Abstract

Purpose:

This study aimed to identify early risk factors for periprosthetic osteolysis after total knee arthroplasty (TKA) and establish clinically useful predictive biomarkers.

Methods:

A retrospective analysis was conducted on patients who underwent TKA at our institution between January 1, 2018, and October 31, 2023. Initially, 397 patients were screened, and 375 met the inclusion criteria after applying strict eligibility standards. Patients were categorized into an osteolysis group (n = 17) and a non-osteolysis group (n = 358). Data on baseline characteristics (age, gender, BMI, diabetes, and hypertension history) and postoperative laboratory results were collected. Logistic regression analyses identified independent risk factors for osteolysis, with subgroup analyses also performed.

Results:

Multivariable logistic regression analysis showed that each 1-unit increase in leukocyte count (OR = 1.36, 95% CI: 1.13–1.65, P = 0.001) and each 1-unit increase in FIB-4 index (OR = 1.94, 95% CI: 1.20–3.14, P = 0.007) were associated with higher odds of osteolysis, whereas each 1-unit increase in the uric acid-to-creatinine ratio (UACR) was associated with lower odds of osteolysis (OR = 0.57, 95% CI: 0.36–0.88, P = 0.012). Subgroup analyses suggested that the strength and significance of these associations varied by sex, age, diabetes status, and hypertension status.

Conclusion:

This study demonstrates that leukocyte count, FIB-4 index, and UACR are independent risk factors for early periprosthetic osteolysis after TKA. These findings may assist in the early identification and management of high-risk patients, thereby reducing postoperative complications and improving patient outcomes.

1 Introduction

Periprosthetic osteolysis remains one of the major causes of failure after total knee arthroplasty (TKA). Severe osteolysis can lead to prosthetic loosening and dislocation (1, 2).This not only directly results in suboptimal surgical outcomes or treatment failure, but also presents significant challenges for revision procedures (3, 4). With the increasing volume of TKA procedures, osteolysis has become one of the main complications affecting long-term prosthesis survival and postoperative function (5).In recent years, as TKA has been increasingly applied to patients with end-stage joint diseases such as knee osteoarthritis, more attention has been given to the management and prevention of postoperative complications.

Osteolysis is thought to be driven primarily by chronic local inflammatory responses to wear debris, particularly polyethylene particle (6–8). Wear particles released from the implant can be recognized by immune cells, especially macrophages, which then produce pro-inflammatory mediators, promote osteoclastogenesis, and ultimately lead to progressive periprosthetic bone resorptio (6, 7, 9). Recent reviews further suggest that this process involves not only the macrophage–osteoclast axis, but also a broader inflammatory cell infiltrate and complex osteoimmune interactions within the periprosthetic microenvironme (6–8). In addition to these biological mechanisms, osteolysis may also be influenced by broader host-related factors. Emerging evidence suggests that liver fibrosis-related metabolic burden and disturbances in renal or uric acid metabolism may be linked to bone remodeling and bone loss in other clinical settings, raising the possibility that hepatic and renal function could also be relevant to periprosthetic osteolysi (10–13).

Imaging modalities such as plain radiography, CT, and MRI are commonly used to evaluate osteolysis after TKA. Among them, conventional radiography remains the most widely used first-line examination because of its accessibility and low cost, but its sensitivity for early osteolytic lesions is limited (14, 15). Although CT and MRI can improve lesion detection and better define lesion size, location, cortical involvement, and surrounding soft tissue changes, they are not routinely performed in all patients during long-term follow-u (15). As a result, early radiographic osteolysis may still be difficult to identify in a timely manne.

While chronic inflammation and metabolic disturbances are increasingly recognized as relevant to osteolysis, simple and accessible laboratory markers for early risk assessment after TKA remain lacking. Previous studies have explored a variety of candidate biomarkers for aseptic loosening, but no widely accepted marker has yet entered routine clinical use (16). Therefore, in this retrospective pilot study, we screened patients for early-stage radiographic osteolysis after TKA and systematically evaluated its associations with routine laboratory biomarkers. The aim was to identify clinically accessible markers that may assist in early risk stratification and postoperative monitoring.

2 Materials and methods2.1 Study design and patients

This was a single-center retrospective study approved by the institutional ethics committee. Patients who underwent total knee arthroplasty (TKA) between January 1, 2018, and October 31, 2023, were screened through the hospital medical record system. A total of 397 cases were initially identified.

Periprosthetic osteolysis was diagnosed on follow-up radiographs based on radiolucent lines >2 mm around the prosthesis, focal bone defects, or loss of trabecular bone structure compared with earlier postoperative radiographs. Early postoperative and approximately 3-month radiographs served as baseline reference images, and radiographs obtained at 1 year or later were used for outcome assessment. Only osteolysis without visible prosthetic loosening or displacement was classified as early-stage osteolysis. All radiographs were independently assessed by two orthopedic surgeons, and cases were included only when both reviewers agreed on the radiographic findings. In cases of disagreement, the final classification was determined by a senior orthopedic surgeon. Because of the retrospective design of the study, formal blinding to all clinical and laboratory information was not implemented. To ensure the completeness and consistency of the dataset, predefined inclusion and exclusion criteria were applied. Eligible patients were required to have standard anteroposterior and lateral radiographs of the operated knee obtained in the early postoperative period and at approximately 3 months after surgery, which served as baseline and early follow-up imaging. In addition, at least one radiographic follow-up assessment at 1 year or later, together with corresponding laboratory test results obtained during the follow-up period, was required for evaluation of radiographic osteolysis and its association with laboratory biomarkers. Only early-stage osteolysis cases, defined as radiographic osteolysis without visible prosthetic loosening or displacement, were included. Patients were excluded if follow-up laboratory data or imaging were unavailable , or if revision surgery had been performed. After screening, 375 patients met the eligibility criteria, including 17 patients with radiographically confirmed osteolysis and 358 without osteolysis.

Baseline demographic and clinical variables were extracted from the records, including age, sex, body mass index (BMI), follow-up interval, history of diabetes (coded as 0 = no diabetes, 1 = diabetes without complications, 2 = diabetes with complications), and history of hypertension . Full baseline characteristics are presented in Table 1. In addition, routine hematological and serological parameters from postoperative follow-up were collected for statistical analysis.

VariableTotal (n = 375)Non-osteolysis (n = 358)Osteolysis (n = 17)P-valueAge, mean ± SD (years)63.00 ± 9.4962.74 ± 9.3968.47 ± 10.180.015Sex, n (%)1.000 Men85 (22.67%)81 (22.63%)4 (23.53%) Women290 (77.33%)277 (77.37%)13 (76.47%)BMI, mean ± SD (kg/m2)25.36 ± 3.6125.31 ± 3.6226.48 ± 3.280.193Diabetes status, n (%)0.008 No diabetes330 (88.00%)318 (88.83%)12 (70.59%) With diabetes, no complications35 (9.33%)33 (9.22%)2 (11.76%) With diabetes + complications10 (2.67%)7 (1.96%)3 (17.65%)Hypertension, n (%)0.778 No186 (49.60%)177 (49.44%)9 (52.94%) Yes189 (50.40%)181 (50.56%)8 (47.06%)

Baseline characteristics of the study population (n = 375).

Data are presented as mean ± standard deviation or number (%). Statistical significance was evaluated using t-test for continuous variables and chi-square test for categorical variables.

2.2 Statistical analysis

Continuous variables were expressed as mean ± standard deviation (x¯ ± s), while categorical variables were presented as frequencies and percentages (n, %). Comparisons between groups were performed using the Student's t-test for continuous variables and chi-square test for categorical variables to evaluate baseline comparability.

To explore associations with periprosthetic osteolysis, univariate logistic regression analysis was first conducted to screen potential predictors. Variables with P-values < 0.05 in univariate analysis were then entered into a multivariate logistic regression model to identify independent risk factors. A bidirectional stepwise regression approach was used to optimize model performance. This method combines forward selection and backward elimination, aiming to balance model complexity and predictive accuracy. The stepwise procedure was as follows: Initialization: The model was initialized either from an empty model (adding variables sequentially) or from a full model (removing variables one by one). Forward selection: Variables with P-values < 0.05 and the greatest incremental improvement to the model were added. Backward elimination: Variables with P-values ≥ 0.05 and minimal contribution were removed. Iteration continued until all variables in the final model had P-values < 0.05. All statistical analyses were conducted using the Statistical analyses were conducted using the Storm Statistics Platform (Zstats, Beijing, China; https://www.zstats.net), R version 4.3.3 (R Foundation for Statistical Computing, Vienna, Austria), and IBM SPSS Statistics version 27.0 (IBM Corp., Armonk, NY, USA). Because several inflammation-related variables, such as leukocyte count, monocyte count, neutrophil count, SIRI, and PIV, are likely to capture overlapping information on systemic inflammation, we first assessed collinearity among them using Spearman correlation analysis and variance inflation factors (VIFs). Given the small number of osteolysis events, these highly correlated markers were not included in the same multivariable model. We therefore constructed several alternative parsimonious models, each including one inflammation-related marker together with FIB-4 index and UACR. The final model was chosen by considering collinearity, model fit, discrimination, and clinical interpretability.

3 Results

A total of 375 patients were included in this study. Among them, 358 patients (95.47%) did not develop radiographic evidence of periprosthetic osteolysis, while 17 patients (4.53%) were diagnosed with early-stage osteolysis. The average age of patients without osteolysis was 62.74 years, while that of those with osteolysis was 68.47 years; the difference was statistically significant (t = −2.45, P = 0.015).

In terms of BMI, the non-osteolysis group had a mean of 25.31 kg/m2, and the osteolysis group had a mean of 26.48 kg/m2; however, the difference was not statistically significant (t = −1.30, P = 0.193).

Regarding diabetes status, in the non-osteolysis group, 88.83% had no history of diabetes, 9.22% had diabetes without complications, and 1.96% had diabetes with complications. In the osteolysis group, 70.59% had no diabetes, 11.76% had uncomplicated diabetes, and 17.65% had diabetes with complications. The distribution difference was statistically significant (P = 0.008).There was no significant difference in hypertension distribution between the two groups (χ2 = 0.08, P = 0.778), nor in sex distribution (χ2 = 0.00, P = 1.000). Detailed results are shown in Table 1.

3.1 Univariate logistic regression

In the univariate analysis, several laboratory and clinical parameters were significantly associated with an increased risk of periprosthetic osteolysis. These included leukocyte count, monocyte count, neutrophil count, SIRI (Systemic Inflammation Response Index), PIV (Pan-Immune-Inflammation Value), age, diabetes status, FIB-4 index, serum creatinine, and direct bilirubin (all P < 0.05). Notably, UACR (uric acid to creatinine ratio) and ALT/AST ratio emerged as protective factors, both demonstrating inverse associations with osteolysis risk. In contrast, BMI, SII, NLR, and several other inflammatory markers showed no statistically significant correlation. Detailed regression results are presented in Table 2.

Significant Variables (P < 0.05)VariableOR (95% CI)P valueLeukocyte count (WBC)1.68 (1.20–2.36)0.003*Monocyte count1.62 (1.18–2.23)0.003*SIRI (Systemic Inflammation Response Index)1.45 (1.13–1.86)0.003*Diabetes group/status2.93 (1.43–6.02)0.003*Neutrophil count1.60 (1.17–2.19)0.004*PIV (pan-immune-inflammation value)1.40 (1.09–1.81)0.010*Age2.25 (1.21–4.16)0.010*FIB-4 index1.55 (1.10–2.16)0.011*UACR (Uric Acid to Creatinine Ratio)0.47 (0.26–0.87)0.016*Creatinine1.39 (1.06–1.83)0.016*ALT/AST ratio0.48 (0.25–0.94)0.033*Direct bilirubin1.53 (1.02–2.28)0.038*Non-significant Variables (P ≥ 0.05)VariableOR (95% CI)P valueCystatin C1.33 (0.99–1.77)0.055Neutrophil percentage1.53 (0.96–2.43)0.071SII (Systemic Immune-Inflammation Index)1.31 (0.97–1.76)0.074Adenosine deaminase1.35 (0.96–1.90)0.082NLR (neutrophil/lymphocyte ratio)1.27 (0.97–1.68)0.086Lymphocyte percentage0.66 (0.41–1.07)0.091Mean platelet volume (MPV)1.44 (0.93–2.22)0.104APRI (Aspartate aminotransferase-to-platelet ratio index)1.29 (0.94–1.77)0.110Urea/creatinine ratio0.61 (0.34–1.12)0.111Large platelet ratio1.43 (0.92–2.23)0.113Indirect bilirubin0.61 (0.33–1.14)0.120Chloride0.70 (0.44–1.10)0.123Alkaline phosphatase (ALP)0.54 (0.23–1.26)0.152BMI (body mass index)1.41 (0.85–2.35)0.187LDL cholesterol0.73 (0.43–1.21)0.219Creatine kinase-MB0.64 (0.31–1.32)0.225BMI grade1.39 (0.79–2.44)0.252Sodium1.34 (0.81–2.23)0.254Eosinophil percentage (EDTA anticoagulated)0.67 (0.34–1.34)0.258Alpha-L-fucosidase0.73 (0.42–1.26)0.260Weight1.30 (0.81–2.10)0.276HALP score1.25 (0.83–1.89)0.282Monoamine oxidase0.75 (0.44–1.27)0.284ALBI grade1.48 (0.72–3.04)0.284Triglycerides0.70 (0.36–1.37)0.302Magnesium1.27 (0.80–2.02)0.307Red blood cell count0.79 (0.50–1.25)0.313Platelet distribution width1.25 (0.80–1.97)0.325Immature granulocyte count1.15 (0.86–1.54)0.341C-reactive protein (CRP)0.43 (0.08–2.46)0.344CAR (C-reactive protein/albumin ratio)0.45 (0.08–2.48)0.362Lipoprotein(a)0.65 (0.25–1.71)0.380LMR (lymphocyte/monocyte ratio)0.76 (0.41–1.40)0.380Amylase1.20 (0.79–1.81)0.395Hemoglobin0.82 (0.52–1.30)0.404Hematocrit0.83 (0.52–1.30)0.406Blood urea1.18 (0.78–1.77)0.437GGT/ALP ratio1.18 (0.78–1.79)0.439Complement C1q0.81 (0.48–1.38)0.440Aspartate aminotransferase (AST)1.14 (0.80–1.63)0.472Uric acid/fasting glucose ratio0.82 (0.48–1.40)0.476ALBI score1.21 (0.71–2.05)0.488Follow-up interval (days)1.14 (0.75–1.75)0.538Prealbumin0.83 (0.47–1.49)0.540Potassium1.16 (0.71–1.90)0.544GFR (glomerular filtration rate)0.85 (0.51–1.44)0.552Red cell distribution width-SD1.13 (0.75–1.72)0.560Mean corpuscular volume (MCV)1.15 (0.71–1.87)0.577Albumin/globulin ratio0.87 (0.53–1.43)0.578Bile acid1.12 (0.74–1.68)0.598Apolipoprotein A-I1.13 (0.70–1.83)0.606Basophil percentage (EDTA anticoagulated)6.19 × 10−12(2.09 × 10−55–1.83 × 10+32)0.613Red cell distribution width-CV0.87 (0.51–1.49)0.620Lactate dehydrogenase (LDH)1.12 (0.71–1.78)0.626Total cholesterol0.88 (0.54–1.46)0.629Alpha-hydroxybutyrate dehydrogenase1.12 (0.70–1.78)0.632Mean corpuscular hemoglobin (MCH)1.13 (0.68–1.87)0.634Gamma-glutamyltransferase (GGT)0.79 (0.28–2.18)0.644PALBI score1.12 (0.68–1.86)0.649Creatine kinase1.07 (0.71–1.61)0.734Apolipoprotein B0.92 (0.56–1.51)0.735Lymphocyte count1.07 (0.67–1.72)0.772Eosinophil count (EDTA anticoagulated)0.92 (0.53–1.60)0.777Hypertension0.87 (0.33–2.30)0.778PAR (platelet/albumin ratio)0.93 (0.57–1.53)0.782Phosphate1.06 (0.66–1.71)0.811Globulin1.06 (0.66–1.70)0.821GPS0.84 (0.19–3.79)0.821Basophil count (EDTA anticoagulated)1.05 (0.66–1.68)0.831Uric acid/albumin ratio0.95 (0.57–1.57)0.835BMI group1.08 (0.51–2.30)0.837Height0.87 (0.22–3.46)0.839Platelet count0.95 (0.58–1.55)0.843Calcium0.96 (0.61–1.51)0.858Monocyte percentage1.04 (0.64–1.67)0.886Mean corpuscular hemoglobin concentration (MCHC)0.97 (0.59–1.57)0.887PLR (platelet/lymphocyte ratio)0.97 (0.59–1.59)0.898Fasting blood glucose1.03 (0.65–1.64)0.904HDL cholesterol1.03 (0.64–1.66)0.905Total protein1.03 (0.63–1.67)0.918Alanine aminotransferase (ALT)0.98 (0.59–1.62)0.922Uric acid0.98 (0.60–1.60)0.924PNI (prognostic nutritional index)1.02 (0.63–1.67)0.926Homocysteine0.98 (0.59–1.61)0.926Serum iron0.98 (0.60–1.60)0.931Sex0.95 (0.30–2.99)0.931Albumin0.98 (0.61–1.60)0.943Immature granulocyte percentage0.98 (0.59–1.63)0.945Plateletcrit1.01 (0.62–1.65)0.958Total bilirubin0.99 (0.61–1.61)0.960GPR (Gamma-glutamyl transpeptidase to platelet ratio)0.99 (0.60–1.64)0.979Uric acid-CRP index1.00 (0.61–1.63)0.988Uric acid/HDL ratio1.00 (0.62–1.63)0.997

Univariate logistic regression variables.

OR, odds ratio; CI, confidence interval.

Bold values indicate statistically significant results.

Several inflammation-related variables were significant in the univariable analysis, including leukocyte count, monocyte count, neutrophil count, SIRI, and PIV. Because these markers are biologically related and likely capture overlapping inflammatory information, collinearity was further assessed. Strong inter-correlations were observed, particularly between leukocyte count and neutrophil count, and between SIRI and PIV (Supplementary Tables S5–S6).

3.2 Multivariate logistic regression

Multivariable logistic regression with bidirectional stepwise selection identified leukocyte count, UACR, and FIB-4 index in the final model. After adjustment, higher leukocyte count and higher FIB-4 index were associated with higher odds of osteol ysis, whereas higher UACR was associated with lower odds. Specifically, each 1-unit increase in leukocyte count was associated with a 36% increase in the odds of osteolysis, each 1-unit increase in UACR was associated with a 43% decrease in the odds of osteolysis, and each 1-unit increase in FIB-4 index was associated with an approximately 94% increase in the odds of osteolysis. Full model outputs are shown in Table 3.

VariableβOR95%CIP valueFIB-4 Index0.6612301.9371.196–3.1360.0072UACR (Uric Acid to Creatinine Ratio)−0.5703170.5650.362–0.8840.0124Leukocyte count (WBC)0.3091711.3621.127–1.6460.0014

Multivariable logistic regression analysis results: factors associated with the occurrence of periprosthetic osteolysis.

OR, odds ratio; CI, confidence interval.

Several inflammation-related variables were significant in the univariable analysis and appeared to capture overlapping information. Collinearity analysis showed substantial inter-correlation among these markers. In the subsequent comparison of alternative parsimonious models, the WBC-based model showed the most favorable overall performance, with the lowest AIC, the highest AUC, and the highest specificity while maintaining acceptable sensitivity. Although the neutrophil-based model showed higher sensitivity, this gain was accompanied by lower specificity and less favorable overall model fit, suggesting