This multicenter parallel randomized controlled trial included patients with a displaced femoral neck fracture. The study was approved by the Regional Ethics Committee (NL19200.098.07, METC07 –118). The trial was registered in the Netherlands Trial Registry NTR 1508 (http://www.trialregister.nl). The protocol was published before start of the study [14].

All patients were admitted to one of the participating hospitals (Reinier de Graaf hospital, Delft; Rijnstate hospital, Arnhem and Canisius Wilhelmina hospital, Nijmegen), large district hospitals in the Netherlands. Inclusion was between August 2008 and June 2012. Included were patients aged 70 years or older, with a displaced femoral neck fracture (Garden type III or IV) suitable for hemiarthroplasty. Excluded were patients with a pathological fracture, a fracture older than seven days or ASA-IV or V classification. Orthopedic residents, trained for this study, performed inclusion.

All patients gave informed consent. In case of (mental) incompetence of the patient, his or her legal representative was consulted to obtain informed consent. Patients were randomized following a simple randomization procedure in the operation theatre by the orthopedic surgeon through opaque sealed envelopes. These were prepared by A.J.V. and kept at the operation theatre of each of the three hospitals. 200 opaque sealed envelopes were prepared. However, 16 patients could not be included in our trial due to variable reasons (Fig. 1), which forced us to prepare another 16 envelopes.

The patients were blinded for the type hemiarthroplasty they received, although we acknowledge the possibility that they might be able to tell after seeing their radiographs during the outpatient clinic visits. Surgeons and outcome assessors were aware of the allocated arm.

Patients received a cemented hemiarthroplasty, type Müller Straight Stem (Zimmer - Biomet, 1800 West Center St. Warsaw, Indiana, USA) or an uncemented hemiarthroplasty, type DB-10 (Zimmer- Biomet, 1800 West Center St. Warsaw, Indiana, USA). The cemented hemiarthroplasty, the Muller straight stem has a small proximal collar and two longitudinal grooves to enable good cement adhesion. The non-cemented DB-10 is a straight collared stem with metaphyseal anchoring and on the surface full hydroxyapatite coating on macro-structured titanium and grooves. If complications occurred during the procedure, the surgeon could change the procedure to ensure best medical practice. Operating technique was according to the manufacturer instruction. In the participating hospitals there was experience with both cemented and uncemented hip arthroplasty. Either an orthopedic surgeon or registrar performed the operation. Cementing technique involved vacuum mixing, cement plug, saline pulsed lavage and retrograde introduction of cement with a cement gun. The approach was up to the surgeon’s preference, as Parker’s Cochrane analysis has shown that insufficient evidence is available for superiority of either approach [3].

Each patient received physiotherapy therapy, analgesia and trombo-embolic prophylaxis according to the protocol of the hospital in which they were treated.

Preoperatively, social demographic data (age, sex, place of residence), ASA— (American Society of Anesthesiologists) classification [15], Body Mass Index (BMI), Minimal Mental State Examination (MMSE) [16] were obtained. Patients were asked to score their pre-fracture mobility and Health Related Quality of Life (HRQoL) using the New Mobility Score (NMS) [17], Groningen Activity Restriction Scale (GARS) [18] and the SF-12 [19]. Patients were asked if they mobilized with an aid indoors and outdoors with or without aid and whether they received homecare. The baseline hemoglobin level was measured. The surgical approach, the type of surgeon (consultant or registrar) and kind of anesthesia were registered.

Outcomes measured during operation were operation time (defined as skin-to-skin surgical time, measured in minutes) and blood loss (in centiliter, estimated by the surgeon).

Length of stay, decrease in hemoglobin level and transfusion rate were measured postoperatively.

All patients were invited for follow up at six, 12 and 52 weeks postoperatively. When the patient was not able to visit the outpatient clinic, the questionnaires were mailed to the patient or its relatives. During follow-up functional outcome was measured using Timed-Up and- Go (TUG) score [20], GARS [18] and NMS [17]. HRQoL, expressed in the SF-12 [19], was measured. The SF-12 was divided in a Physical Component summary Score (PCS) and a Mental Component summary Score (MCS). Mid-thigh pain (defined as pain explicitly in the front and mid part of the femur) pain and place of residence were registered. Complications during surgery, hospital stay and the year thereafter were recorded. The complications were defined and ranked in the modified Elixhauser mode, as described by Parvizi [21]. Mortality was scored meticulously by repeated consultation of the population registers of the counties in the region of the hospital as well as the hospital’s patient registration systems for the full length of follow-up.

A radiograph was obtained on the first postoperative day and after six weeks, 12 weeks and 1 year. Adequate positioning of the stem was defined as within 10° varus or valgus position with respect to the femoral axis. Fissures, fractures, subsidence and loosening were noted.

In total 201 patients were analyzed (Fig. 1) 91. Were randomized to an uncemented, 110 to a cemented hemiarthroplasty. In 15 of the 91 (16%) patients randomized to an uncemented hemiarthroplasty a cemented hemiarthroplasty was used instead. In ten patients this was due to intraoperative complications (i.e. fracture of the femur). In four patients the necessary instruments or prosthesis were not present and in one patient the reason was unknown. Four of the 110 (4%) patients randomized to a cemented hemiarthroplasty received an uncemented hemiarthroplasty. In none of these cases the reason for this breach of the protocol was clear.

Secondary outcomes;

There was no difference in the number of patients who returned to their baseline place of residence after 1 year (28 patients (72%) vs. 37 patients (80%) p = 0.88).

The SF-12 MCS did not differ between the cemented and the uncemented group. (Table 3) However, the SF-12 PCS was lower at six and 12 weeks postoperatively in the uncemented hemiarthroplasty group. This difference resolved after 1 year (Table 3 and Fig. 2). Analyzing the results according the as treated analysis showed a lower PCS for uncemented hemiarthroplasty at six weeks (30.3 vs. 34.8 p = 0.01), and a difference at 12 weeks (34.0 vs. 37.9) which nearly did reach significance (p = 0.056). There was no difference at 1 year after surgery (36.6 vs.37.6 p = 0.65).

Radiographs were taken direct post-operative and after 6 and 12 weeks and 1 year. Five patients deceased before the post-operative radiograph was obtained. Eight varus or valgus deviations in the uncemented group and seven in the cemented group (p = 0.76) were found on the post-operative radiograph. Loosening or subsidence was observed in 13 (20%) of the uncemented and five (6%) of the cemented hemiarthroplasties (p = 0.007) any time during follow up. Four (2%) revision operations were performed: three due to loosening (all in the uncemented group) and one for infection (in the cemented group) (p = 0.162).

There were no differences in length of stay between both groups (mean 11 (SD 7.7) days uncemented vs. 11 days (SD 8.3) cemented p = 0.83), loss in hemoglobin level (g/dl) after surgery (uncemented mean 2.2 (SD 1.4) vs. cemented mean 2.0 (SD 1.5) p = 0.31) and transfusion rate (uncemented 17 (24%) vs. cemented 22 (26%) p = 0.74). Surgeon estimated blood loss was larger in the uncemented group (mean 288 mL (sd 213) vs. mean 220 mL (sd 143) p = 0.03) (Additional file 1: Table S1).

The most important finding of our study was that major local complications were more frequent (odds ratio; 95% CI) (3.36; 1.40 to 8.11) in the uncemented hemiarthroplasty group compared to the cemented group. In elderly patients with a displaced femoral neck fracture hemiarthroplasty is a widely accepted treatment of choice [2]. Previous randomized trials comparing cemented and uncemented hemiarthroplasty give conflicting results on this [4, 11–13, 22].

A periprosthetic fracture was the most common major local complication in the uncemented group (15%). Previous papers comparing fracture rate are heterogenic. Some studies found a higher fracture rate in the uncemented group, ranging from 5.5 to 12% [3, 12, 13, 23] whereas others demonstrated no difference in fracture rate [4, 11, 22]. The fracture rate in the current study is higher than previous papers demonstrate. The teaching of registrars might have attributed to this. This can be due to the design of the DB-10 stem (proximal fitting) compared to the stems used in the other papers. Furthermore, teaching of registrars might have attributed to this as well: the level of experience of the operation surgeon is not always mentioned, but for example in the studies of Inngul, DeAngelis and Parker the operations were always performed by consultant orthopaedic surgeons [4, 11, 13]. However this high complication rate might better reflect the everyday practice with registrars often performing this type of operations.

One-year mortality rate was higher in the uncemented group (25 (27.4%) compared to the cemented group 21 (19.0%) but did not reach significance (p = 0.16). This is in contrast to other randomized controlled trials [4]. The register studies show higher mortality in the first operative days in cemented hemi arthroplasty [24–26]. However the Australian and British register shows lower mortality in cemented hemiarthroplasty the year thereafter [10, 27]. Power analysis in our study was not performed on finding differences in mortality. In our trial one patient died intraoperatively, probably due to Bone Cement Implantation Syndrome (BCIS). BCIS is a major side effect of cement implantation, it has no agreed definition but is characterized by a number of clinical features with amongst others hypoxia, hypotension, cardiac arrhythmias, increased pulmonary vascular resistance and cardio-respiratory collapse [5]. Guidelines to minimize the risk for BCIS by both surgeon and anesthetist are recently published [28].

In contrast to our hypothesis and literature, we did not find a difference in operation time between groups [3, 22]. Intraoperative complications in the uncemented group might have affected this, however equality in the mean operation time persisted when we analyzed our data in an As Treated analysis. Teaching residents during might have affected the operation time in such way that the difference disappeared.

Functional outcome (GARS score and TUG test) was not different between both groups. However TUG was measured only in 53% of all patients at 6 weeks, 51% at 12 weeks and 48% at 1 year. Probably TUG is not a very useful outcome measure in this frail population as mobility was too poor to measure well. Mobility expressed as the NMS was comparable between the two groups. In literature different outcome measures for functional outcome have been used (Oxford or Harris Hip Score [12, 13, 22] Older Americans Resources and Services Instrument [11] Barthel Index [22]) A meta-analysis pooled results of five trials (491 patients) and found that patients with an cemented hemiarthroplasty had a better hip function after 1 year [7].

Mid-thigh pain is known to be more prevalent in uncemented prostheses, however the reported incidence differs tremendously [4, 7, 29]. In our study presence of mid-thigh pain was comparable between groups. Several factors can be of influence on post-operative mid-thigh pain such as sizing, design and stiffness of a prosthesis [29].

Radiological follow up showed loosening or subsidence in 13 uncemented hemiarthroplasty, which led to a revision in three cases. Subsidence in uncemented (hemi) arthroplasty is a common finding; one trial found a larger subsidence in the uncemented hemiarthroplasty [12]. In The Norwegian Hip Fracture Register patients treated with an uncemented hemiartroplasty had a 2.1 times increased risk of revision compared with patients treated cemented prostheses. This increased risk of re-operation was due to peri-prosthetic fracture HRR 17 and aseptic loosening HHR 17 [30]. A combined analysis of the Norwegian and Swedish registers (33.205 hip fractures in patients older than 60 years treated with hemiarthroplasty) also found more reoperations in uncemented stems (HR2.2) [9].

PCS of HRQoL was lower in patients treated with an uncemented hemiarthroplasty at six weeks and three months after surgery. PCS HRQoL is known to decrease in the first three months and recover thereafter [31]. The larger complication rate might have led to this lower PCS HRQoL, although we would have expected to find a difference in functional and mobility scores as well. One previous trial found higher HRQoL (expressed in EQ5days) in the cemented group [13] at 4 and 12 months, another trial did not [22]. The latter did not find a difference in complications either [22]. HRQoL was an secondary outcome in this trial, thus no power calculations were made and dropout at follow-up was quite high: therefore the difference we found might be due to coincidence and has to be verified in further trials.

The large number of patients, the randomized design and outcome measures on both functional and radiological outcomes make the current study worthwhile. Furthermore, we did not exclude patients with cognitive disorders. The latter makes our study generalizable to all elderly hip fracture patients treated with hemiarthroplasty.

However, our study does have limitations. First, many (293) patients (or their caretakers) declined to participate in our study or were not asked to participate, thus selection bias might be present. Second, poor registration has led to incompleteness of some of the baseline data. This led to the exclusion of 12 patients after randomization because of missing baseline data, 5 patients were excluded after randomiz ation due to other reasons (Fig. 1). Deviation from the protocol occurred in 19 patients (9% of analyzed cases). Therefore both as treated analysis and intention to treat were performed. More deviations from protocol were present in the uncemented group (15 vs. 4) which might have caused a bias.

Furthermore, we had a substantial percentage (24%) of patients who were lost to follow-up. This might be due to the inclusion of patients with cognitive disorders (38% had an MMSE less than 24) and high age of the participators, resulting in their caretakers to refrain from extra stress by those patients by filling in follow-up forms.

This trial adds value to the discussion whether to use a cemented or an uncemented hemiartroplasty in femoral neck fractures. Conflicting evidence on this matter is published the last few years [4, 11–13, 22]. Trial design, prosthesis design, inclusion criteria and whether the trail was performed in a teaching hospital or not might all have been of influence of these published results.

Elderly patients with a displaced femoral neck fracture treated with an uncemented hemiarthroplasty had more periprosthetic fractures, loosening, reoperations and lower quality of life compared to patients with a cemented stem. Operation time, functional outcome and mid-thigh pain were comparable between groups. Based on these findings, and earlier work [3, 12, 13] we conclude that in elderly patients with a displaced femoral neck fracture a cemented hemiarthroplasty is favorable compared to an uncemented stem.