Total hip arthroplasty (THA) is the gold standard of the treatment for end-stage hip diseases [[1], [2], [3]]. Conventional cementless hip stem is the first choice to provide satisfactory long-term fixation and pain-free function in patients of all ages, especially for young and active patients. However, conventional stems have some disadvantages, including proximal-distal mismatch, suboptimal load transfer, loss of proximal bone, and thigh pain [4,5]. To address these issues, short stem metaphyseal-engaging implants are designed to enhance the preservation of proximal femoral bone stock and to pursue a more physiological load transfer in the proximal femur [1,2,[4], [5], [6], [7]]. In-vitro tests have confirmed that short stem metaphyseal-engaging implants have the potential to provide a better biomechanical environment [8]. However, the initial stability of the stem fixation is compromised when the length of the stem has been shortened [2,9]. Meanwhile, the shortened stem might lead to malposition of the stem, which has been recognized previously [10,11].

Custom design has the potential to further improve the fit and fill between the stem and proximal femoral medullary canal, which can maximize surface contact, minimize initial micromotion, reduce malposition, avoid corrosion at the interface of a modular stem and optimize the biomechanical load transfer [11]. In-vitro and finite element analyses have proven that custom design could further improve the initial stability of stem, which is also confirmed by clinical follow-up studies [8,12,13]. Chow et al. reported satisfactory mid-term outcomes for a computed tomography-based, custom-made, metaphyseal-engaging short stem femoral implant, and radiographic evaluation showed improved bony ingrowth and preservation of bone stock in the proximal femur [11]. However, there is no consensus regarding the overall system for the manufacture of custom short femoral stem implants.

Traditionally, custom design and manufacture rely on Computer-assisted design and manufacture system, which belongs to the subtractive manufacture system. Computerized numerical control machine system is commonly applied to support the manufacture of custom-made implants, which is related to lower efficiency, longer manufacture cycle and poor quality regarding the manufacture of the free-form surface at the lateral portion of custom stem. The additive manufacturing (AM) technique is capable of manufacturing both the surface coating and stem body. It has the potential to facilitate the manufacture of complicated custom stem implants, with improved manufacture efficacy and shortened manufacture cycle [14]. Recently, attempts have been made to fabricate custom femoral stem with regular length using AM system [[14], [15], [16], [17], [18]].

Understanding the fatigue behavior of metallic implants is important for clarifying their long term effect in the human body. Fatigue properties of additive manufactured titanium alloy products have been studied extensively [[19], [20], [21], [22], [23]]. So far, most attention has been paid on the fatigue behaviour of the additive manufactured titanium alloy femoral stem with a regular length. For examples, Marie et al. fabricated titanium alloy hip stems using the electron beam melting (EBM) technique, and found that the fatigue limits of hip stems are dependent on their surface roughness [14], i.e., the coarser the surface, the lower the fatigue limit. Yang et al. reported that the stress distribution under cyclic deformation in the EBM titanium alloy hip stems is determined by their prototype [15]. Bruno et al. designed and fabricated titanium alloy femoral stems with a range of porosity using the selective laser melting (SLM) technique. They found that the porous titanium alloy stems have lower stiffness than the dense ones. Therefore, the stress shielding effect of the implant can be alleviated by introducing porosity into titanium alloys [16,17]. Croitoru et al. manufactured Ti6Al4V alloy femoral hip stems with fenestrated architecture similar to the natural structure of bones. The fatigue limit of the femoral hip stems is higher than the minimum value specified by ISO standard [18]. However, the geometric parameters of the custom-made short stem are different from those with regular length. The geometry shape may affect the stress distribution and thus the fatigue properties of femoral hip stems. In addition, the influence of microstructure and defects on the fatigue properties of custom additive manufactured femoral stems is not clear. Thus, it is crucial to investigate the fatigue behavior of the custom-made short stems fabricated by the AM technique in order to understand its prototype-microstructure-properties relationship.

The current study was for the first time to apply EBM in the manufacture of the titanium alloy custom short stem, and systematically investigated its fatigue property before and after hot isostatic pressing. The results indicated that the fatigue properties of the studied stems were closely related to the geometric parameters and the underlying mechanism was discussed in details.

The custom short stem was designed based on a series of CT images of a 40-year-old male with a resolution of 512 × 512 pixels and a layer thickness of 1.5 mm. The three-dimensional model of the patient's femur was reconstructed with Medraw software (Image Medraw Technology (Shanghai) Co., Ltd, Shanghai, China). Fig. 1 summarizes the workflow of the custom short stems with 4 different lengths. To maximize the fit and fill of the proximal femoral canal, the most significant anatomical region of proximal femur was chosen for the design of the custom short stem, covering 20 mm above and 10 mm below the middle of the lesser trochanter. Three specific axial planes were defined, including plane P1 at 20 mm above the middle of the lesser trochanter, plane P2 at the middle of the lesser trochanter, and plane P3 at 10 mm below. By manually dragging the control points and dropping on the boundary of effective filling area of the specific planes, one closed boundary curve was generated for each section. A smoothly lofting surface was constructed by connecting boundary curves (Fig. 1(a)). An offset surface with 2 mm distance medially from lofting surface was used to build a complete solid stem model. Finally, four CAD models of custom short stem in STL format were generated as inputs for the manufacture of custom short stems, including the stem body and surface porous coating (Fig. 1(b)).

The Ti-6Al-4 V powders for EBM process were provided by Arcam AB (Sweden). The shape of the powders is nearly spherical and the average particle size is 70 μm (Fig. 2). The chemical compositions obtained by wet chemical and gas analyses for Ti-6Al-4 V powder are shown in Table 1.

Ti-6Al-4 V stems were produced using an Arcam A1 EBM machine. In the EBM process, a voltage of 60 kV, electron beam size of ∼ 200 μm were used and the process was kept under vacuum at 10^-3 mbar, controlled by using helium as regulating gas. The process started with preheating the powder prior to melting. The powder layers were preheated to 730 °C. Following preheating, samples were produced with a speed function of 39, scan spacing of 0.2 mm, and layer thickness of 0.05 mm. In order to know the mechanical properties of the stems, some cylinders with dimension of φ10 mm × 60 mm were fabricated together on the start plate directly and the building direction was parallel to that of short stem.

Some of EBM samples were HIP-treated in a QIH-15 hot isostatic pressing furnace at a temperature of 930 °C with pressure of 130 MPa for 3 h and were then cooled to room temperature in the furnace.

In order to characterize the pores distribution of the EBM sample, micro-CT was conducted on a Zeiss Versa 500 Micro-CT system operating at an accelerating voltage of 160 KV and current of 62.5 μA. The sample was rotated 360°, and 1600 projections (at a step of 0.225°) were acquired using a charge-coupled device detector with 3 s exposure time. The voxel size was 20 μm. The Micro-CT 3D data was analyzed using the Avizo 8.0 software.

Microstructural of the studied samples was examined using optical microscopy (OM, ZEISS-AXIO) and scanning electron microscopy (SEM, JSM-6510A). The specimens for the OM and SEM analysis were mechanically polished and then etched in a solution consisting of 2 vol.% HF, 5 vol.% HNO₃, and 93 vol.% H₂O. The fracture surface of the samples after fatigue failure was observed by SEM (secondary electron mode). Before observation, they were cleaned using solvents in the sequence of acetone, alcohol, and water. The grain size was measured using the linear intercept method, and more than 50 grains are measured for each sample. Phase constitutions were examined on a Brucker D8 Discover 2D X-ray diffractometer (XRD) using a CuKα radiation source with an accelerating voltage of 40 kV and a current of 250 mA.

Tensile samples with the geometry in accord with the ASTM E8 standard were machined from the cylinders that fabricated together with short stems. The uniaxial tensile tests were carried out at a strain rate of 2.5 × 10^-4 s^-1 in air at room temperature using an Instron 8872 machine.

The fatigue test of customized femoral stems was conducted according to the ISO-7206-4-2010. Fig. 3 shows a drawing of the test apparatus and a picture of a stem fixed in a metallic cup (using bone cement) under testing. The cement holds the distal end of the stem starting at 80 mm from the center of the femoral head. The drawing in Fig. 4 illustrates this configuration, being defined by the ventral angle of 10°, alateral-medial angle of 9°. The fatigue tests were performed at a stress ratio R = 0.1, a frequency of 10 Hz and a sinusoidal cyclic waveform loading in dry media and at room temperature. A compressive cyclic load ranging between 230 N and 2300 N is applied on the femoral head component.

The stress and strain distribution in the short stem was analyzed using Abaqus/Standard 2017 (Dassault Systemes Inc., Vélizy-Villacoublay, France). The models of the stem were established according to the ISO7206-4 (Fig. 4). Hexahedral mesh was generated for bone cement model while tetrahedral mesh for short stem model. The modulus and Poisson’s ratio of the Ti-6Al-4 V alloy stem and cement material used for Abaqus/Standard 2017 were 110 GPa and 0.36, 2.3 GPa and 0.3, respectively. All materials were assumed to exhibit homogeneous, isotropic and linear elastic behavior. The boundary conditions were set to simulate the ISO7206-4 loading. The FE models in this study were static model since we were interested in the maximum stress in the stem. A maximum vertical force of 2300 N was applied on the femoral head, and all the nodes of the external surface of the cement were fixed (Fig. 3). The interface between the stem and the cement was considered to be a friction contact with a friction coefficient of 0.3.

Fig. 5 shows the optical micrograph of the lamellar α + β structure of as fabricated EBM samples. Analyses on the SEM images show that the average thickness of α lamellae is about 1.5 μm, and thin β phase was observed between the coarse α lamellae (Fig. 6(a)). As seen in the Fig. 7, the studied EBM Ti-6Al-4V stems mainly consist of α phase with a small amount of β phase.

In the samples after HIP treatment, significantly coarsened α lamellae (∼ 2.5 μm) were observed (Figs. 5(b) and 6(b)). The coarsening of α phase is mainly due to the high HIP temperature (930 °C), which is close to β transus temperature (980 ± 10 °C), and the subsequent slow cooling (furnace cooling).

The morphology of the porous coating on the stem is shown in Fig. 8. It is seen that the porous structure is trabecular like and the average pore size is about ∼ 650 μm. The thickness of the coating layer is about 2 mm. It can also be seen that the surface between the coating layer and the substrate is very rough (Fig. 8).

The CT scan results (Fig. 9) indicate that the pores are the main defect in as-built EBM stems. The pores in the studied stems are almost spherical with an average size of 40 μm. There are two reasons that may cause pore defect in the EBM materials. Residual argon gas may adhere to the powder interiors during powder production by the gas atomization method, which results in blowholes inside the EBM samples [24]. Besides, during EBM processing, powders can be easily gasified producing a large amount of metal vapor [25], which cannot be completely exhausted with the moving molten pool and will remain in the form of pores in the EBM sample [26]. Pore defect in the as-fabricated stems can be significantly eliminated through the hot isostatic pressing (HIP) treatment, as demonstrated in the Fig. 9.

The tensile properties of the studied EBM Ti-6Al-4 V short stems are estimated via testing the accompanying cylinders. The tensile results of the as-fabricated and HIPed samples are summarized in Table 2. It is seen that the as-fabricated samples have higher yield strength (YS) and ultimate tensile strength (UTS) (YS: 951 MPa; UTS: 1037 MPa) but a little lower elongation (∼14.5%), compared to the HIPed samples. The difference in the strength is supposed to be originated from the different thickness of α lamellae in the two conditions. It is well documented that the relationship between the strength σ_ys and the thickness of α lamella (d) for two phases titanium alloy can be estimated from the Hall-Petch equation [27,28]:

where σ₀ is the frictional stress of dislocation motion, Ky is strength coefficient, and d is lamella thickness. As can be seen from the equation, the strength of the alloy decreases when the thickness of α lamella increases, which is exactly the case of the HIPed samples. Meanwhile, coarsened α lamellae have higher capacity in activating and accommodating more types of dislocations during tensile deformation, which can result in better ductility [29].

3.4.1. Fatigue life

The fatigue lives of the studied short stems are summarized in Table 3. It can be seen that the fatigue life of short stems decreases with the increase of stem length, and Model 20 and 40 short stems passed 5 × 10⁶ cycles without failure. HIP treatment can improve the fatigue life of short stems significantly.

Fig. 10 shows the fracture position of the studied short stems. As listed in Table 4, all stems fracture near the cement surface, and the area of cross section of stems at the fracture decreases with the increase of stem length.

3.4.3. Morphology of fracture surface

As shown in Fig.11, the fracture surfaces of the studied stems consist of crack initiation area, stable crack propagation area and fast fracture area. In the crack initiation area, fatigue cracks in stems initiate from the interface between the coating and substrate, rather than the voids in the stem. Due to the complex loading condition, several crack initiation sites were observed in one sample. In the stable crack propagation area, the cracks propagate towards the whole stems, and a large amount of secondary cracks perpendicular to the initial crack propagation direction were observed. In the fast fracture area, a large number of dimples were observed, which indicates a ductile fracture mode of all the studied titanium alloy stems.

The stress distribution of the studied short stems under a constant loading is shown in Fig. 12. The maximum tensile stress was at the surface of the stem body, and the value increases with the stem length. The position of maximum tensile stress was 1-4 mm above the interface between the stem and cement, which was in agreement with the fracture position of the stem in Fig. 10. Therefore the maximum tensile stress in the stem under a pressure may result in its fracture. It should be noted that different from other models, the maximum tensile stress in Model 20 stem concentrated near the interface between the coated and non-coated parts as shown in Fig. 12, which was due to its lowest imbed position in the cement. This indicates that the position of the stress concentration can be adjusted by changing the stem length. Fig. 13 shows the cross section of stems at the position with maximum tensile stress. As shown in the figure, the area of cross section at the maximum tensile stress gradually decreases with the increase of stem length.

For EBM fabricated titanium alloy implant, the pores are usually inevitable [30]. Extensive studies have confirmed that the fatigue cracks can easily originate around these pores, which significantly decreases the fatigue strength of samples [[30], [31], [32], [33], [34]]. In the current work, detailed fractographic studies demonstrated that fatigue cracks always initiated at the surface of the substrate near the coating/substrate interface rather than the pores, although a large amount of pores did exist in the as-fabricated samples. Such discrepancy may be due to the difference in imposed stress levels. It has been reported that for the fatigue life of the samples containing pore defects, the number of cycles to failure for a given stress level can be correlated to the critical stress intensity amplitude, K_cr, near the pores [[35], [36], [37], [38]]. Specimens fail if critical stress intensity amplitude K_cr is reached. Below K_cr, fatigue cracks may initiate at the pores but do not propagate until failure. In this work, at the studied stress level, the stress intensity around pores in the studied stems may not reach the K_cr and the fatigue crack does not propagate. Thus the cracks initiate from the much sharper notches in the rough interface between the coating and matrix. The role of the pores to the stem fatigue life may be related to the load bearing area of stem. It is reported that the pores defects in as-fabricated specimen can cover up to 5% of the fracture surface [34]. The reduction of area may increase the local stresses around the rough surface. HIP treatment closes most of pores contained in the stem and thus increases the loading bearing area [39,40], which may alleviate the local stress near the rough surface and improve the fatigue strength of the studied EBM stems, as evidently shown in the data in Table 3.

Fatigue performance of the bone implant is critical to ensure its long-term use in human body. For metal implant, its high cycle fatigue life is mainly determined by its fatigue initiation life and fatigue propagation life, and the fatigue initiation life is dominant to the fatigue propagation life [41]. In the titanium alloy implant fabricated by electron beam technique, voids are inevitable. They are the sources of fatigue crack initiation sites and are easy to engender cracks at the early stage ^[32-34]. However, as discussed in Section 4.2, the rough surface appears to be the dominant factor in fatigue strength of the studied EBM short stems. According to the results from the fracture surface morphology and finite element analyses of stem, the cracks initiated at the surface of the stem with maximum tensile stress. Moreover, the maximum tensile stress increases and the corresponding area of cross section decreases with the increase of stem length, which means a higher stress concentration at the rough surface in the longer stems. Consequently, the fatigue crack may initiate easier and the fatigue life of stem will decrease with the increase of the stem length. It is noted that for HIPed stem, the model 60 and 80 stem showed similar fatigue life. This means that the fatigue performance of as-fabricated short stem is more sensitive to the concentrated stress on the stem surface caused by different stem length. To sum up, it seems to be an effective way to improve the fatigue life of short stem by tailoring the stem length to weaken the stress concentration, which ensures the confidence in the long term use of femoral stem fabricated by EBM in the human body.

The fatigue property of Ti-6Al-4 V alloy custom short stems with four different lengths fabricated by electron beam melting were investigated according to ISO 7206-4 standard, and the following conclusions are drawn:

(1)The maximum tensile stress concentrated at the lateral side of the stem body. Under the identical applied load, the magnitude of tensile stress increased with the increasing of stem length, while the corresponding area of the axial section decreased.

(2)That fatigue cracks mainly initiated on the rough surface of the short stem where the maximum tensile stress concentrated. Due to the complex stress loading condition, there were several crack initiation sites in the stem.

(3)The fatigue strength decreased when the length of the stem increased, because of the higher concentrated stress on the surface of a longer stem.

(4)Most of the pores were closed after HIP treatment, and the fatigue strength increased since the local stress near the rough surface was alleviated by the increase of load bearing area.

(5)The stress concentration on the surface of the stem can be mitigated via optimizing the stems length, which can potentially improve its fatigue property.

This work was supported partially by the Chinese MoST (No. 2017YFC1104903), the Key Research Program of Frontier Sciences, CAS (No. QYZDJ-SSW-JSC031-02), the National Natural Science Foundation of China (Nos. 81772425, 51631007 and 51871220), the Science and Technology Commission of Shanghai Municipality (No. 16441908700), and the Shanghai Jiao Tong University (No. YG2016MS11).