Introduction

Knee osteoarthritis is the most common form of arthritis and a leading cause of pain and disability1, particularly in the older population. Conservative treatment options include weight reduction (if appropriate), education, physiotherapy, and analgesics. For patients with severe symptomatic knee osteoarthritis, knee arthroplasty remains the main therapy. Data from the US demonstrated that the volume of arthroplasty substantially increased by 161% from 1991–20102, and is projected to further increase in the coming decade3. Similar increase in burden of knee arthroplasty were projected in other developed countries4. Although knee arthroplasty is a successful procedure for many patients, up to a quarter of patients with knee osteoarthritis experience chronic pain, disability, a poor quality of life and dissatisfaction following knee arthroplasty5. With the increasing burden of knee arthroplasty performed for knee osteoarthritis, understanding the modifiable risk factors for outcomes after knee arthroplasty is essential to inform appropriate care provision.

Pre-operative psychosocial factors have been found to be associated with patient-reported outcomes after knee arthroplasty, with anxiety and pain catastrophizing having the strongest relationships with post-operative outcomes6,7. However, from other studies, psychological factors have been found to have low predictive power for knee arthroplasty outcomes7,8, explaining less than 10% of the variability of chronic pain. Therefore, identifying more influential and modifiable pre-operative psychological factors that may affect the outcomes of knee arthroplasty would be important.

Pain catastrophizing is defined as an exaggerated negative reaction towards pain, both real and imagined9. It describes a response style to painful experiences, characterized by a tendency to focus excessively on pain sensations, magnification of pain sensations and to perceive oneself as being unable to control pain symptoms. The presence of pain catastrophizing impacts a person’s belief system and coping strategies to pain and increases the likelihood of psychological distress. There has been evidence to support the prognostic role of catastrophizing in the development of chronic musculoskeletal pain10. For people with knee osteoarthritis, high pain catastrophizing was associated with pain progression over time to a pattern of persistent and unpredictable pain and associated disability11.

For patients undergoing knee arthroplasty, pain catastrophizing has been found to be associated with prolonged hospital stay12 and persistent pain after knee arthroplasty13,14,15. Data from systematic reviews generally supports pain catastrophizing, together with other psychological factors, as predictor of post arthroplasty outcomes16,17,18,19,20, yet a few studies reported the contrary21,22,23,24. Variation in results might be due to non-prospective study design22, small sample sizes, variability of pain catastrophizing severity across studies25, heterogeneity in knee arthroplasty outcomes being evaluated and failure to control for baseline confounding variables such as pain, function and other psychological factors17,20. Therefore, the current study aimed to evaluate the association of pain catastrophizing with pain and function outcomes at 6 and 12 months after arthroplasty for knee osteoarthritis. We hypothesized that a lower (better) preoperative pain catastrophizing score may result in a greater improvement in pain and function following knee arthroplasty.

Methods

Patients

In this prospective study, consecutive patients with knee osteoarthritis enlisted for primary knee arthroplasty (including total or unicompartmental knee arthroplasty) were offered participation during their pre-operative assessment visit (1–2 weeks prior to knee arthroplasty) from May 2017 to September 2018. We excluded people illiterate to both English and Chinese, those with cognitive impairment via a Mini-Mental State Examination ≤ 24/30, as this may affect the understanding and completion of patient reported outcome measures. We excluded cases of knee arthroplasty for indications other than primary knee osteoarthritis and those listed for revision knee arthroplasty. The study procedures were performed in accordance with the Declaration of Helsinki. The study protocol was read and approved by the SingHealth Central Institutional Review Board (CIRB Ref E2016/3168), and all patients provided informed written consent before participation.

Main outcome variable

The severity of pain and functional limitations in the index knee pre-operatively and at 6 months and 12 months after knee arthroplasty was assessed using the Western Ontario and McMaster University Index (WOMAC)26. The WOMAC is a 24-item disease-specific questionnaire with 3 subscales measuring knee pain (5 items), knee stiffness (2 items) and physical function (17 items) for knee osteoarthritis on a 5-point Likert scale, with higher scores indicating worse symptoms. The WOMAC has been validated for use in knee osteoarthritis in Singapore27. We converted the ordinal WOMAC pain and function scores using a Rasch model derive transformation table into interval scales for the calculation of change scores28. The score range of WOMAC pain (0–20) and function (0–68) in interval scale remained as the original. The conversion table for WOMAC pain and function from ordinal to interval scale is shown in Supplementary Table 1.

Patients were contacted by telephone by a trained staff at 6- and 12-months after surgery to complete the WOMAC. The primary outcome of interest for this study was the change in interval scale of WOMAC pain and function28 at 6-months after knee arthroplasty. An interval scale is a quantitative measurement scale where units have an order with the difference between any unit points along the scale is equal and provides the basic requirement for linear measurement over time29,30. We chose 6-months as the primary outcome as improvement in pain and function after knee arthroplasty often plateau within 3-months, and the 6-month time point is frequently used as clearly representing persistent post-operative pain31. In addition, we explore the change in WOMAC pain and function at 12-months after knee arthroplasty as secondary outcomes. During the 6- and 12-month telephone interviews, we also collected data on whether the arthroplasty was conducted, patient global assessment (0–10) of their knee arthritis and their overall satisfaction to the surgery.

Study factors—the pain catastrophizing scale

Pre-operative pain catastrophizing was assessed using the Pain catastrophizing Scale (PCS)32. The PCS asks patients to reflect on past painful experiences to indicate the degree they experienced 13 thoughts or feelings of pain, on 5-point Likert scales from (0) not at all to (4) all the time. The higher the total score, the more one exhibits pain catastrophizing behavior. Accumulated evidence suggests that the PCS has a second-order, three-factor structure which includes rumination (PCS-R) (score ranges 0–16), magnification (PCS-M) (score ranges 0–12) and helplessness (PCS-H) (score ranges 0–24), which aggregates to yield a total score (PCS-T) (score ranges 0–52). The internal consistency, construct validity and structural validity of PCS as a measure of pain catastrophizing in patients with knee osteoarthritis have been demonstrated in our previous study33.

Other variables

Pre-operative sociodemographic variables collected include age, sex, ethnicity, and education level. Body weight and height were measured at recruitment to calculate body mass index (BMI) in kg/m2. Routine anterior–posterior weight standing and weight bearing knee radiographs taken before knee arthroplasty for the index knee were scored by an experienced musculoskeletal radiologist (SWB) blinded to clinical data, using the Kellgren–Lawrence (KL) grading scheme34. The index knee was designed as the knee undergoing knee arthroplasty or the dominant side in case of bilateral knee arthroplasty. Seventy knee radiographs were read twice 12-week apart by the same assessor. The intraclass correlation coefficients of the Kellgren–Lawrence grade of the tibiofemoral compartment was 0.75 (95% confidence interval, CI: 0.60–0.84).

All patient-reported outcomes were self-administered by patients, in either English or Chinese according to patients’ preference. Questionnaires available in both English and Chinese cover 98% of the multi-ethnic general population locally35. Self-reported pre-operative comorbidities were collected using the Functional Comorbidity index (score range 0–18)36. Duration of pre-operative knee pain was self-reported in the following categories (< 1, 1–2, 3–5, 6–8, 9–12, 13–15 and over 15 years).

Psychological factors including anxiety and depression may be confounding the relationship between pain catastrophizing and knee arthroplasty outcomes6,7,37 and therefore were assessed. Pre-operatively, patients completed the Hospital Anxiety and Depression Scale (HADS)38, which consists of two separately scored 7-item subscales of anxiety (HADS-A, score range 0–21) and depression (HADS-D, score range 0–21), with higher score indicating higher anxiety/depression.

Statistical analysis

Demographic variables are presented using descriptive statistics. Continuous variables are presented as mean (standard deviation (SD)), while categorical variables are reported as frequency and percent. We constructed regression models to evaluate the associations between variables and knee arthroplasty outcomes (change in WOMAC pain and WOMAC function at 6-month and 12-month) using generalized linear regression. In addition to pain catastrophizing, other variables previously shown to be associated with pain and functional outcomes after knee arthroplasty were considered in the models. These variables included age, sex37,39,40, ethnicity (Chinese versus, vs. non-Chinese)41, education level (none vs. primary vs. secondary or above), BMI42, radiographic severity (KL 4 vs. KL 3 vs. KL 0–2)43, type of arthroplasty (total knee arthroplasty vs. unicompartmental arthroplasty)44, anxiety/depression37, Functional Comorbidity Index, and pre-operative WOMAC scores17,18,19. Variables with a P-value of > 0.1 in the univariate models were excluded from entering the multivariable analysis.

We further explored a pain catastrophizing phenotype in predicting knee arthroplasty pain and functional outcomes. As there is no standardized cut-off for the PCS25, we defined “pain catastrophizer” by two methods. First, above the 4th quartile of PCS. Individuals who scored > 75th percentiles on the PCS are generally considered at moderate risk of developing chronic symptoms. Second, we defined “pain catastrophizer” with a PCS cut-off of ≥ 30 which has been considered as candidates for targeted intervention program32. We compared pain catastrophizers to controls in predicting WOMAC pain and functional outcome at 6-month with adjustment with confounding variables.

Further sensitivity analyses conducted in a group excluded patients with bilateral knee arthroplasty to demonstrate consistency of results at 6-months.

P values of < 0.05 were taken as statistically significant in all analysis. All analyses were conducted using IBM SPSS Statistics for Windows, version 25 (IBM Corp., Armonk, N.Y., USA).

Results

Prior to surgery, out of the 1539 patients approached, 1379 were eligible, and 1201 patients were recruited (response rate 87%). There were 1136 (94.6%) patients who received arthroplasty. Of these, 1102 and 1089 patients provided questionnaire data at 6- and 12-month after surgery (Fig. 1). Baseline characteristics of the 1136 patients who received knee arthroplasty are shown in Table 1. The mean (SD) age of participants was 65.9 (7.0) years, 794 (69.9%) were female, 84% were Chinese, mean (SD) BMI was 28.3 (4.7) kg/m2, number of comorbidities was 2.2 (1.1) and 59.2% had received secondary education or above. In terms of surgery, 92% of participants had a total knee arthroplasty, 8% had unicompartmental arthroplasty and 5.4% had bilateral knee arthroplasty. At the end of 6- and 12-months, 12 and additional 49 patients respectively had arthroplasty on the contralateral knees.

Figure 1
figure 1

Flow diagram.

Full size image
Table 1 Baseline characteristics of participants who had primary arthroplasty for knee osteoarthritis.
Full size table

Change in WOMAC pain and function at 6-months

In the univariate analysis of change in WOMAC pain and function, age, sex, BMI, radiographic OA severity, anxiety, depression, baseline WOMAC pain, and PCS were significantly associated with the change in both WOMAC pain at 6 months (Table 2). The R-squared of the univariate model showed the proportion of variance of change in WOMAC pain as explained by each variable. Overall, PCS explained a modest proportion of total variance of change in WOMAC pain (R2 = 0.073), as compared to baseline WOMAC pain (R2 = 0.244). In the multivariable analysis, only sex (P = 0.009), radiographic severity (P = 0.003), baseline WOMAC pain (P < 0.001) and PCS (P < 0.001) remained significant for the change in WOMAC pain at 6 months (Table 2). Every one-unit increase in PCS was associated with a decrease of only 0.04 units in the change of WOMAC pain score at 6 months.

Table 2 Variables associated with change in WOMAC pain* at 6-months.
Full size table

As for change in WOMAC function at 6-months, it was only statistically significantly associated with BMI (P = 0.040), radiographic severity (P < 0.001), baseline WOMAC pain (P < 0.001), and PCS (P = 0.010) (Table 3).

Table 3 Variables associated with change in WOMAC function* at 6-months.
Full size table

Change in WOMAC pain and function at 12-months

Results for regression models for change in WOMAC pain and function at 12-months were generally consistent with those at 6-month. PCS was significantly associated with change in WOMAC pain (P < 0.001), and change in function (P < 0.001) at 12-months after adjustments in the multivariable analysis (Supplementary Tables 2 and 3).

Exploratory analysis

Results of the exploratory analysis for pain catastrophizers defined as (1) 4th quartile of PCS (PCS cut-off > 19); and (2) PCS above cut-off of 30 (n = 113); compared to non-catastrophizers were generally consistent with the primary analysis. Pain catastrophizers were negatively associated with change of WOMAC pain and function (Supplementary Tables 4–7).

Sensitivity analysis

Results in the sensitivity analysis excluded patients who had bilateral knee arthroplasty (n = 1075) were consistent with the total cohort. PCS was significantly associated with change in WOMAC pain and function at 6-months after adjustments of confounders (Supplementary Table 8).

Discussion

We evaluated pre-operative PCS and post operative change in pain and functional in a large Asian cohort. The demographic of this knee arthroplasty cohort was similar to cohorts of patients with knee osteoarthritis in Asia45,46, consisted of patients in their 60 s and predominantly female. We found that pre-operative PCS was predictive of poorer pain and function outcomes after knee arthroplasty after adjusting for potential confounders. This was consistent for shorter term (6-months) and longer term (12-months) after knee arthroplasty.

Previous studies have found that psychological and cognitive factors are associated with poor outcomes after knee arthroplasty17,18,20,47. Although some studies have identified pain catastrophizing as a predictor of pain and function after knee arthroplasty16,17,18,19,20, other studies have found no association21,22,23,24. Previous research on this topic has been limited by factors such as small sample size, non-prospective study design, heterogeneity in knee arthroplasty outcomes and incomplete follow up17,20. Our current prospective cohort study with a large sample size added to the current literature that pain catastrophizing was associated with poorer pain and functional outcomes after knee arthroplasty. However, the magnitudes of associations were relatively small compared to other variables like sex, pre-operative pain and radiographic severity. The association between pain catastrophizing and functional outcomes post-knee arthroplasty were not consistent in the literature. Several smaller prospective studies have shown null association21,22,23,24, while Birch et al. has reported an association of PCS with change in Oxford knee score at 4- and 12-months post-knee arthroplasty in 615 patients48. In the current prospective study using interval scale of WOMAC, we observed an association of PCS with change in WOMAC function at both at 6- and 12-months.

Our study has several unique strengths. First, it is the first prospective study in Asia with a large sample size to examine the association of pain catastrophizing with change in pain and function after knee arthroplasty. The cohort study had high follow-up rates of 97% and 96% at 6- and 12-months. We used a validated instrument, the WOMAC, for the measurement of knee osteoarthritis outcomes27. Separate WOMAC pain and function scores were available to reveal the differential effect of association with PCS. Furthermore, we transformed the WOMAC pain and function scores from ordinal to interval scale28, and all the changes in score used as primary outcomes were in interval scores that follows arithmetic orders to inform the magnitude of change accurately29,30. This ensured any unit change in WOMAC pain and functional scores along these interval scales were equal regardless of the baseline scores. We adjusted for potential confounding variables in the multivariable analysis that may influences post-operative pain and function, including age, sex, ethnicity, BMI, radiographic severity, comorbidities, anxiety/depression, and type of knee arthroplasty. The results were generally consistent in the models with a “pain catastrophizing phenotype” classified by two methods. Given that pain is a common reason why patients undergone knee arthroplasty, this implies that improving pain catastrophizing experience, especially for those with a high pre-operative PCS may improve pain outcomes after knee arthroplasty. In a randomized controlled trial in 402 persons with pain catastrophizing undergoing knee arthroplasty, cognitive behaviourally based pain coping skills training delivered by telephone prior to operation did not find significant benefit for post-operative pain or functional up to 12-month compared to standard care or arthritis education49. Further studies in different modality or design of psychosocial interventions targeting pain catastrophizing before knee arthroplasty may be relevant.

Our study had several limitations. PCS was only collected at baseline and no information is available to inform whether there was a change in PCS over time. It has been shown that PCS changes with improvement of pain49,50. However, we were not able to differentiate whether the trait (baseline PCS) or the state (change of pain catastrophizing experience) has the greater influence on knee arthroplasty pain outcomes. We also acknowledge that we do not have data on pre-operative knee strength, knee alignment, and range of motion that may influence post-operative pain and function. The use of telephone interview post-operatively has limited the board and depth of data we could collect. Finally, we do not have data on revision knee arthroplasty over the 12-month follow-up period. As the number of revision knee arthroplasty in our center have been small as previously reported51, this small number should not have a big impact on WOMAC outcomes of the group.

In conclusion, our study adds to the current knowledge regarding the potential impact of psychological distress on change in pain and function after knee arthroplasty. Pain catastrophizing was associated with less improvement in pain but was not associated with functional improvement at 6-months and 12-months after knee arthroplasty after adjustment with confounding variables. Further studies are required to evaluate the effect of interventions to modify pain catastrophizing in improving knee arthroplasty outcomes.