Complement factors C3 and C5 as tissue biomarkers for differentiating periprosthetic joint infection from aseptic implant failure

Abstract

Background

The diagnosis of periprosthetic joint infection (PJI), and particularly low-grade ((LG) PJI) infections, remains a clinical challenge. Systemic inflammatory markers sometimes yield false-negative results in inapparent cases, complicating the differentiation from aseptic loosening (AL). This exploratory study evaluated tissue-based complement factors C3 and C5 as biomarkers compared to terminal factor C9 and established systemic parameters.

Methods

Fifty-five patients undergoing revision arthroplasty were categorized into high-grade PJI (HG) PJI, N = 20), (LG) PJI (N = 14), and AL (N = 21). Serum parameters (C-reactive protein (CRP), white blood cells (WBC)) were analyzed by immunohistochemistry for tissue-based C3, C5, and C9. Diagnostic performance was assessed via ROC analysis.

Results

Pathogen spectra differed significantly, with (LG) PJI dominated by Cutibacterium acnes and (HG) PJI by staphylococci. ROC analysis showed that serum CRP accurately identified (HG) -PJI, but failed in (LG) PJI. WBC demonstrated limited accuracy. In contrast, C3 detection in tissues exhibited superior performance, particularly for identifying (LG) PJI versus (HG) PJI. C5 detection remained robust in both PJI types. Terminal factor C9 yielded lower overall discrimination and significant cross-reactivity in AL cases.

Conclusions

Central complement factors C3 and C5 serve as robust tissue biomarkers for PJI, outperforming C9 and systemic parameters. Their implementation may improve clinical decision-making, especially in (LG) PJI cases, where standard diagnostic criteria remain inconclusive.

Introduction

The complement system is a fundamental component of the innate immune system, playing a pivotal role in pathogen detection and elimination [1]. Activation occurs via three initiation pathways—the classical, the alternative, and the lectin pathway—which all converge in the cleavage of the central component C3 and lead to the terminal complement pathway [2–4]. This hierarchically organized cascade is essential to the host immune response. Complement factor C3 acts as a central hub; its cleavage product, C3b, serves as a primary opsonin that facilitates phagocytosis by immune cells [5, 6]. Subsequently, complement factor C5 is cleaved by C5 convertase to initiate the terminal phase. The resulting anaphylatoxin, C5a, is a potent chemotactic agent that recruits immune cells to inflammatory sites, while C5b directs the assembly of the membrane attack complex (MAC) [7, 8]. Finally, the polymerization of C9 forms the MAC, inducing osmotic lysis of the target pathogen [9–11].

Recent studies have highlighted the potential of C9 as a biomarker for PJI [12, 13]. PJI is the most severe complication following joint replacement procedures and is associated with high morbidity and mortality [14–17]. Management typically requires extensive surgical revision and prolonged antibiotic therapy [18–20]. The clinical presentation of PJI varies widely, ranging from acute, febrile courses to inapparent, chronic presentations known as low-grade PJI. Typical symptoms include localized pain, restricted motion, and classic inflammatory signs such as erythema, swelling, and joint warmth [18, 21, 22]. Standard PJI diagnosis involves a multimodal strategy combining clinical findings, laboratory data, and microbiological analysis, as established by the International Consensus Meeting (ICM) [23]. Laboratory diagnostics rely on serum inflammatory markers, such as CRP and WBC, alongside synovial fluid analyses, including leukocyte count and neutrophil percentage. Additionally, intraoperative findings, histopathology, and aerobic/anaerobic cultures of synovial fluid and periprosthetic tissue are utilized [23–28].

Low-grade PJI remains a significant diagnostic challenge. Clinical presentations are often subtle, frequently manifesting only as persistent pain, which complicates the differentiation from aseptic loosening. In these cases, established serum markers (CRP, WBC) often remain within normal or marginally elevated ranges, leading to high false-negative detection rates [19, 29–31].

Low-grade PJIs are primarily caused by less virulent organisms [32], such as coagulase-negative staphylococci (CoNS), which represent the most common pathogen group in PJI overall, followed by Cutibacterium species. In contrast, high-grade infections are dominated by virulent pathogens such as Staphylococcus aureus [19, 30, 33]. Since the low-virulent organisms are part of the normal skin flora, their presence in cultures is often misinterpreted as a contamination rather than the causative agent of a genuine infection [34]. The estimated rate of culture-negative PJIs (CN-PJI) ranging from 5 to 45% [33, 35–38], further highlights current diagnostic limitations. Reliable diagnostics must provide high specificity to distinguish subtle infectious signals from non-specific inflammation, such as the inflammatory response induced by wear particles in aseptic loosening. Currently, no standard biomarker exists for PJI diagnosis [25, 28, 39, 40]. In recent years, recent approaches have been developed to reduce diagnostic uncertainty, particularly through the analysis of synovial fluid. While synovial CRP [41, 42] and alpha-defensin [43–45] have shown promising results in studies, they possess inherent limitations. Synovial CRP remains a non-specific marker susceptible to influence from non-infectious tissue damage [46]. The alpha-defensin immunoassay, though accurate, is not routinely available, it is costly [47, 48], and can yield false-positive results in the presence of metallosis [40].

Previous research on complement activation as a PJI biomarker focused on the detection of the terminal factor C9; however, this approach showed cross-reactivity with aseptic processes [13] and did not sufficiently distinguish between (HG) PJI and (LG) PJI. Given the role of complement activation in defence against pathogens and the limitations of current markers, this exploratory study focuses on tissue-based diagnostics using immunohistochemistry. We investigate the presence of central complement factors C3 and C5 in periprosthetic tissue, comparing their diagnostic utility to terminal factor C9 and established systemic inflammatory markers.

Methods

Study cohort

In this study, a total of 55 patients were included who underwent revision surgery of a total hip arthroplasty (THA) or a total knee arthroplasty (TKA). Demographic data of the patients, including age, gender, and type of implant, were collected. Patients with known inflammatory joint diseases (such as rheumatoid arthritis (RA)) were defined as an exclusion criterion and were not included in the study. During revision surgery, 3 to 5 periprosthetic tissue specimens were harvested per patient from macroscopically inflamed and altered areas of the periprosthetic synovium for routine histopathological and microbiological diagnostics. For the experimental investigations of this study, additional tissue specimens from macroscopically altered areas were collected for processing.

Diagnostic classification and PJI classification

Diagnostics and treatment of the patients were performed according to the in-house algorithm for the identification of periprosthetic joint infection (PJI), based on the criteria of the Musculoskeletal Infection Society (MSIS) and the ICM [23].

The patients were assigned to the infectious cohort (PJI) if at least one major criterion or multiple minor criteria were met:

• Major criteria: Evidence of two positive microbiological cultures with the same organism or positive tissue culture.

• Minor criteria: Elevated clinical suspicion (medical history), serum CRP > 5 mg l⁻¹, serum WBC >10 Gpt/l, or histological evidence of typical signs of infection (neutrophil infiltration).

For further differentiation, the infectious cohort was subdivided into (LG) and (HG) PJI. This classification was performed post-hoc within the framework of the study and was based on modified criteria of the World Association against Infection in Orthopaedics and Trauma (WAIOT) [49], adapted to the available diagnostic findings of the clinic. Here, a scoring system consisting of “Rule-In” and “Rule-Out” tests was applied:

• Rule-Out tests: Every negative test result was scored −1 (positive tests = 0).

  • ◦
    Serum CRP (threshold >10 mg L⁻¹)
  • ◦
    Synovial WBC (threshold >1,500/μL⁻¹)
• Rule-In tests: Every positive test result was scored +1 (negative tests = 0).

  • ◦
    Clinical examination: Presence of pus, sinus tract, or exposed prosthesis.
  • ◦
    Synovial fluid: Positive culture evidence or WBC >3,000/mL⁻¹.
  • ◦
    Histology (frozen section): >5 neutrophils in at least 3 “High-Power-Fields” (HPF).

Based on this score and the clinical presentation, the classification was made as follows:

1. (LG) PJI: Assumed in the absence of acute local signs of inflammation, but presence of otherwise unexplained pain, swelling, or functional impairment. A prerequisite was a ratio of positive “Rule-In” to negative “Rule-Out” tests of ≥0 as well as positive histology and/or at least one positive microbiological sample.

2. (HG) PJI: Assumed in a clinical condition with two or more acute local signs of inflammation (pain, swelling, erythema, hyperthermia, functional loss). A prerequisite was a ratio of positive “Rule-In” to negative “Rule-Out” tests of ≥1 as well as positive histological and/or microbiological findings.

Microbiological diagnostics

The collected periprosthetic tissue samples were first mechanically minced and subsequently homogenized using an Ultra-Turrax Drive Control disperser (IKA®-Werke GmbH & Co. KG, Staufen, Germany) (6,000 rpm for 2 min with interval direction change). The sample homogenate was inoculated onto Columbia agar with 5% sheep blood (Becton Dickinson, Heidelberg, Germany), chocolate agar, and Schaedler agar (Oxoid, Munich, Germany) and incubated under aerobic conditions with 5% CO₂ as well as anaerobically at 35 ± 1 °C. Additionally, inoculation into thioglycolate and Schaedler broth (bioMérieux, Marcy L'Etoile, France) was performed for 14 days at 35 ± 1 °C. Identification of the detected pathogens was performed using Matrix-Assisted Laser Desorption/Ionization Time-of-Flight Mass Spectrometry (MALDI-TOF MS) using the VITEK® MS System (bioMérieux, Marcy L'Etoile, France).

Immunohistochemical staining and image analysis

The periprosthetic tissue was fixed in 4% formaldehyde for 24 h and embedded in paraffin. To standardize the processing of the non-uniform tissue, specimens were examined by experienced medical technical laboratory assistants who selected macroscopically conspicuous regions and embedded them into 1 to 3 paraffin blocks per patient. From these, one tissue block per patient exhibiting the most pronounced alterations was selected and cut into 4 μm thick sections. Immunofluorescence staining was performed using antibodies against C3, C5, and C9; corresponding IgG antibodies served as isotype controls. Epitope unmasking was performed for C3 and C9 using heat in citrate buffer (pH 6) and for C5 enzymatically using pepsin (0.25 mg ml⁻¹ for 45 min at 37 °C). Primary antibodies were diluted in 4% bovine serum albumin (BSA) and incubated overnight at 4 °C. The specific concentrations and dilution ratios were as follows: anti-C3 (0.16 mg ml⁻¹, 1:100), anti-C5 (1.0 mg ml⁻¹, 1:200), and anti-C9 (0.62 mg ml⁻¹, 1:250). An Alexa Fluor 555 conjugate (1:1000 in PBS) was used as the secondary antibody, followed by nuclear staining with DAPI. First, the tissue section was scanned at low magnification (100x) to identify regions of cellular infiltration and distinct immunofluorescence 'hotspots'. Within these identified hotspots, five representative images were acquired per sample at 400x magnification using a Zeiss Axio Observer.Z1 microscope, avoiding tissue folds, staining artifacts, or extensive necrotic areas. The positive immunofluorescence area was calculated as a percentage of the total tissue area using ImageJ software, applying a fixed intensity threshold determined by the corresponding isotype controls to subtract background noise. To account for variations in cellularity, this percentage area was normalized to the number of cell nuclei detected in each field. To ensure the reliability of the measurements, the background signal of the corresponding isotype control was subtracted from the signal of the antibody-treated sample. The final value for each patient was determined by calculating the median of the five images.

Statistical analysis

Statistical analysis was performed using GraphPad Prism software version 10 (GraphPad Software, La Jolla,CA, USA). The Shapiro-Wilk test was applied to test for normal distribution. For significance tests between the groups, the Kruskal-Wallis test was used. ROC curve analysis was used to evaluate diagnostic performance, utilizing Youden's index to calculate optimal threshold values (cut-offs) for the biomarkers. A P-value <0.05 was considered statistically significant.

Results

Cohort description

This study included 55 patients undergoing revision surgery for THA TKA. The cohort was divided into a periprosthetic joint infection group (PJI; N = 34) and AL group (N = 21). Following the WAIOT criteria, the PJI cohort was further categorized into (HG) PJI (N = 20) and (LG) PJI (N = 14) infections. Mean age was comparable across all groups (Table 1).

Table 1.

Demographic and clinical characteristics of the study population

Cohort	Number of samples	Age [years]	Gender	Type of arthroplasty
(HG) PJI	20	72 ± 9	♀ 7	TKA 12
			♂ 13	THA 8
(LG) PJI	14	71 ± 7	♀ 2	TKA 6
			♂ 12	THA 8
AL	21	72 ± 9	♀ 10	TKA 12
			♂ 11	THA 9

The gender distribution of the (HG) -PJI cohort consisted of 7 women and 13 men (12 TKA, 8 THA). The (LG) PJI cohort included 2 women and 12 men (6 TKA, 8 THA), and the AL cohort comprised 10 women and 11 men (12 TKA, 9 THA). Detailed demographic data are provided in Table 1.

Microbiological analysis revealed distinct pathogen distributions between the groups. (HG)-PJI cases were primarily associated with high-virulence pathogens, specifically S. aureus (25%) and Staphylococcus epidermidis (20%). In contrast, the (LG) PJI spectrum was dominated by low-virulence, indolent organisms, notably Cutibacterium acnes (43%) (Fig. 1A). Serum WBC counts did not differ significantly across the three groups (Fig. 1B). Analysis of serum CRP levels showed significantly higher values in the (HG) PJI group (median: 48.92 mg l⁻¹) than in the AL cohort (median: 1.99 mg l⁻¹). However, no statistically significant difference in CRP levels were found between the (LG) PJI cohort (median: 3.26 mg l⁻¹) and the AL group (Fig. 1C).

Fig. 1.

Pathogen spectrum and systemic inflammatory markers by infection severity. (A) Pathogen spectrum comparison between (HG) and LG infections. Data represent the relative proportion (%) of each total cohort. (B) WBC counts: No significant differences were observed between the groups. (C) Serum CRP levels: (HG) PJI median 48.92 mg l⁻¹, IQR [9.68–102.30 mg l⁻¹], P < 0.0001 ((HG) PJI vs. AL); (LG) PJI median 3.26 mg l⁻¹, IQR [1.33–7.15 mg l⁻¹], P > 0.05 (LG) PJI vs. AL); AL median 1.99 mg l⁻¹, IQR [1.48–4.53 mg l⁻¹]. Data are shown for patients with AL (N = 21), (LG) PJI (N = 14), and (HG) PJI (N = 20). Statistical analysis utilized the Kruskal-Wallis test followed by Dunn's multiple comparison test. Significance levels: ***P < 0.001, ****P < 0.0001; non-significant values (ns) are not shown. Bars indicate medians with interquartile ranges

Tissue staining of C3 and C5 can differentiate between PJI and AL

Immunohistochemical examination of the synovial membranes enabled the quantification of local complement activation in all 55 patients. The area of positive immunofluorescence for C3, C5, and C9 (red) was quantified and normalized to the number of cell nuclei (blue), while the background signal was corrected by subtracting the values of the corresponding isotype control. Microscopically, C3 staining was primarily localized within infiltrating immune cells, particularly neutrophil granulocytes (Fig. 2A). Quantitative analysis showed an about 20-fold increase in C3 stained area in both PJI groups compared to the AL control. In the (HG) PJI group, the median positive area was 0.00033%. Similarly, the (LG) PJI group showed a significant increase with a median of 0.00058%. In contrast, C3 signals were nearly undetectable in the AL cohort (0.00000%). No significant difference was found between the (HG) PJI and L(G) PJI cohorts (P > 0.9999) (Fig. 2D).

Fig. 2.

Immunohistochemical detection and quantification of complement factors C3, C5, and C9 in periprosthetic tissue. (A–C) Representative images of the synovial membrane stained for (A) C3, (B) C5, and (C) C9. Panels display samples from (HG) PJI (top), L(G) PJI (middle), and aseptic loosening (bottom). Target antigens appear in red, with DAPI counterstaining for nuclei (blue). Images were captured at 400x magnification (scale bar: 50 µm). (D–F) Quantification of fluorescence areas for: (D) C3 (HG) PJI median 0.00033%, IQR [4.97 × 10⁻⁵–0.00177%], P = 0.001 ((HG) PJI vs. AL); (LG) PJI median 0.00058%, IQR [0.00024–0.00108%], P = 0.0005 ((LG) PJI vs. AL); AL median 0.000%, IQR [0.00–0.00010%]. (E) C5 (HG) -PJI median 0.00021%, IQR [1.36 × 10⁻⁵–0.00161%], P < 0.0001 ((HG) PJI vs. AL); (LG) PJI median 0.00016%, IQR [2.55 × 10⁻⁶–0.00120%], P = 0.0007 ((LG) PJI vs. AL); AL median 0.000%, IQR [0.00–0.00%]. (F) C9 (HG) PJI median 0.00513%, IQR [0.00104–0.01140%], P = 0.0496 ((HG) PJI vs. AL); (LG) PJI median 0.00290%, IQR [0.00038–0.01078%], P = 0.6539 ((LG) PJI vs. AL); AL median 0.00088%, IQR [0.00029–0.00574%]. Positive fluorescence areas were normalized to nuclei count and presented as scatter dot plots for aseptic (N = 21), (LG) PJI (N = 14), and (HG) PJI (N = 20) groups. Statistical analysis utilized the Kruskal-Wallis test with Dunn's multiple comparison test. Significance levels: *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001; non-significant values (ns) are not shown. Bars indicate medians with interquartile ranges (IQR)

Factor C5 exhibited a similar distribution pattern. Morphologically, C5 staining occurred predominantly at the vessel walls of capillaries and in vessel-associated cells (Fig. 2B) Both (HG) PJI (median: 0.00021%) and (LG) PJI (median: 0.00016%) showed significantly larger stained area than the AL cohort (median: 0.000%). Again, there was no significant difference between the two infectious subgroups (P > 0.9999) (Fig. 2E).

Analysis of the terminal factor C9 showed lower diagnostic discrimination than the central factors C3 and C5. Microscopic examination revealed diffuse, perinuclear C9 staining without clear cell-type specificity (Fig. 2C). A six-fold increase in C9 stained area was observed compared to the AL group (median: 0.00088%) and was detectable only in the (HG) PJI cohort (median: 0.00513%; P = 0.0496). No significant difference was observed between the (LG) PJI (median: 0.00290%) and AL groups (Fig. 2F).

The ROC curve shows a good predictive value for C3 and C5 to discriminate between PJI and aseptic loosening also in case of low-grade PJI

Receiver operating characteristic (ROC) analysis revealed that serum CRP serves as an excellent marker for distinguishing (HG) PJI from AL (AUC: 0.90), yet it provides insufficient diagnostic accuracy for discriminating between (LG) PJI and AL (AUC: 0.54) (Fig. 3B). Serum WBC counts demonstrated low diagnostic utility for both (HG) PJI (AUC: 0.64) and (LG) PJI (AUC: 0.65) compared to the AL cohort (Fig. 3A).

Fig. 3.

Receiver operating characteristic (ROC) analysis comparing (HG) and (LG) PJI to AL. Diagnostic performance was evaluated for the differentiation of (HG) PJI (black curves) and (LG) PJI (grey curves) from the AL.(A) WBC; (HG) PJI AUC: 0.64; (LG) PJI AUC: 0.65). (B) CRP; (HG) PJI AUC: 0.90; (LG) PJI AUC: 0.54). Normalized fluorescence areas for (C) C3 ((HG) PJI AUC: 0.80; (LG) PJI AUC: 0.91), (D) C5 ((HG) PJI AUC: 0.87; (LG) PJI AUC: 0.85), and (E) C9 ((HG) PJI AUC: 0.72; (LG) PJI AUC: 0.62). The dashed diagonal line represents random classification (AUC: 0.5)

In contrast, ROC analysis confirmed the high diagnostic utility of tissue-based C3, particularly for the identification of (LG) PJI versus AL (AUC: 0.91). At the optimal threshold determined by the Youden index, C3 provided a sensitivity of 71.43% and a specificity of 100% for this subgroup. For (HG) PJI versus AL, C3 remained a good marker, with slightly lower accuracy (AUC: 0.80; sensitivity: 65.00%; specificity: 85.71%) (Fig. 3C).

Similarly, factor C5 showed good diagnostic performance across both infection severities. For (HG) PJI versus AL, C5 achieved an AUC of 0.87 (sensitivity: 80.00%; specificity: 90.48%), while for (LG) PJI versus AL, it maintained a high predictive value with an AUC of 0.85 (sensitivity: 78.57%; specificity: 85.71%) (Fig. 3D).

The terminal factor C9 exhibited overall lower diagnostic discrimination. While (HG) PJI reached an AUC of 0.72 (sensitivity: 95.00%; specificity: 47.62%), the discrimination for (LG) PJI was less effective (AUC: 0.62; sensitivity: 35.71%; specificity: 90.48%) (Fig. 3E).

Discussion

This study investigated the immunohistochemical detection of complement factors C3, C5, and C9 in the synovial membrane as potential biomarkers for PJI in patients following THA or TKA revision. Diagnosing PJI remains a clinical challenge, as indolent cases are often difficult to distinguish from AL. Our data emphasize the need for local, tissue-based markers to bridge the diagnostic gap between (LG) PJI and AL.

As expected, pathogen spectra differed significantly between the infectious phenotypes. While the (HG) PJI was predominated by pathogens such as S. aureus and S. epidermidis, the (LG) PJI spectrum was characterized by low-virulence organisms, particularly Cutibacterium acnes (43%). This observation aligns with current research identifying staphylococci as the most frequent causes of PJI [21, 50] whereas C. acnes and other anaerobic pathogens primarily appear in delayed PJI or chronic cases [20, 51].

Consistent with existing literature, PJI in our cohort was associated with elevated systemic inflammatory parameters [29, 52, 53]. Although serum CRP levels were significantly higher in (HG) PJI patients, the diagnostic power of this marker was highly dependent on the infectious phenotype. ROC analysis revealed that while CRP is an excellent marker for distinguishing (HG) PJI from AL, it provides insufficient diagnostic accuracy for discriminating between (LG) PJI and AL. Indeed, a significant proportion of septic patients lacked pathological CRP levels, while some patients in the aseptic group showed elevated values. Notably, in the (LG) PJI cohort, no statistically significant difference were found compared to the AL group. These findings confirm that the predictive value of serum CRP for PJI is limited in indolent cases, as systemic markers are often influenced by comorbidities or other inflammatory processes [23, 53, 54]. Similarly, serum WBC counts showed no significant differences between the groups and demonstrated low diagnostic utility for both (HG) PJI and (LG) PJI, matching observations from other studies [23, 55, 56].

In contrast to the limited discriminatory power of systemic markers, the immunohistochemical detection of central complement components C3 and C5 provided a clear indication of infection. A key finding of this study is that C3 and C5 expression in tissue occurs regardless of clinical severity or pathogen virulence. Both (HG) PJI and (LG) PJI showed a significantly higher presence of these factors compared to the AL cohort. High AUC values in the ROC analysis underscore the potential of these proteins to indicate the presence of pathogens directly at the tissue level. Specifically, C3 demonstrated high diagnostic performance for identifying (LG) PJI versus AL (AUC: 0.91), while remaining a strong marker for (HG) PJI (AUC: 0.80). Similarly, factor C5 exhibited robust diagnostic performance across both infection severities, achieving AUC values of 0.87 for (HG) PJI and 0.85 for (LG) PJI. This is consistent with investigations by Fröschen et al. [57], who demonstrated the high diagnostic utility of proximal fragments (C1q, C3b) in synovial fluid. Our results mirror these observations at the tissue level and suggest that the local complement cascade in PJI is initiated up to the central nodes C3 and C5, making them suitable candidates for identifying clinically subtle infections. While still speculative and requiring extensive technical and clinical validation, the immunohistochemical detection of tissue-based C3 and C5 might hold interesting prospective potential for future diagnostic algorithms, such as rapid frozen-section analysis or preoperative tissue biopsies. Similar to rapid diagnostic staining in oncological surgery, this approach could eventually serve as an adjunctive tool in ambiguous PJI cases where conventional standard criteria yield inconclusive results.

Analysis of the terminal factor C9 revealed significantly lower diagnostic power. While (HG) PJI reached a moderate AUC of 0.72, the discrimination for (LG) PJI was markedly less effective (AUC: 0.62). A significant increase in C9 expression was detectable only in (HG) PJI patients, while (LG) PJI samples showed no statistical difference from the aseptic group. This suggests that in chronic infections, the assembly of the MAC is either less efficient or inhibited by bacterial protective strategies, such as the biofilm matrix [9, 58, 59]. Furthermore, a relevant C9 presence in the aseptic cohort can be attributed to material-induced complement activation. As demonstrated by Meinshausen et al. [13], abrasion particles from cobalt-chromium alloys can act as foreign surfaces that trigger MAC formation. Therefore, quantifying C9 alone is insufficient to distinguish bacterial infection from particle-induced inflammation. The higher specificity of C3 and C5 compared to C9 suggests that proximal markers are less affected by metal-induced background activation, likely due to more efficient local regulation of these early cascade points in aseptic cases [6, 57, 60].

The microscopic localization of these factors offers functional explanations for their diagnostic performance. The presence of C3 primarily in neutrophilic granulocytes reflects its physiological role as an opsonin, where C3 fragments label pathogens for recognition by phagocytes [5, 6]. Conversely, C5 was predominantly localized vascularly and perivascularly, similar to observations in acute anterior cruciate ligament injuries [61]. Because the cleavage product C5a acts as a potent anaphylatoxin that regulates vascular permeability and induces chemotaxis, this localization corresponds to the maintenance of the local inflammatory response and the recruitment of immune cells [7, 62]. Interestingly, C9 showed perinuclear accumulation rather than a purely surface-bound signature, supporting the hypothesis of rapid MAC endocytosis as a cellular defence against osmotic lysis [63, 64]. In diagnostics, this internalization causes the C9 signal to reflect chronic accumulation rather than acute infection, explaining its lack of specificity in aseptic loosening.

A major limitation of this study is the moderate cohort size of 55 patients, with a particularly small (LG) PJI subgroup. This limits the statistical robustness of our subgroup analyses and results in wider confidence intervals around our ROC estimates Furthermore, the absence of an independent validation cohort limits the generalizability of our findings. However, as an exploratory pilot study, these initial data establish a useful baseline for future, larger-scale validation studies. Additionally, since histological and microbiological findings were utilized for the initial clinical group assignment and are biologically linked to the local immune response, the potential for incorporation bias must be acknowledged. Regarding our tissue analysis, sampling was not standardized, which may account for some variability in the measurements. Moreover, we acknowledge that tissue processing, image acquisition, and field selection were conducted in a non-blinded manner by a single investigator, representing a potential source of observer bias and limiting formal inter- and intraobserver variability assessments. To minimize this risk, we used a standardized computer-assisted, quantification workflow with fixed thresholds, ensuring highly objective measurements. Another methodological limitation is that the optimal diagnostic thresholds (cut-offs) were generated and evaluated within the same patient dataset utilizing the Youden index. Because internal validation techniques or an independent external validation cohort were not utilized in this pilot study, the reported sensitivities and specificities for C3 and C5 may be overly optimistic. Therefore, these diagnostic values should be considered preliminary and require validation in independent cohorts and broader clinical settings before any clinical translation can be considered. Consequently, these results should be interpreted with caution and must be viewed as exploratory and hypothesis-generating. In addition, our study did not account for potential confounding factors, such as relevant comorbidities, antibiotic treatment before revision, or the duration of implant retention. While univariate baseline comparisons showed no significant differences in demographic characteristics between our cohorts, the lack of multivariable statistical adjustments due to the limited sample size remains a limitation. Moreover, since our septic cohort consisted mainly of Gram-positive pathogens, which mirrors the clinical epidemiology of PJI, it remains to be evaluated whether these findings are transferable to Gram-negative, polymicrobial, or fungal infections. Finally, patients with RA were excluded due to the known involvement of the complement system in RA pathophysiology. Because the impact of pre-existing RA on local complement activation or potential cross-reactivity remains unclear, no definitive conclusions can be drawn regarding the diagnostic value of these markers in patients with rheumatic diseases.

Conclusion

In conclusion, our study identifies the central complement factors C3 and C5 as promising tissue-based biomarkers for differentiating periprosthetic joint infection from aseptic implant failure. Unlike systemic inflammatory markers or the terminal complement factor C9, C3 and C5 demonstrate high diagnostic accuracy across all infectious phenotypes, including indolent LG PJI. Ultimately, these exploratory findings provide a valuable scientific foundation for future research to evaluate whether tissue-based C3 and C5 detection can eventually support clinical decision-making, particularly in challenging LG PJI cases where conventional diagnostic criteria remain inconclusive.

Abbreviations

AL

aseptic loosening

BSA

bovine serum albumin

CN-PJI

culture-negative PJIs

CoNS

coagulase-negative staphylococci

CRP

C-reactive protein

(HG) PJI

High grade periprosthetic joint infection

HPF

High-Power-Fields

ICM

International Consensus Meeting

(LG) PJI

low-grade periprosthetic joint infection

MAC

membrane attack complex

MSIS

Musculoskeletal Infection Society

PJI

periprosthetic joint infection

RA

rheumatoid arthritis

ROC

Receiver operating characteristic

THA

total hip arthroplasty

TKA

total knee arthroplasty

WAIOT

World Association against Infection in Orthopaedics and Trauma

WBC

white blood cell count

Appendix Statistical Power Analysis

A post-hoc power analysis was performed using G*Power (v3.1.9.7) to evaluate the statistical strength of the study. For the primary outcomes, the observed effect sizes resulted in high statistical power for C3 (1 - β = 0.98). For complement factor C5, the achieved power was 0.73 (effect size f = 0.40), which, considering the high level of significance (P < 0.0001), indicates a robust diagnostic signal at the tissue level.

In contrast, the terminal factor C9 reached a lower power of 0.49 (effect size f = 0.31), reflecting its limited diagnostic discrimination due to the observed overlap between the septic and aseptic cohorts. Regarding systemic markers, C-reactive protein (CRP) demonstrated a high power of 0.99 (effect size f = 0.78), primarily driven by the significant elevation in the high-grade PJI group. However, the white blood cell count (WBC) showed a notably low power of 0.19 (effect size f = 0.18), confirming its lack of diagnostic utility across the clinical groups. These results suggest that while the study was sufficiently powered to detect strong local and specific systemic inflammatory signals, terminal complement activation and routine serum parameters provided lower diagnostic reliability in this exploratory cohort.

Table A1.

Diagnostic performance and statistical metrics of systemic markers and local complement factors for the differentiation of periprosthetic joint infection (PJI) and aseptic loosening.

Marker	AUC (95% CI)	p-value	Optimal Cut-off *	Sensitivity (95% CI)	Specificity (95% CI)
C3 HG-PJI	0.80 (0.66–0.94)	0.0011	>0.0001624	65.00% (43.29%–81.88%)	85.71% (65.36%–95.02%)
C3 LG-PJI	0.91 (0.79–1.00)	<0.0001	>0.0004148	71.43% (45.35%–88.28%)	100% (84.54%–100%)
C5 HG-PJI	0.87 (0.76–0.99)	<0.0001	>0.0000063	80.00% (58.40%–91.93%)	90.48% (71.09–98.31%)
C5 LG-PJI	0.85 (0.71–1.00)	0.0005	>0.0000017	78.57% (52.41%–92.43%)	85.71% (65.36%–95.02%)
C9 HG-PJI	0.72 (0.56–0.88)	0.0164	>0.0004828	95.00% (76.39%–99.74%)	47.62% (28.34%–67.63%)
C9 LG-PJI	0.62 (0.43–0.82)	0.2254	>0.009141	35.71% (16.34%–61.24%)	90.48% (71.09%–98.31%)
WBC HG-PJI	0.64 (0.46–0.82)	0.1271	>6.25	85.00% (63.96%–94.76%)	52.38% (32.37–71.66%)
WBC LG-PJI	0.65 (0.47–0.84)	0.1297	>6.26	85.71% (60.06%–97.46%)	52.38% (32.37%–71.66%)
CRP HG-PJI	0.90 (0.81–1.00)	<0.0001	>4.65	95.00% (76.39%–99.74%)	80.95% (60.00%–92.33%)
CRP LG-PJI	0.54 (0.34–0.74)	0.6862	>2.21	64.29% (38.76%–83.66%)	57.14% (36.55%–75.53%)

*Fluorescence area normalized to the total number of cell nuclei. AUC: Area Under the Curve; CI: Confidence Interval; WBC: White blood cell count; CRP: C-reactive protein.

Funding Statement

Funding sources: Jasmin Damm received a doctoral scholarship from the Faculty of Medicine of the Ott von Guericke University Magdeburg.