This study sought to summarise the main shape variations of the proximal femur. The rationale behind building four separate models for each patient cohort (male and female) was to illustrate the real-life variability that orthopaedic surgeons face intraoperatively, enabling direct comparison between native and prosthetic femoral version (NFV vs PFV).

The longevity of the femoral component in THA is strongly dependent upon the alignment of the implant [9, 15]. When considering uncemented femoral stems, the orthopaedic surgeon has limited rotational control over the PFV, as these implants achieve press-fit fixation and are guided into a best-fitting position determined by the internal cavity [16]. This, combined with the inability to accurately predict the final PFV, can lead to problems such as excessive anteversion or retroversion of the femoral stem [16] in cases where cementless fixation is chosen for patients with excessive and insufficient intramedullary version.

Post-operative hip joint instability has been associated with component malpositioning [17]. A stem with correct anteversion (between 10 and 20°) [18, 19] maintains the femoral head within the acetabular cup. If excessively anteverted or retroverted, there is a greater risk of anterior or posterior impingement and dislocation, respectively [3, 19]. For the patient, this has implications of pain and restricted range of motion [18, 19]. Implant longevity is also affected by improper PFV as the physiological load distribution and forces transmitted to the implant and surrounding bone may be uneven, potentially leading to edge loading and accelerated wear. Muscle function and efficiency may also be compromised.

Cross-model analysis of the SSM results demonstrates the lack of correlation between the internal and external anatomy of the femur. This more thoroughly defines the problem as the differences in the shape of the internal femoral canal and external anatomy of the native femur can now be visualised using statistical analysis. To the authors’ best knowledge, this is the first study to illustrate these modes of variation. The differences between the NFV and PFV are highlighted in the results from model 1 (native proximal femur) and model 4 (reconstructed femur), with the PFV exhibiting a greater range than the former. This further questions the use of the NFV to guide the PFV. Therefore, inter-model differences exist, reinforcing the need to consider the shape and torsion of the internal femoral canal during 3D surgical planning.

Although it is acknowledged in the literature that the internal femoral canal dictates the final position of a cementless femoral stem [16], there is little appreciation and study of the shape of the canal and how it can be used to better predict the PFV. Though there are similarities in the morphometric features that are generated by each of the four models, the patterns of variation differ and there are clear anatomical differences between the external and internal shapes. This further emphasises the need to discard the sole use of the NFV to inform PFV. If a direct correlation between the models did exist, the preoperative prediction of PFV would not pose a challenge, as the patterns of variation in NFV would directly reflect in the PFV. However, the discrepancies between PFV and NFV reported in this study and other authors such as Hirata et al. [2] suggest otherwise.

When comparing the model of the native proximal femur with the model of the internal femoral canal after neck osteotomy, there are differences in the patterns of variation as well as in the range of the NFV and intramedullary version. In the male internal femoral canal model, intramedullary version is not identified as a key variable, whereas NFV is the most prominently changing feature in the first model for the male patient cohort. These differences emphasize the dissimilarities in the internal and external geometry of the proximal femur, particularly in terms of femoral version. When considering other features, such as the varus/valgus orientation, there is minimal variation across the models.

Another goal of this study was to study sex differences in the femoral morphology. This analysis revealed that in terms of the varus/valgus alignment of the femur, the female patient cohort showed a range from − 2° varus to 16° valgus (with the mean shape being 7° valgus). On the other hand, the male patient cohort exhibited a range from 0°, in other words neutral alignment, to 12° valgus (mean shape was 7° valgus). Therefore, the average native femoral anatomy for both sexes was 7° valgus, relative to the vertical line. The anatomical axis of the femur is known to be around 5°–7° valgus relative to the mechanical axis, so our results are in line with existing evidence [20].

The range of both the NFV and PFV measured from the SSMs was greater for the female patient cohort (NFV from 9° retroverted to 38° anteverted, PFV from 12° retroverted to 44° anteverted) compared to the male patient cohort (NFV from 13° retroverted to 28 degrees anteverted, PFV from 8° retroverted to 43° anteverted). For the mean femoral anatomy, the NFV was also measured to be greater in the female patient cohort (15°) compared with the males (7°), which is supported by several studies. Lerch et al. [21] reported a mean femoral version of 15° for male patients and 22° for female patients in their study looking at the prevalence of femoral version abnormalities in symptomatic hips attributed to Femoroacetabular Impingement (FAI) or hip dysplasia. Although their patient cohorts are different, similar patterns in the NFV were found, suggesting a tendency for the femur to be more anteverted in females compared with males. Additionally, Chadayammuri et al. [22] reported a significantly (p < 0.001) greater mean femoral torsion in their female patient cohort compared with the males (17° versus 9°).

Another key finding from this study was that intramedullary version was not prominently changing within the first three modes for the male cohort but is evidently changing in the first PC in the female model. This is suggestive of a greater variability in the axial rotation of the intramedullary cavity in females than males; this patient cohort may require additional considerations when choosing the fixation method (cemented versus uncemented) as there is a greater risk of the femoral stem being seated in a position not dictated by the external geometry.

The sex differences identified through these SSMs could potentially better inform surgical planning, technique and prosthetic design.

To overcome the challenge of limited rotational control in THA, several strategies have been introduced to assist orthopaedic surgeons in achieving the desired PFV. Modular neck femoral stems offer intraoperative flexibility to adjust leg length, femoral offset, and stem version [6, 23, 24]. However, concerns about modular junction failure, fretting, and corrosion ultimately led to their recall [23,24,25]. Short stems are now increasingly favoured in cementless primary THA due to their focus on metaphyseal fixation rather than diaphyseal engagement, allowing for some degree of proximal version adjustment [26, 27]. Nonetheless, their versatility is limited [28, 29], and the ability to fine-tune PFV heavily depends on the stem’s design. Rectangular cross sections, common in these implants, tend to occupy the proximal canal fully, reducing adjustability. Alternative approaches include modified broaching techniques or opting for cemented fixation. While these methods aim to enhance the surgeon’s control over femoral component orientation, they do not directly address the unmet need for better planning and prediction of stem version.

The authors of this study acknowledge limitations, the main limitation being the limited sample size. Firstly, the 62 patients are a relatively small patient cohort for building SSMs, in order to capture broad anatomical variability. Therefore, the results should be viewed and interpreted with caution and should not be generalised to larger, more diverse populations (e.g. patients with other hip abnormalities). However, we ensured the inclusion of a balanced patient cohort between the sexes, to avoid any skewness or bias of the data. Secondly, in the models where the femoral head was not included (models 2 and 3), version had to be approximated using surrogate landmarks on the osteotomy plane. This may introduce inaccuracies, which we aimed to minimise by systematically picking the most medial and lateral point on the face of the neck post-osteotomy to define its orientation relative to the posterior condylar axis.