificity 98% (95% CI 95% to 99%), positive predictive value 60% (95% CI 26% to 88%), negative predictive value 88% (95% CI 83% to 92%) and a diagnostic accuracy of 87% (95% CI 82% to 91%).

Table 6

Parsimonious multivariable (adjusted) binary logistic regression analysis of persistent postoperative pain 6 months following total knee arthroplasty

Table 7

Functional and psychological outcomes and quality of life assessments at 6 months following total knee arthroscopy in patients with and without CPSP

Classification analysis for CPSP identified APSP at the highest level of classification (figure 2). The STAI-S score ≥47 was also identified as a split point for classification of CPSP. Seventy-three per cent of patients with an APSP>4 and a STAI-S score >47 had CPSP at 6 months. Binary characteristics of the classification tree model were sensitivity 23% (95% CI 11% to 41%), specificity 98% (95% CI 96% to 99%), positive predictive value 72% (95% CI 39% to 94%), negative predictive value 89% (95% CI 84% to 93%) and a diagnostic accuracy of 88% (95% CI 83% to 92%).

Figure 2

Conditional classification tree analysis of variables entered into the multivariable binary logistic model for CPSP. The overall prevalence of CPSP was 13.9% (95% CI 10.2% to 19.3%). The data in the terminal branches presented as number with CPSP/number in group, percent incidence in group, 95% CI. CPSP, chronic postsurgical pain; STAI, State-Trait Anxiety Inventory; TKA, total knee arthroplasty.

NRS pain scores at follow-up intervals following discharge from the hospital are shown in figure 3. Pain scores were greater in patients with CPSP at all follow-up intervals. Functional outcomes at 6 months are shown in table 6. The overall WOMAC score and the score for each of the subscales showed less improvement in the CPSP group compared with the no-CPSP group. The Physical Component Summary score of SF-36 questionnaire demonstrated more interference in physical function in the CPSP group. Patients with CPSP described more pain descriptors with greater intensity in the continuous, intermittent and affective descriptor categories of the SF-MPQ-2. Opioid analgesia use beyond 3 months was reported by 8% of the patients and was not different between the CPSP and no-CPSP groups (p=0.40).

Figure 3

Box plots of numeric rating score for pain were 0=‘no pain’ and 10=‘worst pain imageable’ in patients with and without persistent postoperative pain. The horizontal line is the median; the box ceiling and floor are the 25th to 75th percentiles; the whiskers are the 10th and 90th percentiles, and the filled circles are the 5th and 95th percentiles. Upper panel is for pain at rest. Median difference (99.5% CI of the difference) in NRS score for pain at rest between the persistent postoperative pain and no persistent postoperative groups at 3 weeks, 6 weeks, 3 months and 6 months are 1.5 (95% CI 0 to 4, p=0.004), 1.5 (95% CI 0 to 3, p=0.001), 2 (95% CI 1 to 4, p<0.001) and 2 (95% CI 0 to 4 p<0.001), respectively. Lower panel is for pain with movement. Median difference (99.5% CI of the difference) in NRS score for pain with movement between the persistent postoperative pain and no persistent postoperative groups at 3 weeks, 6 weeks, 3 months and 6 months are 3.5 (95% CI 1 to 4.25, p<0.001), 2.5 (95% CI 0 to 4.25, p<0.001), 3 (95% CI 2 to 6, p<0.001) and 6 (95% CI 4 to 7, p<0.001), respectively. P values corrected for 10 comparisons using the Bonferroni method. CPSP, chronic postsurgical pain.

Weighted mean NRS pain scores and oral and intravenous opioid analgesics administered in the intervals 0–24 hour, 25–48 hour and 49–72 hour are shown in figure 4. NRS pain scores were not different between the CPSP and no-CPSP groups for the 0–24 hour and the 25–49 hour epochs. The median (quantile) weighted NRS score was greater 4.3 (3.2–5.5) in the CPSP group compared with the no-CPSP group 2.4 (1.6–3.7) in the 49–72 hour epoch, difference 1.9, 99.5% CI of the difference 0.6–3.3, p<0.001. Morphine consumption was not different between the CPSP and no-CPSP groups for any of the epochs.

Figure 4

Box plots of weighted mean pain scores and oral and/or intravenous opioid administration by epoch (0–24 hours, 25– 48 hours and 49–72 hours) following total knee arthroplasty. The horizontal line is the median; the box ceiling and floor are the 25th to 75th percentiles; the whiskers are the 10th and 90th percentiles and the filled circles are the 5th and 95th percentiles. Upper panel is weighted mean pain scores. Median difference (99.5% CI of the difference) in NRS score for pain between the persistent postoperative pain and no persistent postoperative groups at rest at 0–24 hours, 25–48 hours, 49–72 hours are 1.1 (95% CI −0.6 to 2.4, p=0.05), 0.9 (95% CI −0.6 to 2.0, p=0.07) and 1.9 (5% CI 0.6 to 3.3, p<0.001), respectively. Lower panel is for oral and/or intravenous opioids in milligram oral morphine equivalents. Median difference (99.5% CI of the difference) in milligram morphine equivalents for pain between the persistent postoperative pain and no persistent postoperative groups at rest at 0–24 hours, 25–48 hours, 49–72 hours are −6 (95% CI −18 to 24, p=0.95), −2 (95% CI −29 to 36, p=0.59) and 0 (95% CI −38 to 41 P<0.83), respectively. P values corrected for 10 comparisons using the Bonferroni method. CPSP, chronic postsurgical pain; NRS, Numeric Rating Scale.


The important finding of this prospective study was the association of APSP with CPSP in patients 6 months following TKA even after adjusting for patient characteristics, disease and psychological confounders. We also found that high state anxiety, greater impairment in physical functioning and female gender were also significant independent predictors of CPSP in our cohort. CPSP was associated with significant worse outcomes in all subcategories of the WOMAC index as well as decreased impairment in activities of daily life as measured by the SF-36 questionnaire and an increase in number and severity of pain descriptors in the SF-MPQ-2 at 6 months. Taken together this demonstrates the important clinical significance and burden that accompanies CPSP following TKA.

We observed an incidence of CPSP of 14% and our loss to follow-up rate was 17%. The incidence of CPSP in our study is consistent with other prospective observational trials where rates of unfavorable long-term pain outcomes ranges between 10% and 20% in high-quality studies with low loss to follow-up.4 We examined pain patterns early (3 weeks) following surgery to limit false classification of CPSP based on spurious values at 6 months and to help construct patterns or trajectories that could be used for early identification of CPSP. Our findings are also in agreement with studies suggesting that examination of pain trajectories early (within the first 3–4 weeks) following surgery are useful for identifying patients with persistent pain. 10 ,26 Unlike the findings of Lenguerand et al, we found that WOMAC physical functional status prior to surgery was an independent predictor of long-term pain outcomes using multivariable modeling.16

Prior prospective studies have assessed the influence of APSP on the development of CPSP. Sayers et al examined the association of pain at rest and with movement averaged over the first 3 days postoperatively (VAS 0–10) with chronic pain as assessed using the WOMAC pain subscale at 12 months following total hip replacement and TKA as a secondary analysis of data from the Arthroplasty Pain Experience trial.9 27 Patients were randomized to receive either an intraoperative local anesthetic infiltration or a standard postoperative analgesia regimen consisting of non-steroidal anti-inflammatory drugs and opioids. Like the current study, preoperative pain was positively associated with APSP, but APSP was not associated with chronic pain after adjusting for preoperative pain (WOMAC subscale) using structural equation modeling. The authors suggest that their findings suggest that measures to limit acute postoperative pain will not likely impact CPSP and efforts should focus on reduction in preoperative pain. The model was adjusted for patient characteristics, gender and socioeconomic status, but not for psychological factors that have been shown to influence APSP. Pain was assessed at 08:00, 12:00 and 17:00 daily for 3 days and the mean pain determined as the average of these three samples. In contrast, we assessed pain every 4–6 hours and integrated the area under the curve to weight the APSP calculation.

Grousu et al found that the cumulative maximal pain intensity measured on the first 3 days following TKA was associated with the presence of chronic pain at for up to 1 year on univariable analysis.8 Multivariable analysis to adjust for confounding factors was not performed. In their study, the incidence of pain at rest and with movement at 6 months were 22% and 35%, respectively. The prevalence of opioid use at 6 months and 1 year following surgery was 9%, like that observed in the current study.

Thomazeau et al and Levand’homme et al examined postoperative pain trajectories beginning with the first postoperative day with the risk for the development of chronic pain following TKA. In the aforementioned study, a high pain intensity (mean daily NRS score >5), not regular physical activity in adulthood, a brief pain inventory walking score ≥7 and a high school level of education or greater were identified as independent risk for chronic pain.10 In the later study, patients with higher pain trajectories over the first week following surgery were more likely to report pain at 3 months, although this effect was only significant for patients with chronic neuropathic pain.26 Patient and psychological factors were assessed, but no preoperative factors were found to be significant predictors of chronic pain. They also found that pain of neuropathic origin had greater impairment in quality of life measures than those with chronic pain of non-neuropathic origin.

Our exploratory analysis suggests that difference in APSP reported by patients was significantly different as early as 49–72 hours postoperatively, despite the use of a multimodal pain regimen that included a non-steroidal anti-inflammatory agent, an alpha-2 delta ligand and opioid analgesics. Some separation in APSP was evident in the 25–48 hour time epoch as the patients were transitioned to oral analgesic following discontinuation of the regional anesthesia. Yet, opioid consumption was not different on days 2 and 3 suggesting that difference in pain may represent inadequate analgesic effectiveness in patients that develop CPSP or that the multimodal regimen used in this study does not adequately inhibit pain and inflammatory pathways involved in the transition of APSP to CPSP. With the desire for earlier discharge following TKA, the current study demonstrates the need for more effective analgesic regimens that could be used early in the postoperative period to limit the potential of an increasing the rate of CPSP in patients receiving inadequate analgesia. Conversely, this difference may be unmodifiable, studies suggest that genetic influences on pain are polygenic and may involve single-nucleotide polymorphism mutations in the OPRM1, COMT, ADRB2, and genes related to the GIRK channels KCNJ6 may influence both acute and chronic pain.28

We did observe an association of psychological factors, catastrophizing and anxiety with the both acute and chronic pain, although the effect of catastrophizing appears to primarily influence APSP. Given the significance of the associations between preoperative pain, catastrophizing and high-state anxiety with APSP were observed suggesting that interventions to reduce these factors may have either direct or indirect benefit in reducing CPSP. Systematic reviews of the influence of catastrophizing and anxiety support our findings.29 30

The results of our study should only be interpreted in the context of its limitations. We studied only patients undergoing primary TKA that did not expect to undergo an additional orthopedic procedure within the follow-up period to improve the fidelity of our follow-up assessment. We excluded patients that were chronically taking opioids prior to surgery, patients with a body mass index ≥40 kg/m2, patients with widespread pain or pain likely of neuropathic origin and those with clinically severe depression and significant comorbidities; all these factors have been shown to influence pain and functional outcomes following TKA. We did not assess pain at movement during the first 72 hours and cannot contrast the differences between APSP at rest compared with movement in the early postoperative period. Approximately 15% of our sample received a peripheral nerve block or intravenous patient-controlled analgesia rather than epidural analgesia for the first 24 hours postoperatively, although we saw no difference in the incidence of CPSP based on the primary postsurgical anesthesia type. Also, recruitment and follow-up for the study occurred over a 6-year period (2011–2016), and although we observed no difference in the incidence of CPSP by study years, unmeasured difference in practice may have affected the results of this study.

The intention of our study was to evaluate the association between APSP and CPSP while adjusting for confound factors and not to develop predictive models for CPSP. Nevertheless, predictive models were developed and although the models had high overall diagnostic accuracy, they had low sensitivity and a low positive predictive value and would be most useful for identifying patients at low risk for development for CPSP. Clinically, the model developed using classification tree analysis would not require a substantial amount of additional data to be collected, could be easily be implemented at discharge and may be helpful for stratifying patients into low-risk, intermediate-risk and high-risk categories for CPSP.

In conclusion, we found that APSP is a risk factor for CPSP following TKA even after adjusting for confounding variables such as patient characteristics, pain catastrophizing, anxiety, depression and functional status. Studies are needed to determine if APSP is a modifiable risk factor for the development of CPSP.


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
  21. 21.
  22. 22.
  23. 23.
  24. 24.
  25. 25.
  26. 26.
  27. 27.
  28. 28.
  29. 29.
  30. 30.


  • Funding This study was supported in part by an investigator-initiated unrestricted grant from Pfizer (WS735224), to Rush University Medical Center with Principal investigator as Asokumar Buvanendran of the Department of Anesthesiology. The funding organization had no role in design and conduct of the study; collection, management, analysis and interpretation of the data; preparation, review or approval of the manuscript and decision to submit the manuscript for publication.

  • Competing interests None declared.

  • Ethics approval The study was approved by the Institutional Review Board of Rush University (10081110).

  • Provenance and peer review Not commissioned; externally peer reviewed