Using a convenience sampling approach, we enrolled patients undergoing TKA at a tertiary Grade-A general hospital in China between June 2024 and February 2025. Inclusion criteria: Age ≥18 years; Scheduled for primary unilateral TKA; Willing to participate and provide informed consent. Exclusion criteria: Required pain rescue medication ≥2 times within 24 h; Impaired Chinese language comprehension or communication; Chronic opioid use; Participation in other clinical trials during the study period; Development of severe acute complications during observation. Hertzog's (14) study pointed out that a cohort of 200 people can achieve more than 80% statistical efficiency at five time points. Considering a 20% loss to follow-up rate, the sample size should be no less than 240 cases.

This study adopted a longitudinal multi-timepoint design with data collection at the following intervals: 1–2 days preoperatively (T0), postoperative day 1 (T1), day 2 (T2), day 3 (T3), day 5 (T4), and 3 months postoperatively (T5). Before survey administration, researchers explained the study purpose, significance, and questionnaire completion methods in detail to participants and obtained informed consent. Patients completed questionnaires independently, while for those unable to do so, researchers conducted face-to-face interviews and faithfully recorded responses. At T0, researchers administered paper-based versions of the general information questionnaire, NRS, PCS, HADS, and APGAR questionnaire through face-to-face interviews in orthopedic wards. For postoperative assessments at T1, T2, T3, and T4, patients' self-reported pain levels during activity (daily postoperative exercises, walking, and flexion/extension movements assisted by a rehabilitation physician) were collected using NRS at 5 PM each day through face-to-face interviews. At the 3-month postoperative follow-up (T5), patients' joint functional recovery was assessed via telephone using the WOMAC scale. To ensure data accuracy and reliability, this study implemented rigorous quality control measures. First, all collected data were processed and entered by two independent researchers. Second, all patients received standardised basic analgesia. When breakthrough pain persisted for 30 min (21), a rescue dose of 50 mg of buccinnazine hydrochloride was administered via intramuscular injection. The basic analgesia protocol included preoperative pain management education and intravenous infusion of cyclooxygenase-2 (COX-2) inhibitors for prophylactic analgesia; intraoperative periarticular “cocktail” injection with a formulation of ropivacaine, epinephrine, ketorolac, and morphine, diluted with normal saline to a total volume of 40 ml; postoperative intravenous infusion of nonsteroidal anti-inflammatory drugs and oral tramadol tablets; and concurrent use of ice packs, ear acupuncture, Chinese herbal poultices, and moxibustion for traditional Chinese medicine analgesia. Third, to best capture the natural progression of postoperative pain, pain scores were recorded immediately before any rescue medication administration. Fourth, the Patient-Controlled Analgesia (PCA) weaning protocol: On the morning of the first postoperative day, when the patient's pain is stably controlled (NRS rest score consistently ≤4), without severe side effects, and once mobilization has begun, the process is initiated. First, discontinue the background infusion of the PCA pump while retaining the PCA bolus function for rescue use, and simultaneously initiate regular oral administration of tramadol. Monitor the patient's frequency of rescue oral medication requests and pain scores. If over the next 4–6 h, the patient's pain remains well-controlled without frequent use of PCA bolus (e.g., usage frequency <2 times/4 h), completely discontinue the PCA. If pain becomes uncontrolled (NRS ≥7) after discontinuation, restart the PCA background infusion and reassess after 4 h. The study did not interfere with clinical analgesic decisions, prioritizing patients' pain management needs throughout. This study was approved by the hospital ethics committee (Approval No. 2024-10-005).

Data analysis was performed using SPSS 27.0 and Mplus 8.3 software. Categorical variables were described using frequencies and percentages, with between-group comparisons conducted using chi-square tests or Fisher's exact test. Measurement data following a normal distribution are described using mean ± standard deviation. Intergroup comparisons were performed using the independent samples t-test and one-way analysis of variance (ANOVA). Non-normally distributed continuous data were described using medians and interquartile ranges, with between-group comparisons performed using the Mann–Whitney U-test and the Kruskal–Wallis H-test. Latent growth curve modeling was employed to characterize the overall developmental trajectory of APSP in patients undergoing TKA. GMM was employed to examine the changing trajectories of APSP in TKA patients across T1-T4 time points and to identify potential heterogeneous subgroups. Logistic regression analysis was used to assess the influence of relevant variables on APSP, while multiple linear regression was applied to investigate the relationship between trajectory patterns and joint functional recovery at 3 months postoperatively. A p-value <0.05 was considered statistically significant.

A total of 252 questionnaires were distributed. Among these, 11 cases were lost to follow-up, and 14 cases required rescue medication≥2 times within 24 h, resulting in 227 valid questionnaires retrieved. The effective response rate was 89.72%, as detailed in Figure 1. The general characteristics and scale scores of the surveyed participants are presented in Table 1. A comparison of baseline characteristics between the excluded and included groups showed no statistically significant differences in any indicators, with observed effect sizes being minimal. This indicates that the exclusion process did not introduce significant selection bias, and the final sample included for analysis demonstrated good representativeness at baseline, thereby supporting the internal validity of subsequent findings. Details are provided in Supplementary Table S1. Since the proportion of participants lost to follow-up was less than 5%, only descriptive statistics of their baseline information are presented in Supplementary Table S2.

Latent growth curve modeling was employed to fit both linear and quadratic (nonlinear) models. Based on model fit indices, the linear model was ultimately selected for further analysis, as detailed in Table 2. GMM was applied by incrementally increasing the number of classes from 1 to 3. No covariates were included in any of the models. As the number of classes increased, the values of Information Criterion (AIC), Bayesian Information Criterion (BIC), and Adjusted Bayesian Information Criterion (ABIC) progressively decreased. When the number of classes was set to three, the Lo-Mendell-Rubin (LMR) test did not reach statistical significance (P = 0.831). Moreover, the smallest trajectory subgroup accounted for only 5.7% of the total sample, comprising a relatively small number of individuals. This subgroup demonstrated low clinical interpretability and lacked credibility for generalization. Therefore, after comprehensive consideration of clinical utility and the above model fit indices, the two-class linear GMM was ultimately retained as the optimal model. Detailed fit indices are presented in Table 3. The average posterior probabilities for class membership were 0.980 and 0.972 for each class, respectively. The mean pain score distributions across postoperative time points for each class are shown in Figure 2.

The two distinct latent class trajectories of postoperative pain in TKA patients demonstrated different characteristics at each time point. Based on the changing patterns and features of pain trajectories, each latent class was named accordingly. See Supplementary Table S3 for details. In Class 1 (C1), patients exhibited higher initial pain levels (intercept = 6.956, P < 0.001) with a relatively slower decline over time (slope = −0.494, P < 0.001), maintaining moderate pain levels even on postoperative day 5. Therefore, C1 was designated as the “Moderate-High Persistent Pain” group. In Class 2 (C2), patients showed lower initial pain levels (intercept = 5.631, P < 0.001) with a steeper declining trend (slope = −0.631, P < 0.001), demonstrating significant pain relief by postoperative day 5. Consequently, C2 was named the “Moderate-Low Rapid Relief” group.

The results demonstrated statistically significant differences (P < 0.05) in APSP trajectory development among TKA patients based on age, gender, sleep quality, comorbidities, history of knee replacement, preoperative NRS scores, PCS scores, FCIQ scores, and anxiety/depression levels (Table 1).

Using APSP trajectory categories as the dependent variable (with “Moderate-Low Rapid Relief” as reference) and incorporating all univariate predictors with P < 0.05, logistic regression analysis identified age, preoperative NRS, PCS, and FCIQ scores as significant independent predictors of APSP trajectories (P < 0.05, Table 4).

After controlling for age, preoperative NRS, PCS, and FCIQ scores, multiple linear regression analysis was performed with APSP trajectory categories as the independent variable and WOMAC scores as the dependent variable. Dummy variable coding was applied, using the “Moderate-Low Rapid Relief” group as the reference (coded as 0), while the “Moderate-High Persistent Pain” group was coded as 1. As shown in Table 5, compared to TKA patients in the “Moderate-Low Rapid Relief” subgroup, those in the “Moderate-High Persistent Pain” subgroup were associated with significantly worse functional outcomes (β = 0.32, t = 5.64, p < 0.001).

Using GMM, this study identified two latent classes of APSP trajectories in TKA patients: the moderate-high persistent pain group (45.2%) and the moderate-low rapid relief group (54.8%), demonstrating the heterogeneous nature of postoperative pain, Similar to the conclusions of Thomazeau (10). Furthermore, Thomazeau found at the 6-month postoperative follow-up that the high pain intensity group had a significantly higher incidence of chronic pain compared to the low pain intensity group (10). Therefore, healthcare providers need to early identify patients with high pain scores and low rates of pain relief, promptly adjust intervention strategies, and implement stepped, personalized treatment measures. Rehabilitation therapists should adopt differentiated rehabilitation interventions based on distinct pain trajectories. The moderate-high persistent pain group exhibited significant prolonged postoperative pain characteristics, with activity-related NRS scores remaining at relatively high levels during the first 5 postoperative days. This may be associated with preoperative central sensitization (22), health status (7), and psychological factors (23). Therefore, for this subgroup, comprehensive management strategies should be implemented, including enhanced multimodal analgesia, psychological interventions, modified rehabilitation protocols, and surgical optimization to prevent pain chronification. Patients in the “moderate-low rapid relief group” subgroup exhibited a rapid decline in postoperative activity-related pain, suggesting a favorable response to standard multimodal analgesia. This subgroup may derive greater benefit from ERAS protocols, thereby optimizing functional outcomes. The underlying neurophysiological and psychological mechanisms warrant further investigation.

The results demonstrated that the APSP trajectory served as an independent predictor. Compared to patients in the moderate-low rapid relief group, those in the moderate-high persistent pain group exhibited significantly worse WOMAC scores, with this variable alone increasing the model's explained variance by 6.2%. Notably, it is predictive potency (β = 0.32) even exceeded that of variables such as age, preoperative pain level, and pain catastrophizing. These findings indicate that patients experiencing severe acute movement-related pain face substantially elevated risks of poor functional recovery, corroborating previous studies by Singh (9) and Lo (34). The underlying mechanism may involve kinesiophobia induced by intense pain, which significantly reduces patients' willingness and frequency to participate in early rehabilitation exercises, thereby delaying functional recovery (35). Contemporary research in pain medicine has demonstrated that pain is an active process resulting from the interplay of physiological and psychological factors, and that pain perception can be effectively modulated through psychological interventions (36). Therefore, healthcare providers can utilize psychological approaches such as preoperative health education, cognitive-behavioral therapy, mindfulness training, and pain empathy to enhance patients' emotional regulation, alleviate postoperative pain, and improve self-management capabilities. Notably, during the follow-up period, we observed that some patients with high early pain scores showed significant improvement in activity-related pain at 3 months postoperatively, yet demonstrated limited improvement in joint function. This “pain-function recovery dissociation” suggests that pain relief and functional recovery may be mediated by distinct pathophysiological mechanisms. Further investigation into the underlying influencing factors is warranted, as traditional pain-oriented postoperative management strategies may be insufficient to ensure optimal functional outcomes. Future studies should establish a dual-track evaluation system integrating both pain and functional recovery to further elucidate the relationship between these two domains.