We introduced a modified polished Charnley stem (Kerboull^® MKI; Benoist Girard, Stryker Howmedica, Herouville Saint-Clair, France) in 1972. The survival rate of this stem was greater than 90% at 20 years using radiographic loosening as the end point [10, 16, 18, 27]. Later, this polished stem became available with a satin and matte surface finish.

The Kerboull prostheses combined a 22.2-mm stainless steel femoral head and an all-polyethylene socket. The monoblock femoral stem in each study group (Fig. 1) was derived from the original Kerboull stem (Kerboull^® MKI; Benoist-Girard, Stryker-Howmedica). Both stem designs included a stem neck angle of 130°, a 10-mm neck diameter, and a double-taper shape with an angle of 5°. The femoral stem was available in two configurations according to the surface finish. The polished stems had a surface roughness of 0.04 μm (1.6 μin) with a quadrangular cross-section (Kerboull^® MKIII; Benoist-Girard, Stryker-Howmedica), whereas the matte stems had a surface roughness of 1.7 μm (66.9 μin) with an oval cross-section (CMK3®; Vecteur Orthopédic, Marne la Vallée, France). The all-polyethylene sockets were sterilized with 3 Mrad of gamma irradiation in air.

All procedures were carried out by two senior surgeons (JPC, MK) through a lateral approach with a trochanteric osteotomy as has been reported previously [17]. Both components were inserted with CMW Type 1 bone cement (DePuy 1, Exeter, Devon, UK). The cement was mixed in air. The cement was hand-packed on the acetabular side. It was packed with a rasp in the femoral canal, which was blocked with a bone plug before cementation. We used no cement gun, pulse lavage, or pressurizer. The femoral component in each study group was cemented according to the French paradox principles [19], including removal of all diaphyseal cancellous bone to obtain a primary stability of the stem before line-to-line cementation.

Postoperatively, patients received anticoagulation therapy, systemic antibiotics, and nonsteroidal antiinflammatory drugs (100 mg ketoprofen per day) to prevent heterotopic ossification. Passive motion exercises of the involved joint began immediately after the operation with the assistance of a therapist and were continued until active motion of the hip was possible. The patients were free to walk with two supports after 3 days. Full weightbearing was usually allowed after 6 weeks.

All patients were evaluated preoperatively, and most of them had been followed at 6 weeks, 3 months, 6 months, 1 year, and then every 2 years. Attempts were made to interview all 395 patients who had composed the prospective and historic series or to contact their families. The surviving patients were asked to return for clinical and radiographic evaluation. Those who were unable to return were asked to have radiographs made locally and then sent to us for evaluation. Of the 246 patients (276 hips) who were still alive and had not had a revision, 224 (250 hips) were evaluated at a minimum of 10 years clinically and with an anteroposterior radiograph of the pelvis made. The remaining 22 patients (26 hips) were unable to return for a clinical evaluation for reasons not related to the results of the arthroplasty, and they were evaluated by telephone interview by independent observers (FB, NL, MH). Members of the families of the patients who had died were interviewed to determine the function of the hip at the time of death. Hip functional results were rated according to the Merle d’Aubigné grading system [24].

Serial anteroposterior radiographs of the pelvis were analyzed by independent observers (FB, NL, MH, MK). Parameters investigated on the femoral side included progression of radiolucent lines according to the seven zones described by Gruen et al. [9], calcar resorption, and subsidence of the stem. Loosening of the stem was defined according to the criteria of Harris and McGann [12], including stem subsidence equal to or greater than 3 mm, cement mantle fracture, complete radiolucent line equal to or greater than 2 mm, or radiolucent line in Zone I equal to or greater than 2 mm. Periprosthetic cystic or scalloped lesions exceeding 2 mm in diameter that had not been noted on the immediate postoperative radiograph were defined as osteolysis.

Twenty stems were radiographically loose after a minimum 10-year followup. The total rate of femoral loosening, including revision after a minimum 10-year followup, was 0.6% (one of 165), 3.3% (four of 123), and 10.3% (15 of 145) for the polished, satin, and matte stem groups, respectively (Fig. 2). The rates differed (p < 0.0001) for the polished versus the matte stems and the satin versus the matte stems (p = 0.03), but not for the polished versus the satin stems (p = 0.17). Except for one isolated femoral satin stem loosening, the loosened stems in the satin and polished groups of patients were associated with high polyethylene wear and periacetabular osteolysis with limited femoral osteolysis (Fig. 3). On the contrary, among the 15 loosened matte stems, high polyethylene wear was observed only in three hips. All hips with matte stem loosening were associated with extensive femoral osteolysis. Moreover, loosening of matte stems occurred at the bone-cement interface as indicated by clearly delineated radiolucent lines at the bone-cement interface (Fig. 4).

With revision of the femoral component for any reason as the end point, the survival rate was greater (p = 0.003) for the polished versus the matte stems, but not greater for the polished versus the satin stems and for the satin versus the matte stems (Table 3). With radiographic femoral loosening including loosened revised stems as the end point, polished stems had greater (p < 0.0001) survival than the matte stems and the satin greater (p = 0.0004) than the matte stems, but the polished and satin stems had similar survival (Table 3). The apparent mechanisms of loosening differed. The intraoperative findings at revision were different among the stem surface groups. In the hips with a polished or satin surface, the cement mantle was continuous, and during removal, the stem was easily separated from the intact cement. In contrast, in the hips with a matte surface, the failure occurred at the bone-cement interface. The cement remained firmly attached to the stem, and the cement-stem composite was usually removed in one piece (Fig. 5).

The stem surface did not influence either the global hip score or any of the three parameters of the Merle d’Aubigné scale (Fig. 6) for the 246 patients (276 hips) who were alive and had no revision at the time of the minimum 10-year evaluation. There were 165 female and 81 male patients. The mean Merle d’Aubigné functional hip score increased (p < 0.001) from 11.2 ± 2.6 points preoperatively to 17.4 ± 0.6 points at the time of the last followup.

A limitation of this study was the design of the polished stems was not identical; their cross-section was quadrangular as opposed to the oval of the satin and matte stems. However, the coronal and sagittal shapes of the three stems were similar with a regularly decreasing cross-section area according to an angle of 5°. Moreover, none of the stems used in this study had a cylindrical shape, which has been demonstrated to increase loosening as a result of rotational instability [25]. The satin stem series was historic rather than concurrent, and certain aspects of the surgery or postoperative care may have differed. At the time the randomized study (satin versus matte finishes) was conducted, there were no previous data in the literature and we did not perform a sample size power analysis: the numbers of patients were chosen arbitrarily. Moreover, the randomization process was not sophisticated and only based on operating rooms. No statistician participated at that time. Although we did not stratify by diagnosis, comorbidities, or other potentially confounding variables, the three groups of patients were comparable in terms of demographics, preoperative functional status, surgical technique including operating surgeons and mode of cementation, and postoperative regimen. All patients were evaluated clinically and radiographically after a minimum 10-year followup with less than 10% of patients lost to followup.

The data suggest the matte stems had a lower survival rate with radiographic loosening as the end point when compared with polished and satin stems. Similar results have been reported with other stem designs cemented according to modern cementing techniques. Howie et al. [13] compared 20 polished (radius, 0.01–0.03 μm) with 20 matte (radius, 0.7–1.3 μm) Exeter stems (Stryker, Howmedica, Rutherford, NJ) after a minimum 9-year followup. In the matte stems group, revision for femoral aseptic loosening was performed in four hips, whereas no polished stem was revised for loosening. Ong et al. [28] reported on 254 uncoated satin finish Harris Design-2 stems (radius, 0.75 μm) (Howmedica) with a mean followup of 13.5 years versus 514 methylmethacrylate precoated stems (Zimmer, Warsaw, IN) of similar geometry with a mean followup of 8.4 years. Two precoated stems were used, including the Harris Precoat stem (not textured; radius, 0.8 μm) and the Harris Precoat-Plus stem (proximally textured; radius, 2.2 μm). Failure was defined as revision or pending revision, progressive complete radiolucent line greater than 2 mm, stem migration, or stem fracture. The coated stems failed at a higher rate; at least 10 (eight revised, two radiographically loose) of the 254 uncoated stems (3.9%) and at least 49 (43 revised, six radiographically loose) of the 514 precoated stems (9.5%) had failed at a mean followup of 9.9 and 5.7 years, respectively. The most extensive literature that has been published concerning the effect of surface finish has been reported with the Iowa stem (Zimmer) [1–3, 25, 28, 31]. Indeed, this cobra-shaped femoral component has been developed with a polished (radius, 0.1 μm), satin (radius, 0.8 μm), and grit-blasted methylmethacrylate proximally precoated (radius, 2.1 μm) surface finish. Sporer et al. [31] compared 36 satin stems with 45 grit-blasted stems in patients younger than 50 years old after a mean followup of 11.3 and 8.2 years, respectively. Using revision of the femoral component resulting from aseptic loosening as the end point, the survival rate of 93% at 10 years was greater for the satin stems compared with 86% at 7 years for the grit-blasted stems. The survival rate using loosening of the femoral component (with or without revision) as the end point was 91% at 10 years for the satin stems versus 84% at 7 years for the grit-blasted stems. Collis and Mohler [3] reported on 122 polished and 122 grit-blasted Iowa stems after a mean followup of 5.8 years. Six hips in the grit-blasted stem group had femoral aseptic loosening, including four revised stems, whereas none of the polished stems were loosened. Recently, Della Valle et al. [6] compared 64 textured (radius, 1.75–2.5 μm) with 138 satin-finish (0.5 μm) Versys stems (Zimmer) after a followup of 4 to 8 years. They reported more stems (seven hips) in the textured group failed than in the satin surface group (no failed hips). Two prospective, randomized studies compared similar stems differing only by the surface finish [29, 32]. Neither showed differences in femoral loosening and revision rates. However, in both of these studies, the so-called rough stems had a matte finish only on the proximal third of the implant (radius, 4.25 μm and 2.75 μm, respectively), whereas the distal portion of the stem was either highly polished (radius, 0.1 μm) or had a satin surface finish (radius, 0.75 μm). Moreover, the mean followup was 4.8 and 6.5 years, respectively.