Temporal Association of Pain Catastrophizing and Pain Severity Across the Perioperative Period: A Cross-Lagged Panel Analysis After Total Knee Arthroplasty

Abstract

Objective

Although numerous studies show that preoperative pain catastrophizing is a risk factor for pain after total knee arthroplasty (TKA), little is known about the temporal course of the association between perioperative pain catastrophizing and pain severity. The present study investigated temporal changes and their dynamic associations between pain catastrophizing and pain severity before and after TKA.

Design

A secondary data analysis of a larger observational parent study featuring prospective repeated measurement over 12 months.

Setting

Dual-site academic hospital.

Subjects

A total of 245 individuals who underwent TKA.

Methods

Participants completed pain catastrophizing and pain severity questionnaires at baseline, 6 weeks, and 3, 6, and 12 months after TKA. Cross-lagged panel analysis was conducted with structural equation modeling including age, sex, race, baseline anxiety, and depressive symptoms as covariates.

Results

Reduction in pain catastrophizing from baseline to 6 weeks after TKA was associated with lower pain severity at 3 months after TKA (standardized β = 0.14; SE = 0.07, P = 0.046), while reduction in pain severity at 6 weeks after TKA was not associated with pain catastrophizing at 3 months after TKA (P = 0.905). In the chronic postsurgical period (>3 months), pain catastrophizing at 6 months after TKA predicted pain severity at 12 months after TKA (β = 0.23, P = 0.009) with controlling for auto-correlation and covariates, but not vice versa.

Conclusions

We provide evidence that changes in pain catastrophizing from baseline to 6 weeks after TKA are associated with subsequent pain severity. Future studies are warranted to determine whether targeting pain catastrophizing during the perioperative period may improve clinical outcomes for individuals undergoing TKA.

Introduction

Osteoarthritis is a leading cause of chronic pain and disability worldwide. Treatment of advanced osteoarthritis with total knee arthroplasty (TKA) is common and typically leads to significant reductions in pain and improvements in physical function for the majority of individuals. However, despite good outcomes, 15–30% of individuals who undergo TKA report persistent postoperative pain and disability [1–9]. Identifying modifiable nonsurgical risk factors during the perioperative period may improve post-TKA clinical outcomes.

Pain catastrophizing, a maladaptive cognitive and affective response to pain, is a modifiable psychological factor that influences TKA outcomes [1, 10–12]. Robust evidence suggests that preoperative pain catastrophizing contributes to post-TKA pain severity and chronicity, above and beyond the effects of depressive and anxiety symptoms [13–15]. In fact, numerous systematic reviews have suggested that preoperative pain catastrophizing adversely affects clinical outcomes after TKA [16–18], and a meta-analysis has provided further evidence that preoperative pain catastrophizing is one of the strongest predictors of chronic post-TKA pain [1].

Changes in pain catastrophizing during the acute postoperative (i.e., <3 months) period may serve as a novel therapeutic target for improving TKA outcomes, yet scant evidence exists on how changes in pain catastrophizing across the perioperative (i.e., preoperative to 3 months postoperative) period may influence long-term post-TKA outcomes [19]. To our knowledge, only one study has examined temporal associations between pain severity and pain catastrophizing after TKA [20]. That study showed that while controlling for depressive and anxiety symptoms, pain severity and pain catastrophizing measured at baseline and 2 months after TKA predicted future pain catastrophizing at 2 and 6 months, respectively. However, the authors did not report whether changes in pain catastrophizing prospectively predicted post-TKA pain severity [20]. Longitudinal studies assessing the potential bidirectional temporal association between changes in pain catastrophizing and pain severity may provide greater understanding of their causality, which may guide the development of more efficacious pain interventions administered in the preoperative and postoperative periods [21].

In the present study, we conducted a cross-lagged panel analysis, which can reveal the direction of potential causality between variables measured across time [22], in a cohort of patients who underwent TKA. The primary aim of the study was to examine whether temporal changes in pain catastrophizing precede changes in pain severity during the perioperative period. Previous studies using cross-lagged panel analysis revealed that changes in pain catastrophizing precede and contribute to subsequent alterations in pain severity across an acute time period (i.e., days to weeks) in both healthy and clinical pain populations [23–25]. Thus, we expected that there would be a unidirectional association between perioperative pain catastrophizing and pain severity. Specifically, we hypothesized that even when controlling for age, sex, anxiety, and depressive symptoms, baseline to 6-week post-TKA changes in pain catastrophizing would predict pain severity at 3 months after TKA. We also hypothesized that pain catastrophizing at 3 and 6 months after TKA would predict subsequent pain severity in the chronic postsurgical period at 6 and 12 months after TKA. However, we did not expect that temporal changes in pain severity would predict future pain catastrophizing.

Methods

The present study is a secondary data analysis of a larger dual-site parent study featuring prospective repeated measurements over a 12-month period. The parent study collected preoperative measures (e.g., self-report questionnaires, experimental pain testing, daily diaries, physical functioning test) at baseline (i.e., 1–4 weeks before TKA) and postoperative measures at 6 weeks, 3 months, 6 months, and 12 months after TKA. The main outcome report for the parent study, which examines bio-behavioral risk factors related to the development of persistent pain after TKA, is currently in progress. Previously published studies based on the parent study data do not overlap with the present study in terms of research aims and do not use all the longitudinal assessment time points [26–29].

Participants

Participants were recruited through the Johns Hopkins University School of Medicine (JHU SOM), Baltimore, Maryland, and Brigham and Women’s Hospital (BWH), Boston, Massachusetts, between March 1, 2011, and January 15, 2016, through posted flyers, advertisement letters mailed to patients scheduled for TKA, advertisements in local orthopedic clinics, and announcements on hospital and university clinical research websites. Patients were also directly recruited from orthopedic surgery clinics at both institutions. All study-related procedures were approved by the JHU SOM and BWH Institutional Review Boards, and all participants provided informed consent.

Inclusion criteria included 1) age >45 years; 2) met American College of Rheumatology diagnostic criteria for knee osteoarthritis; 3) scheduled or planning to undergo TKA; 4) proficient in English; and 5) on a stable dose of acetaminophen or nonsteroidal anti-inflammatory drugs for at least 1 month before study enrollment. Exclusion criteria included 1) use of opioids within the prior 30 days; 2) recent history of substance abuse or dependence; 3) meeting diagnostic criteria for certain sleep disorders, such as restless legs syndrome; 4) presence of systemic inflammatory or autoimmune disorders; 5) pregnancy; 6) history of Raynaud’s disease; 7) current infection; 8) moderate-to-severe peripheral neuropathy; 9) history of myocardial infarction or other serious cardiovascular condition in the prior 12 months; 10) current use of oral steroids; and 11) demonstration of delirium, dementia, psychosis, or other cognitive impairments that would prevent the completion of study procedures.

We screened a total of 636 patients. Two hundred forty-eight individuals met the inclusion and exclusion criteria and consented to participate in the study. Three participants dropped out before initiating the study. Thus, a total of 245 participants were included in the present study. Because of an administrative oversight at the beginning of the study, baseline data were missing for 33 participants (13.5% of the total sample). However, as this missing data is most likely missing at random (i.e., no relationship between the missingness and the observed data), we used full-information maximum likelihood (FIML) estimation to address the missing data [30, 31] and included these 33 participants in the main statistical analyses.

Study Measures

Pain Severity

The Brief Pain Inventory (BPI) [32] is a well-validated and widely used self-report measure that asks participants to indicate their level of pain intensity on a standard numeric rating scale, from 0 = no pain to 10 = worst pain you can imagine. Pain severity was the mean of four pain items (worst, least, average, and current pain), with higher scores indicating greater pain. Participants completed the BPI before undergoing any experimental pain assessments at each of the visits.

Catastrophizing

The Pain Catastrophizing Scale (PCS) [13] consists of 13 items rated on a five-point Likert scale ranging from 0 (not at all) to 4 (all the time). The measure assesses three dimensions of pain catastrophizing: rumination, magnification, and helplessness. Participants also completed the PCS before undergoing any experimental pain assessments at each of the visits.

Covariates

A number of covariates pertinent to the experience of pain severity and pain catastrophizing were included in the present study: 1) sex, 2) dichotomized race (White vs. racial minorities), and 3) baseline anxiety, measured by the widely used and validated PROMIS Anxiety and Depression Short Form V1.0 [33].

Data Analytic Plan

Before testing our primary hypothesis, we first conducted a repeated-measures analysis of variance (RM-ANOVA) to examine changes in pain severity and pain catastrophizing from baseline to 12-month post-TKA follow-up. We also created pre- and postoperative change scores of pain catastrophizing and pain severity by subtracting baseline scores from 6-week post-TKA follow-up scores. Negative change scores are indicative of improvement (i.e., decrease in either pain severity or pain catastrophizing from baseline to 6 weeks after TKA).

Next, we conducted a cross-lagged panel model using baseline and 6-week post-TKA change scores of pain catastrophizing and pain severity and 3-, 6-, and 12-month post-TKA scores of pain catastrophizing and pain severity. As findings of raw change scores can be influenced by baseline scores [34], we controlled for baseline preoperative pain severity and pain catastrophizing variables in the cross-lagged model (see Figure 1). We also included age, sex, race, and baseline anxiety and depressive symptoms as additional covariates in the model. Appropriateness of model fit was determined by the comparative fit index (CFI), the root mean square error of approximation (RMSEA), and the standardized root mean residual (SRMR). CFI values >0.90, RMSEA values <0.06, and SRMR values <0.08 indicate good fit to the data [35–37]. Missing data were handled by FIML estimation based on a missing-at-random assumption. FIML estimation is one of the best existing missing-data analytic techniques [30, 31]. Variables that were assessed at concurrent time points were allowed to correlate. Mplus 8.2 (Muthen & Muthen) [38] was used to conduct the cross-lagged panel model.

Figure 1.

Cross-lagged model with pain severity and pain catastrophizing. Bold lines indicate statistically significant paths. For visual parsimony, only significant path estimates are shown. All path estimates are standardized regression coefficients. The following covariates were included in the analyses but are not shown for parsimony: age, sex, race, anxiety, and depressive symptoms at baseline. Single-headed arrows indicate regression paths. Double-headed arrows indicate correlations. * P < 0.05, ** P < 0.01, *** P < 0.001.

Results

Participants’ Characteristics

As shown in Table 1, the participants’ mean age was 65.1 years (standard deviation = 8.2), and the majority of participants were female and Caucasian and had at least some college education. Most participants were either married or had a partner and were working full-time or retired.

Table 1.

Baseline sample characteristics

Variables	N = 245
Age, y, mean ± SD	65.1 ± 8.2
Sex, female, % (n)	59.5% (147)
Race, % (n)
White	88.0% (216)
Black	9.1% (22)
Others	2.9% (7)
Education, % (n)
Some high school or less	3.3% (8)
High school graduate/tech graduate/GED	8.7% (21)
Some college	25.6% (63)
College graduate	29.3% (72)
Graduate school degree	33.1% (81)
Marital status, % (n)
Married or living with partner	73.7% (180)
Single	8.2% (21)
Separated or divorced	11.1% (27)
Widowed	7.0% (17)
Employment, % (n)
Full-time	35.1% (86)
Part-time	12.0% (29)
Homemaker	4.5% (11)
Retired	39.3% (96)
Unemployed	3.7% (9)
Disability	5.4% (13)

SD = standard deviation.

Pain Severity and Pain Catastrophizing from Baseline to 12 Months After TKA

Table 2 indicates pain severity and pain catastrophizing at each time point. As expected, RM-ANOVAs indicated that pain severity decreased significantly from baseline to 12 months after TKA (F[2.815, 315.266] = 61.126, P < 0.001). Similarly, pain catastrophizing also decreased significantly over time (F[2.674, 299.498] = 26.971, P < 0.001).

Table 2.

Means and standard deviations of study variables at each time point

Variables	Baseline	6 Weeks After TKA	3 Months After TKA	6 Months After TKA	12 Months After TKA
Pain catastrophizing	13.80 ± 12.10	9.43 ± 10.39	7.87 ± 10.22	6.39 ± 9.20	5.99 ± 9.35
Pain severity	3.76 ± 2.20	2.75 ± 1.78	1.73 ± 1.68	1.61 ± 1.73	1.61 ± 1.80

Findings of the Cross-Lagged Panel Model

The cross-lagged panel model (see Figure 1) had overall good model fit (CFI = 0.987, RMSEA = 0.059, SRMR = 0.016). Changes in baseline to 6-week post-TKA pain catastrophizing significantly (standardized β = 0.15; SE = 0.07, P = 0.046) predicted pain severity at 3 months after TKA while controlling for auto-correlation and covariates. In other words, a decrease in pain catastrophizing from baseline to 6 weeks after TKA was associated with lower pain severity at 3 months after TKA. On the other hand, baseline to 6-week post-TKA changes in pain severity did not predict pain catastrophizing at 3 months after TKA (standardized β = –0.004; SE = 0.07, P = 0.952) while controlling for auto-correlation and covariates.

There were no significant cross-lagged paths from 3 months after TKA to 6 months after TKA while controlling for auto-correlation and all other covariates. Although baseline pain catastrophizing did not predict pain severity at 6 months, baseline anxiety symptoms significantly predicted pain catastrophizing at 6 months after TKA (standardized β = 0.24; SE = 0.08, P = 0.002), and baseline depressive symptoms significantly predicted pain severity at 6 months after TKA (standardized β = 0.19; SE = 0.08, P = 0.019).

With regard to 12-month post-TKA outcomes, pain catastrophizing measured at 6 months after TKA significantly predicted pain severity at 12 months after TKA (standardized β = 0.24; SE = 0.09, P = 0.006) while controlling for auto-correlation and covariates. On the other hand, pain severity measured at 6 months after TKA did not significantly predict pain catastrophizing at 12 months after TKA (standardized β = 0.13; SE = 0.08, P = 0.106) while controlling for auto-correlation and covariates.

Discussion

This study investigated whether changes in pain catastrophizing prospectively influence subsequent pain severity, or vice versa, in individuals who underwent TKA. Using a cross-lagged panel model, we found that reduction in pain catastrophizing from baseline to 6 weeks after TKA was associated with decreased pain severity at 3 months after TKA. These findings were observed even when controlling for baseline to 6-week changes in pain severity, as well as baseline pain severity, pain catastrophizing, anxiety and depressive symptoms, age, and sex. In contrast, changes in pain severity from baseline to 6 weeks after TKA did not predict pain catastrophizing 3 months after TKA. In addition, when we examined the long-term bidirectional temporal association of pain severity and pain catastrophizing beyond 3 months after TKA, pain catastrophizing predicted subsequent pain severity at 12 months after TKA, whereas pain severity did not predict subsequent pain catastrophizing. Collectively, these results provide further support that perioperative changes in pain catastrophizing (i.e., elevations or minimal reductions in high catastrophizers) may be an important clinical target for reducing chronic postsurgical pain (CPSP) after TKA.

Our findings are consistent with numerous studies demonstrating that the greatest reductions in pain severity and pain catastrophizing occur within 3 months after TKA [20, 39–44], and these reductions are sustained over long-term follow-up [20, 39, 45]. Our results also add to a growing body of literature suggesting that changes in catastrophizing may precede changes in pain severity [23–25]. To date, only one study has examined temporal associations between pain severity and catastrophizing in a population of patients undergoing TKA [20]. Contrary to our findings, that study found that baseline and 2-month post-TKA pain severity predicted subsequent pain catastrophizing through 6-month follow-up, though the association of pain catastrophizing with subsequent pain was not reported. Further research is needed to elucidate the mechanisms underpinning these mixed findings. Participants’ education level, a proxy for socioeconomic status, is a notable difference between these studies. Only 10% of participants in our study had a high school education or less, compared with one third of participants in the previous study [20]. Lower socioeconomic status is associated with greater preoperative pain and catastrophizing in individuals undergoing TKA [46], yet the extent to which socioeconomic status impacts acute and long-term outcomes after TKA has been difficult to discern [47, 48]. Future studies that examine the impact of socioeconomic risk factors on catastrophizing, pain expectations, treatment access, and chronic stressors [49, 50] may help elucidate underlying mechanisms influencing the temporal relationship of catastrophizing and pain.

To our knowledge, this is the first study to demonstrate that reduction in pain catastrophizing in the acute (i.e., 6-week) post-TKA period predicts subsequent pain reduction at the acute-to-chronic postoperative transition (i.e., 3-month post-TKA) period. This finding is important because individuals who experience acute post-TKA pain are at greater risk of developing CPSP [51–54], and progression to CPSP predicts worse long-term outcomes [40, 55, 56]. A recent systematic review that evaluated the effectiveness of psychological interventions for managing post-TKA pain highlighted the need for more rigorous studies that assess the effect of timing of intervention delivery on clinical outcomes, including during the acute-to-chronic pain transition period [57]. Few studies [39, 58] have examined the effectiveness of delivering postoperative adjunctive psychological interventions. Our findings that acute post-TKA changes in pain catastrophizing influence pain severity around the time of onset of CPSP buttress support for more empirical research on when to deliver adjunctive interventions.

Mounting evidence suggests that myriad interventions, including surgery itself, can reduce pain catastrophizing in TKA candidates [45, 59]. Nevertheless, results of psychological interventions targeting individuals with elevated pre-TKA pain catastrophizing remain equivocal. In one study, patients with elevated preoperative pain catastrophizing who received an eight-session pain coping skills intervention within 8 weeks after TKA showed clinically significant reductions in pain and catastrophizing when compared to a historical usual care group [60]. However, these results were not replicated in a subsequent larger multisite clinical trial [39] in which patients received an eight-session pain coping skills intervention that began 2 weeks preoperatively and continued through the acute post-TKA period. The authors suggested that a moderate PCS score at baseline may not be a viable predictive tool for TKA outcomes. Similarly, a randomized controlled trial that investigated the effectiveness of a 10-session postoperative motivational interviewing intervention delivered up to 6 months after TKA for individuals with elevated preoperative catastrophizing did not yield functional improvement compared with usual care [58].

Our findings support the possibility that preoperative pain catastrophizing alone may not provide sufficient utility in predicting which individuals would benefit from adjunctive interventions. It is worth speculating that elevated or unchanged pain catastrophizing at 6–8 weeks after TKA may potentially serve as a more nuanced target for adjunctive interventions to help reduce the development of CPSP [61]. Indeed, the recent adaptation of perioperative pain clinics in North America and Europe [62–66] has demonstrated the necessity and effectiveness of identifying and treating patients who experience intense pain and elevated distress both before and after surgery. As many surgical patients enter perioperative pain programs postoperatively [63, 67], empirical research is needed to investigate whether psychological interventions targeting elevated or unchanged acute post-TKA pain catastrophizing may facilitate adaptation of more tailored real-world treatments.

Although pain catastrophizing predicts poor outcomes independent of the effects of anxiety and depressive symptoms, preoperative anxiety and depressive symptoms are also significantly associated with pain severity 1 to 2 years after TKA [68–70], and psychological distress has adverse effects on post-TKA outcomes in older patients [71]. Consistent with the extant literature, we found that baseline anxiety symptoms predicted pain catastrophizing at 3 months and 6 months after TKA, and baseline depressive symptoms predicted pain severity at 6 months after TKA. Collectively, these findings provide further evidence that early-phase trials are needed to adapt perioperative psychological interventions targeting the modifiable factors that are associated with worse post-TKA outcomes.

These findings must be interpreted in the context of study limitations. First, the sample was largely Caucasian, college educated, employed, and married; thus, our findings may not be generalizable across all individuals undergoing TKA. Additionally, as we were unable to ascertain the total number of TKAs conducted at both study sites during the enrollment period, our findings may not reflect the overall potential sample of patients undergoing TKA within our academic centers. Second, the sample excluded patients with opioid use within 3 months of TKA and represented a cohort with low levels of pain catastrophizing; thus, our findings may not be generalizable across a significant minority of patients on preoperative opioids or with high baseline catastrophizing (i.e., PCS ≥16 [39, 72]). Third, a study administration error resulted in some missing data at baseline assessment. However, we used a modern missing-data analysis technique (i.e., FIML) to adequately handle the missing data. Fourth, it was beyond the scope of this study to examine participants’ use of pharmacological and nonpharmacological analgesia treatments after TKA. It is possible that various treatments could have influenced the temporal relationship of pain severity and pain catastrophizing. Despite these limitations, we think our study contributes important clinical results that warrant further investigation.

In conclusion, we found that changes in pain catastrophizing from baseline to 6 weeks after TKA significantly predict pain severity at the transition from acute to chronic postoperative pain and that pain catastrophizing predicted subsequent pain severity in the chronic postoperative period up to 12 months. Conversely, changes in pain severity did not predict subsequent pain catastrophizing, and pain severity did not predict future pain catastrophizing in the chronic postoperative period up to 12 months. Future studies are warranted to determine how changes in pain catastrophizing during the perioperative period may serve as prognostic indicators for TKA outcomes and how postoperative pain catastrophizing may guide the use of tailored adjunct interventions to improve clinical outcomes for individuals undergoing TKA.

 

Funding sources: This research was funded by R01 AG034982 (RRE), F32DA049393 (CJM), T32 NS070201 (TJS), and the Johns Hopkins Bayview Medical Campus Clinical Research Unit.

Conflicts of interest: The authors have no conflicts of interest to disclose.