The major finding of this study was that the implantation of a rectangular femoral stem with an Al2O3-reduced surface in THA led to unexpected catastrophic failure rates, almost exclusively due to early aseptic loosening. Given these results, it is our ethical duty to report these adverse events to avoid any further potential harm to patients until more conclusive findings concerning this surface treatment are obtained, at least for use in THA.

Analyzing the results of the study, different potential reasons for early loosening must be considered. Long-term stability of the Zweymüller stem relies on its distal anchoring in the diaphysis of the femur [18]. Thus, a distal cortical hypertrophy is frequently seen in uncemented stems and is not thought to affect the clinical outcome or stem stability [19,20,21]. In the present study, more cases of cortical hypertrophy were noted at 2 years in the NT group than in the STD group, especially in the diaphysis around the stem (zones 2, 3, 5, 6). This can be interpreted as a response of the cortical bone to the unsatisfactory osseointegration eventually leading to instability and abnormal load transmission from the stem to the bone [19,20,21]. Patients with NT stems developed slightly less but more severe heterotopic ossification at 2 years follow-up; however, this finding is not believed to be of clinical relevance.

Almost half of patients with an STD stem had RLs present in metaphyseal zones 1 and 7 at 12 months postoperatively, but without progression at 24 months. These data build on the study by Zweymüller et al. [22], who analyzed the extent of RLs in 95 patients with a Zweymüller Alloclassic cementless stem 6 years after surgery. The authors found RLs in 45% of the patients almost all RLs in Gruen zones 1 and 7 with no progression in number and thickness at 10 years. Based on both the observation of Zweymüller et al. [22] and the present study, this suggest that RLs are a frequent finding and are thought to appear early (< 1 year) after THA with a STD stem. Furthermore, if their presence remains unchanged, the stability of the implant should not be jeopardized [22]. In contrast, the NT group of the present study showed about twice as many patients with RLs in zones 1 and 7 with, in addition, a progression tendency to other zones at 24 months. This finding hints at the numerous cases of aseptic loosening seen in the NT group, which can only be attributed to the different surface treatment of these stems.

Stress shielding is an important factor of bone loss in the femoral meta- and diaphysis after uncemented THA [23]. Typically, it takes place during the first 6–12 months, mainly in the Gruen zones 1 and 7 [24,25,26]. Indeed, the BMD values also stabilized after 6 months in the STD group of the present study. This was, however, not the case for the proximal medial meta- and diaphysis (zones 6 and 7) of the femur in the NT group. The surface treatment is not believed to alter the stiffness of the stem, thus excluding stress-shielding differences as the sole reason of failure. A reduced surface roughness of the NT stem could be attributed to the new surface treatment, where lower profile roughness might impair osseointegration and contribute to the worse outcomes in the present study.

Veldstra et al. [27] conducted a double-blind, randomized controlled study with acetabular cups applying the same surface treatment (acid etching, dry ice blasting). No significant differences in BMD and clinical or radiographic outcomes between the standard and alumina-reduced group were noted at 1 year. Thus, the study was unfortunately prematurely interrupted. However, a longer follow-up would have been of great interest, as in the present study, the median time to revision of NT stems was 18 months (range, 14–50 months).

Indeed, the process of grit blasting followed by acid etching, with HF or other types of acids, is still commonly used as a surface treatment for dental implants to promote osseointegration [28, 29]. A study by Li et al. [30], analyzed the effect of grit-blasted titanium tooth implants treated with HF during acid etching in osteoporotic rats and found a higher surface roughness and better osseointegration than in only grit-blasted implants.

On the other hand, histological results of titanium dental implants in rabbits did not provide evidence that residual alumina particles on the implant surface would affect the osseointegration [14]. Comparably, no significant differences in osseointegration or removal torque force to release the implant were found between grit-blasted implants treated with or without acid-etching 10–12 weeks after implantation in rabbits [31]. Similar results were found in more recent studies, although other surface treatments seem even more promising [32,33,34,35,36].

In the same way that alumina particle remnants from the corundum grit-blasting process could interfere with the osseointegration process of the stem, we wonder whether potential residual traces of HF used for the short-acid etching process could contribute to the poor outcome of these NT stems. In contrast, studies demonstrated that fluoride ions on titanium surfaces resulting from HF treatment have the capacity to increase osteoblast differentiation and gene expression [30].

It is nonetheless important to note that in many in vivo rabbit or rat studies involving dental implants, the follow-up time was between 4 and 12 weeks and the implant had a screw shape [14, 30, 31]. These studies cannot reliably represent the in situ and loading conditions, such as those supported by a hip stem over approximately 20 years. Moreover, the optimal surface roughness in dental implants is lower (1–3 μm) than for hip stems [28, 37,38,39]. Beside HF, other types of acid were used for surface treatment and the implants were either pure titanium or a titanium alloy [14, 29].

The primary limitation of the present study is the short follow-up time of 24 months for the BMD measurements, as well as the radiological and outcome data. Although the observation period to calculate survival rates was extended up to 50 months, it remains unknown whether further stems failed thereafter. Surgeries were performed more than 10 years ago, and many patients were lost to follow-up, and it is not possible to know how many stems failed thereafter. Secondary, even at the last follow-up of 24 months, about 35% of patients with the NT stem were not included any more. Thirdly, the implant position or the femoral anteversion, which could also influence the outcome after THA, were not assessed postoperatively.