Sex moderates the association between quantitative sensory testing and acute and chronic pain after total knee/hip arthroplasty

2.1 Participants and procedure

Recruitment took place at the Orthopedics Unit of Hospital of Braga (Portugal), from June 2021 to December 2022. Inclusion criteria were as follows: (a) age >50; (b) unilateral total joint arthroplasty due to knee/hip osteoarthritis; and (c) ability to understand written information and give informed consent. Exclusion criteria were as follows: (a) revision surgery; (b) severe or disabling diseases; and (c) contralateral knee/hip arthroplasty in the previous 6 months. In case of acceptance, the participants signed the informed consent, followed by the baseline assessment. The study was approved by the Ethics Committee from the Hospital of Braga (Ref. 03_2021) and by the Ethics Committee for Research in Life and Health Sciences from the University of Minho (CEICVS 093/2020).

Patients were assessed before surgery, 48 h, and 3 months postsurgery. Presurgical (baseline) assessment focused on sociodemographic, clinical, pain-related, psychological, and QST assessment. The 48 h assessment took place during hospital stay and covered pain frequency, pain intensity, and anxiety. At 3 months, the evaluation focused on pain frequency, intensity, and interference.

This study includes 63 participants with complete baseline data. Assessments at 48 h and 3 months were completed in 62 and 60 participants, respectively. At 3 months, 2 participants were excluded due to revision surgery and 1 due to manipulation under general anesthesia.

2.2 Surgical and anesthetic procedures

Most total knee arthroplasties (TKA) were performed through the medial parapatellar approach (27, 79.4%) and most total hip arthroplasties (THA) were performed via the posterior approach (22, 75.9%).

Anesthetic and analgesic procedures were determined individually by the anesthesiologist in charge. Anesthesia was either locoregional only (subarachnoid block [58, 92.1%] or epidural block [1, 1.6%]), or locoregional plus general anesthesia (4, 6.3%). Spinal anesthesia (locoregional) was achieved with a combination of bupivacaine (0.5%/10 ml) or hyperbaric bupivacaine (20 mg/4 ml) and sufentanil (0.005 mg/ml). Peripheral nerve block was achieved using perineural ropivacaine (7.5 mg/ml) or lidocaine (10 mg/ml).

Postsurgical analgesia was administered by epidural or intravenous routes. Coadjuvant medication (detailed below) was delivered via oral or intravenous route, with indication to administer rescue analgesia if pain intensity was 3 or higher on a 0–10 numerical rating scale (NRS). Epidural analgesia was administered by continuous disposable infusion balloon (DIB [19, 30.2%]) or by patient-controlled epidural analgesia (11, 17.4%) with background infusion. These protocols included a combination of ropivacaine (0.15%) and fentanyl (1.5 μg/ml) and other analgesics such as paracetamol (1 g, 6/6 h) and non-steroidal anti-inflammatory drugs (NSAIDs: ketorolac 30 mg, parecoxib 40 mg, or diclofenac 50 mg, 12/12 h). Rescue analgesia was prescribed individually and administered as needed (ropivacaine 40 mg/20 ml, tramadol 100 mg, methimazole 2 g, pethidine 25 mg). Intravenous analgesia (26, 41.3%) consisted of different combinations of paracetamol (1 g, 6/6 h), ketorolac (30 mg, 12/12 h), and tramadol (100 mg, 8/8 h), and additional medication (pethidine, methamizol) delivered in case of uncontrolled pain (NRS >3). Some patients had an intravenous DIB protocol (7, 11.1%), with a 5 ml/hour drip of tramadol (400 mg) and droperidol (5 mg), along with coadjuvant medication (paracetamol and NSAIDs) and rescue analgesia (pethidine, 20 mg).

Antiemetic treatment (metoclopramide or ondansetron) was used as needed and all patients underwent infection prophylaxis with antibiotics (cefazolin) and thromboembolism prophylaxis with LMWH (low molecular weight heparin – enoxaparin).

2.3 Measures

2.4 Statistical analyses

All analyses were conducted in IBM SPSS Statistics v.29, with statistical significance set at p < 0.05. Data are reported as absolute and relative frequencies (n, %), or mean and standard deviation (mean ± SD). A logarithmic transformation (base 10) was performed to MDT, WUR, and PPT tests to achieve a secondary normal distribution [29]. Data distribution was analyzed through skewness (Sk ± 2) and kurtosis (Ku ± 7), with no substantial differences from a normal distribution [31].

Between-sex differences were explored through independent samples’ t-tests or chi-square (χ ²) tests. Cohen’s d and Phy (ϕ) coefficient were computed as measures of effect size. Pearson correlation or point-biserial testes analyzed the association of study variables with two outcomes: acute and chronic pain (worst pain intensity at 48 h and 3 months). For acute pain, no further analyses were performed because there was only one variable significantly associated with it. The variables significantly associated with chronic pain entered a hierarchical linear regression model, wherein a minimum of 10 events per variable was considered for predictor selection [32]. Multicollinearity was analyzed through the tolerance coefficient (>0.4) and the independence of errors through Durbin–Watson value (2 ± 0.2) [33]. There were no violations of these assumptions.

Moderation models analyzed if sex (moderator, W) interacted with psychological/psychophysical characteristics (independent variables, X) in their influence on postsurgical pain (dependent variables, Y) [34]. The independent variables were psychological and psychophysical characteristics that differed between men and women. Worst pain intensity (48 h and 3 months) was set as dependent variable. Type of arthroplasty was included as a covariate in all models given its relevance for the acute and chronic pain experience after total joint replacement. Additionally, presurgical pain duration and acute postsurgical pain (APSP) were included as covariates in the moderation models with pain at 3 months, given their strong association with that outcome. The continuous predictors were z-transformed. Moderation analyses were performed using PROCESS macro for SPSS [34] (model 1) with 95% bias-corrected confidence intervals and 5,000 bootstrap samples.

The present work is part of a larger study focusing on predictors of pain chronification, and thus, there were no specific power calculations for the analyses. From the larger sample, only a subset of patients completed the QST assessment. Thus, only those with QST at baseline were selected for the present analyses.

3.1 Sample characterization and between-sex differences

This study included 63 participants undergoing TKA (34, 54%) or THA (29, 46%), with a mean age of 64.5 years (SD = 7.43). Thirty-one (49.2%) were female and 32 (50.8%) were male. Baseline sample information and between-sex differences are detailed in Table 1. There were statistically significant differences in body mass index (BMI, t = −2.726, p = 0.008), type of arthroplasty (χ ² = 21.967, p < 0.001), pain duration (t = −2.59, p = 0.013), presurgical pain intensity (t = −2.181, p = 0.033), and interference (t = −2.88, p = 0.006). Most women had TKA, while most men had THA. BMI was higher among women, who also had longer pain duration, more intense pain, and higher pain interference before surgery. ASA score (American Society of Anesthesiologists physical status classification) was retrieved from the clinical record and included 1 (1.6%) case of ASA I, 48 (76.2%) ASA II, and 14 (22.2%) ASA III.

Concerning psychological variables, there were significant differences in anxiety (t = −2.261, p = 0.027) and optimism (t = 3.199, p = 0.002), with women showing more anxiety symptoms and less optimism than men. For QST, there were significant differences in hand MDT (t = −2.667, p = 0.010), WUR (hand: t = −3.646, p = 0.001; joint: t = −2.208, p = 0.033), and CPM at the affected joint (t = −2.061, p = 0.044). Women showed lower mechanical sensitivity (higher MDT) at the hand, higher pain facilitation (higher WUR) at the hand and at the affected joint, and lower pain inhibition (higher CPM) assessed at the affected joint. Individual analysis of the CPM effect revealed the proportion of participants with effective pain inhibition (CPM responders) (Figure 1). An anti-nociceptive effect was shown in 33 (54.1%) participants at the hand (14 [46.7%] males and 19 [61.3%] females), and in 41 (68.3%) participants at the affected joint (23 [76.7%] males and 18 [60%] females).

Figure 1

Individual CPM effect at the hand (a) and affected joint (b) (each bar represents a participant’s score). Men are represented as dashed bars and women are represented as solid bars.

3.2 Surgical outcomes at 48 h and 3 months after total knee/hip arthroplasty

The mean length of hospital stay after surgery was 4.51 days (SD = 1.23), with women having significantly longer hospitalizations than men (t = −2.393, p = 0.02). At 48 h postsurgery, there were no significant between-sex differences in the analyzed variables. However, 3 months after surgery, there were statistically significant differences in pain intensity (t = 2.239. p = 0.029), with women reporting more intense pain than men (Table 2).

3.3 Variables associated with pain intensity at 48 h and 3 months after surgery

The correlation tests did not reveal demographic or clinical variables significantly associated with APSP (Table 3). However, pain at 3 months was significantly associated with sex (r _pb = 0.282, p = 0.029), type of arthroplasty (r _pb = −0.312, p = 0.015), presurgical pain duration (r = 0.353, p = 0.006), and APSP intensity (r = 373, p = 0.004). More intense pain at 3 months was associated with being female, having longer presurgical pain duration, undergoing TKA, and experiencing higher pain at 48 h after surgery. CPM (at the affected joint) was the only QST measure with significant results for either time point, with CPM at the hand being associated with acute pain (r = 0.301, p = 0.019) and CPM at the joint being associated with chronic pain (r = 0.126, p = 0.004). At both moments, pain intensity was higher among patients with less effective pain inhibition (higher CPM score). The analysis of CPM responders vs. non-responders showed comparable results. Participants with effective CPM at the hand had less pain at 48 h (r _pb = −0.288, p = 0.026), and those with effective CPM at the joint had less pain at 3 months (r _pb = −0.353, p = 0.007). None of the psychological variables were associated with pain at 48 h or 3 months after surgery.

Table 4 shows the results of the hierarchical linear regression model combining the previously significant individual predictors of chronic pain. Sex and CPM cease to be significant predictors when the clinical variables are included in the model. Presurgical pain duration (β = 0.276, p = 0.023), type of arthroplasty (β = −0.300, p = 0.034), and APSP intensity (β = 0.370, p = 0.002) are significant predictors of pain at 3 months, with the final model explaining 35% of the variance in CPSP intensity.

3.4 The moderator role of sex

Table 5 shows that sex is a significant moderator in the relationship between WUR (assessed at the joint) (int: b = −1.373, p = 0.046) and acute pain. In males, increased pain facilitation (higher WUR scores) is associated with more intense pain at 48 h (b = 1.486, p = 0.016). For females, there is no relationship between WUR and acute pain (Figure 2a). For this time point, moderation analysis also shows a statistically significant interaction between sex and optimism (int: b = 1.191, p = 0.048), but there are no significant conditional effects revealed (Figure 2b).

Figure 2

Visual representation of the moderator role of sex on the association between WUR (joint) and acute pain (a), optimism and acute pain (b), and CPM (joint) and chronic pain (c).

There is a moderation effect of sex in the association of CPM (at the joint) and CPSP (int: b = 1.625, p = 0.016). Women with less effective pain inhibition (higher CPM scores) have more intense pain three months after surgery (b = 1.543, p = 0.002). The conditional effect for males is nonsignificant (Figure 2c). The effect of the remaining variables on postsurgical pain is not dependent on sex.

4.1 Sex differences

Women had higher pain intensity before and 3 months after surgery than men. Indeed, sex has been suggested as a risk factor for pain chronification [2,3], despite some equivocal evidence [35,36]. It is worth noting that most women in the present study underwent knee surgery, while most men had hip surgery, and it is known that THA has a lower incidence of CPSP than TKA [37]. Since TKA is also associated with more severe postoperative pain and lower improvement rates [38,39], it is possible that the sex-related differences in CPSP are at least partially accounted for by surgery-specific characteristics. Also, women undergo surgery at more advanced arthritis stages than men [40,41], which may contribute to higher presurgical pain intensity and interference in this group. Indeed, this hypothesis is partially supported by longer presurgical pain duration in female participants, though there are no differences in reported disability.

Sex differences in psychological variables are well documented, with different reports in the general population [42] and chronic pain patients [43] showing higher prevalence of anxiety and mood disorders in women. Poorer emotional well-being in women was confirmed in this study in terms of higher anxiety and lower optimism among females but without any differences in depression. There were also no differences in pain catastrophizing, which in some studies was higher in women [44]. Concerning optimism, there are apparently contradictory findings in other studies, which show higher levels in women than men [45]. However, among patients undergoing joint arthroplasty, the results have matched ours [46].

In QST, women showed lower sensitivity to touch (MDT) and increased pain facilitation (assessed by WUR and CPM). Generally, studies do not report sex differences for non-noxious stimuli [47], which makes our finding concerning MDT at the hand rather unexpected, particularly because women were found to be less sensitive than men. It is possible that the inhibitory systems activated to counteract the hyperexcitability of peripheral afferents in the affected joints also contribute to reduced tactile sensitivity [48,49]. In our sample, it could be hypothesized that this phenomenon is more pronounced in women due to longer pain duration and more intense pain in this group.

The absence of differences in PPT was somehow surprising, since studies generally conclude that women are more sensitive to pressure pain than men [22,47]. However, the effect of age should be accounted for, as PPT differences seem to disappear in older age groups [22], in line with our findings.

Women also showed facilitated central integration (WUR at the hand and joint) and more impaired descending modulation (CPM at the joint). Gender differences at this level have been analyzed in a variety of studies, with some evidence of greater facilitation and less effective inhibition in women. However, some non-significant findings in individual studies make these conclusions somewhat inconsistent [50,51]. In the present sample, it is conceivable that women have increased sensitization due to longer pain duration and more intense pain [52].

4.2 Variables associated with APSP and CPSP

Pain-related variables were associated with chronic but not with acute pain. Indeed, pain itself is one of the strongest predictors of future pain [53], with different sources indicating that both presurgical pain and APSP are relevant risk factors for pain chronification [2,36]. The type of arthroplasty should also be accounted for, since TKA is usually associated with higher postsurgical pain intensity than THA [54].

Psychological variables are another relevant category of protective/risk factors, with characteristics such as anxiety, depression, pain catastrophizing, and optimism being frequently associated with postsurgical pain [3,6,7]. Nonetheless, findings concerning psychological domains remain equivocal, with systematic reviews highlighting that many studies fail to find significant associations [36]. Particularly in this study, it is possible that the sample size, along with the globally low levels of emotional distress and low variance in questionnaire scores, accounts for the absence of statistically significant results.

CPM was the only QST measure associated with postsurgical pain, and only in correlation analysis. A recent report also showed that CPM was associated with pain 12 months after TKA but was not a significant predictor in a multivariate model [55]. Nonetheless, less effective CPM still shows a trend towards higher CPSP in this study (p = 0.055), when controlling for the clinical variables. Based on previous reports, it is possible to hypothesize that the predictive value of QST differs according to arthroplasty type [12]. This seems may be visible in the present results since CPM was not a significant predictor when ‘type of arthroplasty’ was included in the regression model.

4.3 The moderator role of sex

Sex was not a moderator in the association between emotional states and postsurgical pain. Previous studies have shown such an effect in chronic pain [19,20] and TKA patients [18], but this was not confirmed by a meta-analysis focusing on postsurgical pain [56].

A significant effect of sex was seen for the association between two QST measures (WUR and CPM) and postsurgical pain. Interestingly, WUR was associated with acute pain in men and CPM with chronic pain in women. The wind-up phenomenon translates a state of neuronal hyperexcitability at the spinal cord that causes a progressive increase in the magnitude of pain evoked by repetitive noxious inputs of the same intensity, being a marker of pain sensitization [47,57]. Indeed, the acute postsurgical period is characterized by continuous nociceptive transmission due to damage and inflammation at the operated joint. Our findings seem to suggest that the impact of increased pain facilitation before surgery on APSP is more pronounced in men.

On the other hand, impaired presurgical descending inhibition seems to be associated with pain chronification only in women. A similar trend was reported by Bossmann et al. [23], reinforcing that accounting for sex may be relevant to improve the predictive value of QST. Our results support the utility of CPM to predict CPSP, at least in a subgroup of patients. CPM is a proxy measure of endogenous inhibitory pain pathways [58], which are important for pain chronification [59]. In patients undergoing thoracotomy, CPM was also shown to predict chronic, but not acute pain [60]. It is likely that endogenous modulation does not have much expression in the acute phase, when the injury itself is likely more relevant [60]. At this stage, the continuous activation of nociceptors (WUR) seems to play a major contribution to increased pain perception.

4.4 Limitations

In this study, the relatively small sample size may have hindered statistical power and limited the ability to detect otherwise significant associations in some analyses.

There are relevant sex differences in clinical variables, which were circumvented by including them as covariates in the moderation analyses (type of arthroplasty, presurgical pain duration, and APSP). Nonetheless, future studies should aim to achieve larger TKA and THA samples, to analyze them independently. The type of arthroplasty may have significant implications for postsurgical pain experience, which deserves more detailed exploration.

Joint pain was not assessed in terms of localized vs. spreading location [61]. Since spreading pain may be associated with greater pain hypersensitivity, this would be relevant to consider in future studies.

Pain medication can affect QST results but, due to the nature of the underlying pathology, patients were not asked to refrain from taking analgesics before the assessment. Subsequent analyses confirmed that there were no differences in QST outcomes between patients who did or did not take medication, so this is also not likely to have influenced results. On the other hand, the variability in pharmacological treatments and analgesic strategies did not allow their inclusion in the statistical models and its influence was not controlled for. The single-center nature of this study limits the generalizability of findings.