1 Introduction

Total knee replacement (TKR) is one of the most common orthopaedic procedures in the UK with 134,652 primary procedures performed in 2024 alone.¹ Increasing demand is driven by an ageing population whilst technological advances have increased expectations of good functional outcomes. Optimal implant sizing is a critical determinant of postoperative pain and patient satisfaction with femoral overhang, particularly when exceeding 3 mm, associated with focal soft-tissue irritation of the iliotibial band, lateral retinaculum and collateral ligaments, leading to a significantly increased risk of post-operative knee pain.² Conversely, mediolateral underhang reduces cortical support at the condylar edges, concentrating load through a narrower footprint and potentially increasing local contact stresses and micromotion at the bone-cement or bone-implant interface. Marked underhang may therefore serve as an indicator of global sizing or rotational compromise undertaken to avoid notching or to achieve gap balance.³

The UK national GIRFT programme advocates for implant standardisation within hospitals, citing benefits including surgical familiarity, reduced inventory complexity and potential cost savings.⁴ At our institution, the Attune (DePuy Synthes, Massachusetts, USA) system is currently used as the predominant TKR implant. However, it remains unclear whether a single implant design provides optimal femoral fit for our local population, particularly given known sex-based and ethnic variation in distal femoral morphology.⁵

Computed tomography (CT) offers a reliable method for assessing distal femoral anatomy and simulating implant positioning. By combining CT-derived ML and AP measurements with implant sizing data, it is possible to model theoretical best-fit implant selection and quantify the amount of over- or under-hang generated by different implant systems.

In this study we therefore aimed to use CT-based measurements to 1) simulate distal femoral resections in a representative local population, 2) determine best-fit femoral component sizing for five commonly used UK TKR systems, 3) compare mediolateral mismatch between implant systems and 4) assess whether the currently used implant provides optimal ML fit for our population.

2 Methods

A retrospective observational study was performed on the most recent consecutive CT knee scans performed at a single Southeast London NHS hospital. A priori power calculation conducted before data collection determined a required sample size of 180 knees. The cohort comprised 90 male and 90 female patients. CT scans were performed for clinical indications (e.g. knee injury, osteoarthritis) unrelated to this study and reviewed retrospectively.

Scans were included if they demonstrated a native, adult knee with image quality sufficient for accurate distal femoral measurement. Studies were excluded if the knee had undergone prior arthroplasty (e.g. TKR, UKR, PFJR), sustained bony injury to the distal femur, or exhibited severe deformity that prevented standard measurement.

All CT scans were reviewed using standard multiplanar reconstruction within local imaging software (Sectra PACS, version 26.11.7701). A simulated distal femur resection was established perpendicular to the mechanical axis, 9 mm proximal to the joint line, corresponding to the thickness of the femoral component. Axial reconstructions in this plane allowed measurement of the mediolateral (ML) width as the distance between the medial and lateral epicondyles. The anteroposterior (AP) dimension of the femur was measured as the maximum distance on sagittal reconstructions, perpendicular to the ML axis of the femur. All measurements were recorded in millimetres. Where bilateral CT scans were available, each knee was treated as an independent anatomical specimen.

According to data from the National Joint Registry, the five most frequently utilised femoral component systems in the UK are DePuy Synthes Attune, DePuy Synthes Sigma, Zimmer Biomet NexGen, Smith and Nephew Genesis II, and Stryker Triathlon.¹ Manufacturer-supplied sizing data for femoral components, including medio-lateral (ML) and anteroposterior (AP) dimensions for each size, were analysed for this ^6,7study^.

For each knee, a Euclidian distance model was applied using ML and AP measurements to identify the best-fitting femoral component size for each implant system. The best-fit size was defined as the implant size minimising the Euclidean distance (D) between native femoral dimensions and implant dimensions in the ML-AP space (Fig. 1).

Close

Fig. 1 Euclidean distance model.

Export to PPT

For each implant system, true mediolateral difference (MLD) was calculated as the difference between the mediolateral dimension of the chosen implant less the mediolateral dimension of the patient's femur (Fig. 2):

Close

Fig. 2 Mediolateral difference calculation.

Export to PPT

Positive values represented ML over-hang; negative values represented under-hang. Absolute mediolateral difference (|MLD|) was calculated as the absolute value of MLD and represented the magnitude of mismatch independent of direction.

Statistical analysis was performed using R (R Foundation for Statistical Computing, Version 4.5.2 31/10/2025, Vienna, Austria). Continuous variables were summarised using means and standard deviations. Normality was assessed visually and using distributional characteristics.

Overall differences in true MLD between implant systems were assessed using a paired Friedman test, accounting for repeated measures within each knee. The same approach was used to compare absolute MLD between implant systems. Pairwise post-hoc comparisons were performed using paired Wilcoxon signed-rank tests with Holm-Bonferroni adjustments for multiple comparisons.

The proportion of knees exceeding clinically relevant mismatch threshold of >2 mm and >3 mm absolute MLD was calculated for each implant system.

Sex-specific analyses were performed to compare male and female knees. Differences in absolute MLD between sexes were assessed using the Wilcoxon rank-sum test, while differences in true MLD were assessed using the same approach. Implant performance stratified by sex was summarised descriptively.

The project was registered and reviewed by the local clinical audit review panel and all data analysed was screened and anonymised by the local business intelligence department (project number 8025).

Statistical significance was defined as p < 0.05.

3 Results

A total of 180 knees were included in the analysis, comprising 90 male and 90 female knees. The overall mean age of the cohort was 61.0 ± 19.8 years. Male patients were significantly younger than female patients, with a mean age of 54.1 ± 19.7 years compares with 68.0 ± 17.4 years, respectively (p < 0.001).

Male knees demonstrated significantly larger distal femoral dimensions than female knees. Mean mediolateral (ML) width was 75.2 ± 4.0 mm in males compared with 66.0 ± 4.1 mm in females (p < 0.001). Similarly, mean anteroposterior (AP) dimension was greater in males at 68.8 ± 4.2 mm compared with females at 62.0 ± 3.8 mm (p < 0.001). Males exhibited a significantly higher ML/AP ratio than females (1.09 ± 0.05 compared with 1.07 ± 0.07, p = 0.003), indicating sex-related differences in distal femoral morphology (Table 1).

Across all implants, there was a statistically significant difference in mean true mediolateral difference (MLD) between implant systems when analysed using a paired Friedman test (P < 0.001). However, when comparing the absolute magnitude of mediolateral mismatch (|MLD|), no statistically significant difference was observed between implant systems (p = 0.5284) (Table 2).

The direction of mediolateral mismatch varied by implant design. Triathlon, Genesis and NexGen demonstrated a tendency towards mediolateral over-hang, with positive mean MLD values, whereas Attune and Sigma demonstrated a tendency towards mediolateral under-hang, with negative mean MLD values. Despite these directional differences, the mean absolute mismatch was comparable across implants, ranging from 1.78 ± 1.36 mm (Attune) to 2.08 ± 1.64 (Sigma).

The proportion of knees exceeding commonly cited thresholds for clinically relevant mediolateral mismatch differed between implant systems (Table 3), with 2 mm often quoted in comparable morphological studies⁸ whilst 3 mm is utilised as a clinically relevant threshold in real life models.²

The percentage of knees with an absolute mismatch greater than 2 mm ranged from 38.3% (Triathlon) to 45.6% (Genesis). When the more relaxed threshold of 3 mm was applied, the proportion ranged from 17.2% (Triathlon) to 23.9% (Sigma).

These findings demonstrate that a substantial proportion of knees exceeded both morphological and clinically relevant mismatch thresholds irrespective of implant design, with variability in both magnitude and distribution across systems.

When stratified by sex, no significant difference was observed in absolute mediolateral mismatch between male and female knees (p = 0.6474) (Table 4). Mean absolute MLD was 1.97 ± 1.57 mm in males and 1.94 ± 1.37 mm in females.

In contrast, true MLD differed significantly by sex. Across all systems, male knees demonstrated a tendency towards mediolateral under-hang, with a mean MLD of −0.777 ± 2.40 mm, whereas female knees demonstrated a tendency towards mediolateral over-hang, with a mean MLD of 0.802 ± 2.23 mm (p < 0.005). Under-hang occurred in 57.8% of male knees compared with 32.7% of female knees, indicating a sex-related directional bias in implant fit.

When implant performance was stratified by sex, differences were observed in both the magnitude and direction of mediolateral mismatch (Table 5).

In male knees, all implant systems demonstrated a tendency towards under-hang, most pronounced with the Sigma system (−2.13 ± 2.44 mm). Mean absolute mismatch in males ranged from 1.70 ± 1.25 mm (Genesis) to 2.59 ± 1.94 mm (Sigma). In female knees, most implant systems demonstrated a tendency towards overhang, particularly Genesis (1.80 ± 2.17 mm) and NexGen (1.43 ± 2.13 mm). Mean absolute mismatch in females ranged from 1.57 ± 1.05 mm (Sigma) to 2.35 ± 1.55 mm (Genesis).

These findings indicate that although overall absolute mismatch was similar between sexes, the direction and relative performance of individual implant systems differed substantially between male and female knees.

When a multi-implant model was considered, availability of two implant systems substantially reduced mismatch across all pair combinations (Table 6). The best-performing pair for our dataset (Attune and NexGen) reduced mean absolute mismatch to 1.06 mm and decreased the proportion of knees exceeding 3 mm mismatch to 3.9%. Compared with Attune alone, this represented an absolute risk reduction (ARR) of 17.2% (95% CI 10.6% - 21.1%). Mean mismatch was reduced by 0.73 mm (95% CI 0.57 – 0.88).

Availability of three-implants further reduced mismatch dispersion. The best performing triple implant combination (Attune, Genesis and NexGen) achieved a mean absolute mismatch of 0.83 mm and reduced the proportion of knees exceeding 3 mm mismatch to 3.3%, corresponding to an ARR of 17.8% (95% CI 12.2% – 23.3%) relative to Attune alone. Mean mismatch reduction under this strategy was 0.95 mm (95% CI 0.78 – 1.13).

4 Discussion

This CT-based simulation study demonstrates that femoral component design significantly influences mediolateral fit within a South East London population. While overall absolute mismatch was similar across commonly used total knee replacement (TKR) systems, the direction of mismatch differed significantly between implants, with some systems tending towards mediolateral overhang and others towards underhang. These findings suggest that implant design influences how mismatch occurs, even when the overall magnitude of mismatch remains comparable.

The observed differences between implant systems are likely attributable to variation in femoral component aspect ratios, mediolateral scaling and the availability of narrow or alternative sizing options. Although mean absolute mismatch did not differ significantly between system, a substantial proportion of knees exceeded morphological and clinically relevant thresholds of 2 mm and 3 mm across all implants. This highlights that mismatch remains common regardless of implant choice and reinforces the importance of understanding both the magnitude and direction of femoral sizing error. Notably, even implant systems offering “narrow” sizing options (e.g. Attune and Sigma) did not clearly mitigate these directional trends within our population, suggesting that the availability of narrow sizes alone does not guarantee improved mediolateral conformity.

Sex-specific analysis revealed important and clinically relevant findings. While male and female knees demonstrated similar absolute mismatch, the direction of mismatch differed significantly by sex, with male knees tending towards underhang and female knees towards overhang. This is consistent with known sex-related differences in distal femoral morphology, including narrower mediolateral dimensions in female knees for a given anteroposterior size.⁹ These findings reinforce that distal femoral morphology of male and female knees are structurally different in predictable ways and that these differences have practical implications for implant sizing. Importantly, implant performance was not uniform across sexes: relative rankings of implants differed between male and female knees and certain systems demonstrated more pronounced overhang in females, and underhang in males. These findings suggest that a single implant system may not optimise femoral fit equally across sexes.

We know from large cohort studies that metal extending greater than 3 mm beyond the cortical edge (particularly laterally) irritates the ilio-tibial band, lateral retinaculum and collateral ligaments, producing focal tenderness and activity-related pain. One study found femoral overhang of greater than or equal to 3 mm in 40% of men and 68% of women with total knee replacements and that this contributed to an odds ratio of 1.9 for clinically significant knee pain at 2 year follow-up.² Conversely, under-sizing the femoral component in the medio-lateral plane leaves unsupported cancellous bone at the condylar edges, transferring load through a narrower footprint and theoretically increasing local contact stress. Previous simulation work on sizing show that altering the femoral component size by as little as 2 mm significantly changes contact forces and ligament tension, potentially shifting stresses unfavourably elsewhere.³

Current GIRFT guidance promotes implant standardisation within hospitals, citing benefits such as surgical familiarity, reduced inventory complexity and potential cost savings.¹⁰ The findings of this study do not argue against standardisation per se, and the authors acknowledge the broader institutional and patient-related advantages of a standardised approach. Rather, these results emphasise the importance of evidence-based implant selection, particularly when standardisation policies are implemented. Selection of a single implant system without consideration of local population morphology may risk systematic over- or under-sizing in specific patient groups.

From a practical clinical standpoint, our findings suggest that maintaining access to two or three implant systems within a department broadens the available spectrum of femoral aspect ratios and mediolateral scaling options, thereby increasing the probability of achieving acceptable conformity across a heterogeneous population. In our population, modelling a:⁃Two-system availability strategy reduced the proportion of knees exceeding the 3 mm mismatch threshold from 17.8% under an Attune-only model to 4%⁃Three-system availability strategy reduced the proportion of knees exceeding the 3 mm mismatch threshold further to 3%

These findings demonstrate that prudent, routine multi-system stocking could substantially reduce clinically relevant mismatch compared with a single-vendor model and suggests that modest increases in implant diversity may provide a pragmatic route to improving population-level morphological conformity.

As a result, where pre-operative CT scanning is not performed, careful templating on calibrated plain radiographs should be considered, ensuring availability of multiple implant systems within the digital templating software to allow meaningful comparison of mediolateral fit to allow optimal implant selection for any given patient.

Our data does show a degree of variability within our own local population such that even procurement of multiple implant options may not provide a theoretical best fit for any given patient. To truly identify the optimal implant system based on distal femoral morphology for any given patient a CT scan of the knee would be required. This is not common practice at our institution, with the majority of patients undergoing plain film examinations only, however given the rise of CT-guided navigation and robotic arthroplasty, recent reviews have commented on an increased use of pre-operative CT scanning and hence, this may be common-place in years to come.¹¹

CT-based modelling provides a reproducible and objective method for evaluating implant-bone compatibility at a population level. By simulating distal femoral resections and applying a consistent best-fit algorithm, this approach allows institutions to assess whether their chosen implant system aligns with the anatomical characteristics of their local patient population. Such data may be valuable on a systems level when reviewing procurement strategies or evaluating the suitability of a single-vendor model and provide useful information on a patient level to help guide optimal implant system choice based on pre-operative templating.

5 Limitations

Several limitations are evident in this study. It a simulation study and does not account for intraoperative judgement, soft tissue balancing or surgeon-driven compromises in component sizing. Calculated component sizing for a given implant and knee were generated using a Euclidean best-fit model and does not account for the myriad of intra-operative considerations required in real life clinical practice. The CT-based measurement may not perfectly replicate intraoperative resections planes and all measurements were performed by a single-observer. Finally, no clinical outcomes were assessed and thresholds for clinically significant overhang and underhang were derived from previously published literature rather than patient-reported outcomes. Finally, although the cohort reflects a real-world local population, the findings may not be generalisable to other regions with differing demographic characteristics.

6 Conclusion

In a CT-based analysis of 180 knees from a Southeast London population, significant differences were observed in the direction of mediolateral femoral mismatch between commonly used TKR implant systems, despite similar absolute mismatch across designs. Sex-specific differences in mismatch direction and implant performance were identified, suggesting that a single implant system may not optimise fit for all patients, rather, availability of two or, if logistically feasible, three implant systems substantially improves morphological compliance. These findings support consideration of local population morphology when implementing implant standardisation strategies and provide objective data to inform discussions aligned with GIRFT recommendations.

Informed consent

Informed consent was not required for this study due to its nature as a review of anonymised CT images.

Institutional ethical committee approval

The study was authorised by the local quality assurance department with reference number 8025.

Authors contribution

TE: Conceptualization, Methodology, Formal analysis, Investigation, Writing – Original Draft; TPD: Conceptualization, Methodology, Data Curation, Writing – Review & Editing; AT: Conceptualization, Writing – Review & Editing, Supervision.

Consent declaration

Informed consent was not required for this study due to its nature as a review of anonymised CT images.

Ethics declaration

This study was approved by the local Clinical Audit Review Panel on the 13th March 2026 with reference number 8025. The study did not deviate from the proposal.

Funding/sponsorship

The authors have no funding or sponsorship to declare.

Funding statement

No external funding or sponsorship was received for this study.