In a retrospective cohort study analyzing thousands of primary knee replacements, researchers have found that an index measuring the technical and mechanical complexity of the operation is associated with a graded increase in subsequent periprosthetic joint infection. Conducted across 2,920 primary total knee arthroplasties performed in 2,350 patients at Whangārei Hospital in New Zealand, the investigation demonstrated that each additional anatomical or construct challenge present at the initial operation raised the subdistribution hazard of septic revision by approximately 61 percent. By tracking patients over a median follow-up of 7.4 years with national joint registry linkage, the investigators observed that knees requiring two or more complexity markers experienced a roughly four-fold rise in five-year cumulative infection incidence compared to procedures with zero or one marker. This association between procedural complexity and joint infection persisted after adjusting for measured patient-level health variables, indicating that the mechanical and structural demands of the surgical procedure itself represent an important axis of clinical risk alongside traditional patient-centered characteristics.
For decades, predictive modeling for periprosthetic joint infection after total knee arthroplasty has concentrated almost exclusively on the patient rather than the operation. Established statistical tools and risk calculators heavily prioritize host-level attributes, including demographic factors, body mass index, baseline medical comorbidities, and laboratory measures. In clinical practice, infection after joint replacement is frequently attributed to these systemic vulnerabilities, leaving the physical complexity of the reconstructive procedure largely unexamined. While surrogates of surgical burden such as extended operative duration, blood transfusions, and prolonged hospital stays are recognized markers of risk, they represent general reflections of systemic stress rather than specific anatomical or reconstructive hurdles. Furthermore, technological innovations such as robotic guidance and computer navigation have not reduced infection rates in matched cohorts, as they alter how surgical targets are achieved rather than diminishing the underlying magnitude of the reconstructive task.
To investigate whether surgical burden carries infection risk independent of the patient, the research team examined prospective clinical records from procedures performed between 2010 and 2024 linked directly to the New Zealand national joint registry. This complete linkage ensured that any subsequent revision surgeries performed at other regional hospitals across the country were fully captured. The final analytical cohort comprised 2,920 primary total knee arthroplasties in 2,350 individuals with comprehensive operative documentation and long-leg radiographic assessments. Patients had a minimum of two years of potential follow-up, with an overall median observation period of 7.4 years and an interquartile range spanning 4.4 to 10.7 years. Within the study group, the average age was 69.4 years, 52 percent were male, the mean body mass index was 31.7, 29 percent presented with an American Society of Anesthesiologists score of three or higher, and 16 percent identified as New Zealand Māori.
Before performing any statistical analyses, the investigators formulated a pre-specified surgical-complexity index based on five objective construct markers documented during the primary operation. These markers included the use of implant constraint, such as a condylar-constrained or hinged prosthetic design, the insertion of a tibial stem, the placement of a femoral stem, the presence or removal of a previous surgical implant, and the correction of a pre-existing extra-articular deformity or prior fracture. These specific criteria were selected because they are recorded objectively rather than through subjective judgment, are identifiable prior to surgery from standard preoperative radiographs and operative plans, and capture aspects of procedural complexity that are typically absent from standard national joint registries. The index was constructed as an unweighted sum ranging from zero to five, avoiding data-driven weighting that could introduce noise into a cohort with rare binary endpoints.
Methodologically, the study addressed the competing risk of death by employing Fine–Gray subdistribution hazard regression models with patient-level clustering to account for individuals undergoing bilateral procedures. Accounting for competing mortality was essential, as 525 patients—representing 18.0 percent of the cohort—died during the follow-up window without undergoing revision surgery. Over the study timeframe, the team recorded 51 septic revisions, defined as any reoperation for periprosthetic joint infection, and 104 aseptic revisions, defined as reoperations for non-infectious causes. In the statistical models, death without revision was treated as a competing event, while each distinct revision type served as a competing risk for the other. Five-year cumulative incidence curves were constructed using the Aalen–Johansen method, adjusted for patient age, sex, body mass index, and an American Society of Anesthesiologists grade of three or greater.
The primary regression analysis revealed that each additional complexity marker was associated with a statistically significant increase in the subdistribution hazard ratio for periprosthetic joint infection, yielding an adjusted estimate of 1.61 with a 95 percent confidence interval of 1.27 to 2.04. The overall concordance of the multivariable model was 0.68. Among measured patient characteristics, male sex demonstrated a strong independent association with septic failure, showing a subdistribution hazard ratio of 2.83 with a 95 percent confidence interval of 1.49 to 5.39. Higher American Society of Anesthesiologists class and elevated body mass index trended in the direction of increased infection risk but did not attain formal statistical significance across the 51 observed septic events. Across the entire cohort, complexity markers were identified in 525 knees, representing 18.0 percent, while 177 knees, or 6.1 percent, exhibited two or more markers.
A notable finding emerged when examining five-year cumulative infection incidence across complexity strata, demonstrating a clear practical threshold. For knees presenting with zero complexity markers, the five-year cumulative incidence of infection was 1.5 percent, with a 95 percent confidence interval of 1.1 to 2.1 percent. Knees with exactly one marker showed an identical five-year cumulative incidence of 1.5 percent, with a confidence interval of 0.6 to 3.5 percent. However, for knees requiring two or more complexity markers, the five-year cumulative incidence rose sharply to 5.9 percent, with a 95 percent confidence interval of 3.2 to 10.8 percent. Sensitivity analyses evaluating robustness against potential unmeasured confounding yielded an overall E-value of 2.60, with a value of 1.86 for the lower confidence bound, indicating that an unmeasured confounder would need an exceptionally strong association with both surgical complexity and joint infection to nullify the observed relationship.
Because extended wound exposure is an established biological risk factor for surgical contamination, the authors examined operative duration to determine whether it explained the complexity effect. Skin-to-skin surgical times were available through registry records for 2,722 knees, or 93 percent of the analytic cohort, with an overall median duration of 85 minutes. Operative duration increased alongside surgical complexity, rising from a mean of 85 minutes for knees with zero markers to 111 minutes for knees with three markers and 122 minutes for knees with four markers. When operative time was added to the multivariable regression model, the subdistribution hazard ratio for the complexity index shifted only slightly from 1.58 to 1.55, maintaining a 95 percent confidence interval of 1.21 to 1.98. Operative time itself showed no independent association with septic revision, and statistical mediation analysis via the difference method indicated that only five percent of the total complexity effect was transmitted through operative duration.
To rigorously evaluate the specificity of their findings, the researchers incorporated coronal mechanical alignment as a negative-control exposure and aseptic revision as a negative-control outcome. In a subset of 1,594 knees with paired preoperative and postoperative long-leg radiographs, neutral coronal alignment—defined as a postoperative mechanical hip-knee-ankle angle within three degrees of neutral—was strongly protective against aseptic mechanical revision, producing a subdistribution hazard ratio of 0.42. Conversely, neutral alignment exhibited no association with periprosthetic joint infection, yielding a null subdistribution hazard ratio of 0.87. Conversely, the surgical-complexity index showed no association with aseptic mechanical revision, producing a hazard ratio of 1.06. This mutual negative-control framework established a double dissociation, indicating that technical complexity selectively tracks infectious failure, whereas coronal alignment selectively influences mechanical failure rather than reflecting generalized institutional or surgical confounding.
The biological mechanisms connecting construct complexity to periprosthetic infection are supported by established microbiological and orthopedic research. Decades of experimental work have documented that the introduction of foreign prosthetic material drastically reduces the bacterial inoculum required to establish active infection, while compromising the local bactericidal efficiency of granulocytes. Constrained and stemmed knee designs introduce substantially larger prosthetic surface areas into the joint space compared to standard unconstrained implants. Furthermore, knees presenting with post-traumatic deformity or retained internal hardware frequently carry occult colonization; prior sonication studies of osteosynthetic implants removed from asymptomatic patients have cultured bacteria in 27 percent of cases, occurring far more frequently in the knee than the hip. These biological dynamics align with previous large-scale registry studies identifying previous open surgery and retained hardware as major independent predictors of prosthetic joint infection.
The authors emphasized that their surgical-complexity index is designed to complement rather than replace published preoperative infection calculators. Existing predictive tools utilize broader combinations of patient comorbidities, clinical variables, and laboratory markers to achieve higher predictive concordance than the 0.68 achieved by the anatomical index. What the complexity index provides is an independent, structural dimension of risk that standard calculators and national registries overlook entirely. Importantly, because each of the five complexity markers can be determined prior to entering the operating room from standard radiographs and procedural plans, this risk profile is accessible well before surgery, unlike post-hoc metrics such as operative time or transfusion requirements that are known only after an operation is concluded.
The investigators carefully outlined several limitations inherent to their study design, noting that the findings represent observational associations derived from a single surgical center and cannot establish direct causality. Because the analysis was based on 51 confirmed septic revisions, statistical power supported evaluating an aggregated index rather than isolating individual component weights, as fitting rare binary markers produced complete statistical separation that required penalized-likelihood estimation. Furthermore, operative duration was unavailable for seven percent of the cohort, surgeon-specific identities and intended alignment philosophies were not recorded, and residual confounding by surgical indication cannot be entirely excluded. Consequently, the authors conclude that while these findings should not be interpreted as a validated clinical calculator, surgical complexity represents a tangible exposure that clinicians should consider alongside host health when assessing infection risk and planning perioperative management.
Subject of Research: Surgical complexity and periprosthetic joint infection risk after primary total knee arthroplasty
Article Title: A pre-specified surgical-complexity index is associated with a graded increase in periprosthetic joint infection after total knee arthroplasty: a competing-risks analysis using coronal alignment as a negative control
Article References: Hoskins, W., Gusho, C., Rosser, K., Lim, M., Dunstall, K., & Vince, K. G. (2026). A pre-specified surgical-complexity index is associated with a graded increase in periprosthetic joint infection after total knee arthroplasty: a competing-risks analysis using coronal alignment as a negative control. Journal of Bone and Joint Infection, 11(6), 705-713. https://doi.org/10.5194/jbji-11-705-2026
Image Credits: AI Generated
Keywords: total knee arthroplasty, periprosthetic joint infection, surgical complexity, competing risks, coronal alignment, septic revision, aseptic revision, orthopedic surgery, joint registry linkage, Fine-Gray model, tibial stem, femoral stem, constrained implants, operative duration, negative control, knee replacement outcomes
News Source: Kristina Jarvis. (November 11, 2026). Surgical Complexity Index Predicts Increased Knee Replacement Infection Risk. Scienmag.
