However, in these patients, the incidence of dislocation ranges from 2% to 9% after THA [3, 4, 15, 17, 26], partly because of different durations of followup for this specific endpoint. Thus, the better clinical functional scores with THA could be offset by an increased risk of dislocation as compared with hemiarthroplasty [14].

In the hemiarthroplasty group (412 patients), 201 patients were men and 211 were women, with an average age of 80 years (range, 65–92 years) and a diagnosis of neck fracture. All patients received a cemented arthroplasty (280 bipolar, 132 unipolar hemiarthroplasties), performed between 1990 and 1994 at the same hospital. The 90-day mortality for patients undergoing hemiarthroplasty for hip fracture was 2.6%. Forty patients died before their 5-year followup and eight had revision surgeries for infection. Additionally, 63 of the 412 patients were lost to followup after 5 years. This left 309 of the 412 patients (75%) with a followup of 5 years, 201 (49%) with a followup of 10 years, 87 (21%) with 15 years, and 32 (8%) with 20 years. We examined their records at their most recent followup or before their death. The average followup, as reported in medical records, was 9 years (range, 1–20 years). Surgeons performed surgery using a posterolateral approach and general anesthesia. The stems were always the same (Ceraver Osteal, Roissy, France), made of anodized titanium alloy (TiAl6V4), smooth, and always cemented. Surgeons fixed femoral components with cement (Palacos® G; Heraeus Medical GmbH, Hanau, Germany) containing antibiotics (gentamicin).

The THA group consisted of 380 patients (175 men, 205 women), with an average age of 79 years (range, 64–90 years). Between 1995 and 1999, all patients received a cemented THA at the same hospital. The 90-day mortality in these patients undergoing THA for hip fracture was 2.8%. Thirty patients died before their 5-year followup. Nine patients had revision surgeries for infection, and 51 of the 380 patients were lost to followup after 5 years. This left 299 of the 380 patients (79%) with a minimum followup of 5 years, 191 (50%) with a followup of 10 years, and 98 (26%) with 15 years. The average followup (as reported in medical records) was 7 years (range, 1–15 years). Surgeons performed this surgery using the same technique, general anesthesia, and stems (also cemented) as used in the hemiarthroplasty group. The conventional acetabular component was a standard PE cup manufactured by Ceraver (all cemented). The alumina head was 32 mm in diameter and anchored through a Morse taper.

Age and gender proportions were similar in the two groups (mean age of 80 years in the hemiarthroplasty group versus 79 years in the THA group).

For postoperative rehabilitation, no patient wore a hip spica cast to minimize the possibility of hip dislocation. We encouraged patients to be upright with weightbearing after surgery. Range of motion of the hip and knees began immediately, and ambulation progressed as tolerated according to the ability of each patient (usually after 48 hours). Medical doctors supervised the physiotherapy for 3 weeks to 45 days.

The followups were at 3, 6, and 12 months, then annually thereafter. At each visit, patients had clinical and radiologic evaluations. The usual hip-rating scores do not reflect the function of THA well in patients with hip fractures because the age or medical status may affect the functional variables. The preoperative and postoperative walking status of patients was graded as unable to walk, able to walk indoors only, able to walk about the community, and able to walk an unlimited distance. All patients, except the eight with hip infections, subjectively showed improved ambulation (they were able to walk about the community or able to walk an unlimited distance) after surgery. Every year until the most recent followup, plain radiographs (AP and lateral views) were obtained. From the charts, we determined whether there were any dislocations (posterior or anterior) or revisions for recurrent dislocation, loosening, or wear of cartilage for hemiarthroplasties. We routinely evaluated patients, in person or via a standardized letter or telephone questionnaire, at 2 to 3 months postoperatively, at 1, 2, and 5 years, and then at each subsequent 5-year interval until revision or death. At each time, we specifically asked patients whether they had experienced dislocation of the hip. We did not consider a sensation of subluxation of the hip to be sufficiently specific to represent an episode of hip instability, and only considered a complete hip dislocation requiring reduction to be a dislocation event.

Three of us (PH, MB, OP) evaluated all immediate postoperative radiographs and all radiographs at last followup to assess wear of the cartilage for hemiarthroplasties, and loosening of the cup for THA. We defined loosening of the socket as cup migration exceeding 5 mm, angular rotation exceeding 5º, and a continuous radiolucent line wider than 2 mm [6]. To assess cartilage wear we used the criteria of Baker et al. [2]. Cup position was assessed according to the acetabular position angle [22].

Recurrent hip dislocation and the cumulative risk of dislocation were greater with THA. Nine patients (nine hips) treated with THA had recurrent hip dislocations: four had two recurrent dislocations, two had three dislocations, one had four dislocations, and two had five dislocations. Therefore, the cumulative number of dislocations (first time and recurrent) for patients treated with 380 THAs was a total of 58 dislocations (13%) at the most recent followup. The cumulative risk of dislocation for surviving patients with THAs who did not have revision surgery was 3% (95% CI, 1.5%–4.5%) at 2 months and 6% (95% CI, 3%–9%) at 1 year. The risk then increased at a constant rate to approximately 0.5% every year, to 13% (95% CI, 9%–18%) at 15 years for patients who had not had a revision by this followup. Six THAs needed revision for recurrent dislocations. In the group of patients who underwent a hemiarthroplasty, the incidence of recurrent dislocations was lower (three patients with recurrent dislocations two times among 412 hips) than in the patients with a THA (nine of 380 patients). Therefore, the cumulative number of dislocations (first-time and recurrent dislocations) for patients treated with 412 hemiarthroplasties was lower (p = 0.01), with a total of 22 dislocations (5%) at the most recent followup. The cumulative risk of dislocation with hemiarthroplasty was 1% at 1 month, 1.9% at 1 year, and 5% (95% CI, 2%–6%) at 15 years (Fig. 2). Only one hip had revision for this problem.

There are several limitations to this study. First, we did not randomize the patients into the two groups. However, we believe it is reasonable to compare the two groups, as we could identify (Table 1) only five studies in which a hemiarthroplasty with cement was compared with a THA with cement [2, 7, 9, 14, 18]. Second, the surgeons performed the two procedures at two different times; however, the surgeons were the same and used the same approaches and perioperative treatments. Third, we lost some patients to followup during this study, and others died. It is possible some patients had a dislocation treated at another hospital and did not report it owing to being cognitively impaired [10] when they became older. Another limit is that the dislocation rate may be affected by the lack of modularity (using 32-mm ceramic heads with limited neck lengths in our series), by other factors [1] than the choice of the implant, including orientation of the acetabular and femoral components, and leg length discrepancy, factors that were not analyzed in this study. The risk of dislocation also might be affected by the surgical approach [8], by the design such as dual-mobility cup and constrained liner, or by the size [21] of the femoral head. THA is a more complex procedure than hemiarthroplasty or fixation, and one concern could be the possibility that the mortality rate [16] would be higher. This did not occur in our series, and the 30-day mortality in another series [20] of more than 7000 patients undergoing THAs for a hip fracture is 2.4%, consistent with our observation. Two randomized trials showed similar mortality rates with THA compared with other treatment options [9, 18]. Therefore, it would seem that increasing age and medical comorbidities [16], rather than the magnitude of the surgical procedure, influence mortality.

Our data (Table 1) suggest that after THA there is a higher initial risk level for first-time dislocation that increases with time, and that, after hemiarthroplasty, the risk of first-time dislocation is lower and decreases after 6 months; thus, the longer patients are followed, there is a greater cumulative risk for patients who underwent THA compared with patients who underwent hemiarthroplasty in the treatment of fractures. Our cumulative risk of dislocation for patients with THAs is higher than expected. The fact that many studies of THA for femoral neck fractures have relatively short followup periods [7, 14, 18], often for only 2 years, might explain this. Hip dislocation rates in two studies [4, 14] are 2% and 4%, but with only 2 to 4 years followup, respectively. However, another series of 51 THAs [3] had cumulative rates of 6% at 10 years and 9% at 25 years. Our data are consistent with this observation.

The risk of revision for recurrent dislocation increased with THA, but it remained lower than the risk of revision for cartilage wear and acetabular protrusion in the hemiarthroplasty group. This conforms to the literature: acetabular erosion has been considered to occur with rates ranging from 2% [5] to 36% [25] with a unipolar design and from 0% [5] to 26 % [24] for a bipolar design. Furthermore when acetabular erosion was graded, Baker et al. [2] reported a 66% rate of acetabular erosion. Hemiarthroplasty, regardless of type, does not seem to fulfill the active, independent patient’s functional demands. Van Vugt et al. [28] concluded that their study group of independent living patients between 71 and 80 years old should not have a bipolar hemiarthroplasty owing to their higher risk of acetabular protrusion and loosening. Three studies [15, 17, 26] have longer-term followups and calculated 10- and 20-year survival of THAs after femoral neck fractures. These studies indicate that the 10-year survival of THA is in excess of 90%. Tarasevicius et al. [26] reported 10-year survival rates greater than 90% for stems and cups in THAs in a series of 135 patients. Mabry et al. [17] reported 93% 10-year survival and 76% 20-year survival of Charnley cemented implants used to treat femoral neck nonunion. Lee et al. [15] reported survival rates of 95% at 5 years, 94% at 10 years, 89% at 15 years, and 84% at 20 years for cemented THAs used as a primary treatment for intracapsular hip fractures.