Methods
Trial design
HemiSPAIRE was a pragmatic, superiority, parallel-group, randomized controlled trial (RCT) (with internal pilot), with blinded assessment, comparing patient outcomes under the SPAIRE and standard lateral approaches to hip hemiarthroplasty for displaced intracapsular fractures. The trial included an embedded qualitative study. The protocol and statistical analysis plan (SAP) are published.9 10 Conduct and reporting followed Consolidated Standards of Reporting Trials guidelines.11 All randomized participants provided informed consent. An independent trial steering committee (TSC) reviewed progress against prespecified stop-go criteria9 and throughout the trial.
Participants
Participants attending one of six hospitals in South West England with a displaced intracapsular fracture requiring hip hemiarthroplasty were recruited. If potentially eligible, the study was discussed with the patient and/or their carer(s). Potential participants were identified by orthopedic surgeons involved in the study, who admitted patients under their care. Identified patients interested in participating were invited to read the participant information sheet and, if interested, to provide informed consent. If a patient lacked capacity and was unable to consent, potential participation was discussed with a healthcare professional who was involved in the patient’s care but not part of the research team, and if deemed appropriate a declaration was signed as well as seeking the opinion of a person whose relationship to the patient made them suitable to act as his/her legal representative. Participants who lacked the mental capacity to consent and did not seem in agreement with any part of the study, even if agreement had been given by another, were not included.
Patients were eligible if they were aged ≥60 years, presented with an intracapsular hip fracture requiring hip hemiarthroplasty, and were resident in South West England. Exclusion criteria were being immobile (unable to walk) before hip fracture; not expected to live until postoperative day 120 due to chronic illness and receiving surgery for palliative care; or use of femoral stems not of a proven stem design, in line with NICE clinical guideline on hip fracture management.4 A sizeable proportion of this population suffers dementia and/or temporary delirium and was not excluded. A systematic review of observational studies found that 19% of people with hip fracture meet formal diagnostic criteria for dementia, and 42% are cognitively impaired.12 Cognitive ability was not part of the HemiSPAIRE eligibility criteria.
Randomization and blinding
Randomization was undertaken as late as practically possible. There was no special preparation required in theater, and no difference in equipment required for either surgical technique. Participants were individually randomized to receive the SPAIRE or lateral procedure in a 1:1 ratio. Concealed allocation was determined by the UKCRC registered Exeter Clinical Trials Unit (CTU) using a validated password-protected web-based system, based on random permuted blocks of varying size (2, 4 or 6). Stratification variables were hospital site and cognition level (impaired vs non-impaired), from information gathered by the research nurse in the participant’s records. The surgeon was informed of allocation by the CTU via email through nhs.net mail. The trial chief investigator and the principal investigator at the hospital site were copied in. Participants, ward staff, all research staff involved in assessments, and the trial statistician, were blinded to treatment allocation. Surgeons and operative teams were unblinded. There is no difference between the SPAIRE and lateral approach techniques in the following: preparation for surgery, patient positioning, skin incision, surgical time taken, application of surgical dressing or postoperative care. For medicolegal reasons, the surgical approach used was specified in the operation notes. A cover sheet was attached to the front of the printed operation notes stating that the patient is a trial participant and reminding the research team to avoid inadvertent unblinding to treatment allocation during postoperative assessments. Provision of information on the study, consenting, randomization and postoperative assessments were carried out by research teams at each site.
Procedures
The SPAIRE technique via the posterior approach to the hip involved a modified muscle sparing posterior approach where insertions of piriformis, superior gemellus, obturator internus and inferior gemellus were spared with division of only obturator externus and part of quadratus femoris. The single divided tendon and posterior capsule were subsequently repaired with a transosseous repair to their initial position prior to closure. The insertions of the abductor muscles were left intact throughout the procedure. Further details of the SPAIRE technique are published elsewhere.13 In the lateral approach, the operation was performed in accordance with criteria set by the study, to minimize issues of standardization with this approach. This means that the gluteus medius and minimus insertions onto the greater trochanter were partially divided anteriorly, leaving the posterior part of their insertions intact. The anterior capsule was divided and subsequently repaired prior to closure, followed by repair of the detached portion of the gluteal muscles.
Outcomes
The primary outcome was function and mobility, reported by the participant (or by proxy by a family member or carer if the participant was unable to answer the questions), measured using the OHS (range 0–48; 48 is the best score),14 15 at 120 days postoperation. The 120-day follow-up was chosen for the primary outcome because it is the time of final review offered to all hip fracture patients by the National Hip Fracture Database. Secondary outcomes included level of function and mobility at 3 days postoperation (and 120 days if possible) using the De Morton Mobility Index (DEMMI) (range 0–100; 100 is best score),16 early (3 days) mobility using the Cumulated Ambulation Score (CAS) (range 0–18; 18 is best score),17 level of pain using a Numeric Pain Rating Scale (NPRS) from 0 (no pain) to 10 (worst possible pain) at 3 and 120 days, health-related quality of life (European Quality of Life 5 Dimensions 5 Level, EQ-5D-5L) (range <0 (health state worse than death) to 1 (full health), with 0 being value of a heath state equivalent to dead)18 19 at 120 days, acute and total length of hospital stay, hip-related complications and mortality within 120 days of operation, acute discharge destination, and place of residence at 120 days. Participants were assessed face to face at 3 days and via telephone at 120 days, by research nurses.
Sample size
The original primary outcome was the DEMMI, and the trial was powered to detect a difference of 6 points16 between trial arms, at 120 days (assuming an SD of 11.9), with 90% power, at the 5% level of significance, allowing for 25% loss to follow-up, giving a recruitment target of 224 participants (112 in each arm). The primary outcome was changed to the OHS (non-substantial amendment approved by the sponsor June 17, 2020 and acknowledged by the Health Research Authority (HRA)). This was because the DEMMI must be assessed in person with the participant, requiring either an outpatient appointment or home visit, which were not possible during periods of COVID-19 restrictions. The OHS can be collected remotely, via the telephone. The recruitment target was updated, based on a minimal clinically important difference for OHS of 5 points20 and an SD of 10 (ie, an effect size of 0.5), with 90% power, at the 5% level of significance, requiring 85 participants per trial arm, that is, a total of 170. Allowing for 25% loss to follow-up, the recruitment target was 228. Review of follow-up rates by the study team during the trial indicated that drop-out rates were higher than originally anticipated and as such the recruitment target was updated again (non-substantial amendment approved by the sponsor December 2, 2021 and acknowledged by the HRA), to allow for a 30% loss to follow-up, requiring a total of 244 participants (122 in each arm).
Statistical analysis
Statistical analyses are detailed in the published SAP.10 The SAP was prespecified and approved prior to final data collection. To aid presentation of trial findings to the trial team, the trial statistician was unblinded at the point of receiving the final trial data, which was after all data queries had been resolved and the trial database had been locked. Analyses were completed using Stata version V.17.0 (StataCorp).
Comparisons of outcomes between trial arms used the all-randomized population, under the intention-to-treat (ITT) principle with participants analyzed according to the trial arm they were randomized to.
The main analyses used a survivor average causal effect (SACE) approach, which allowed estimation of the effect of surgical approach on outcomes in the population of people who would have survived regardless of what surgical approach they received.21 Further details of this method are in the published SAP.10 The main analyses were based on the complete case data. The adjusted analyses are the main analyses. Analyses based on multiple imputed datasets were conducted as additional analyses. Multiple imputation was used to impute missing data, under the assumption that data were missing at random according to Rubin’s rules,22 using the chained equations approach and predictive mean matching.23 50 imputed datasets were generated. Variables used to impute missing data included all outcomes at all follow-up time points, trial arm status, stratification variables, and variables included as adjustment factors in the regression models fitted to outcomes. While all participants were included in the imputation process, no outcomes were imputed for participants who died before the outcome could have been assessed. Further additional analyses were conducted. First, using the trial data (ie, not using multiply imputed data): (1) Analyses using linear (for continuous outcomes) and logistic (for binary outcomes) regression, among those participants who survived—a survivors analysis. The population of interest comprised participants who survived under the surgical approach they received. (2) Analyses using linear and logistic regression, setting those participants who died before the outcome could be assessed to the worst possible score (for continuous outcomes OHS, DEMMI, CAS, NPRS), the score equivalent to being dead (for EQ-5D-5L), the worst score observed among all participants (for lengths of stay), or the worst category (for binary outcomes discharged to prefracture residence, and living at prefracture place of residence at 120 days)—a composite approach. The population of interest comprised all participants regardless of whether they survived or not. (3) Analyses including any outcomes collected outside the prespecified data collection window (ie, outside days 110–130), using a SACE approach. (4) Analyses of all outcomes that could be completed by proxy, excluding those participants for whom data collection was by proxy, using a SACE approach. Second, the following analyses were repeated, based on multiple imputed datasets: (1) main analyses using a SACE approach; (2) survivors analyses; and (3) composite analyses.
Patient and public involvement
Patient and public involvement (PPI) helped shape the trial and inform the primary outcome and premise of the work. During trial development, meetings were held with patients to discuss the importance of the research, recruitment and acceptability of outcome measures. It was clear that mobility was the highest priority. Patient/carer coapplicant (AA) was involved from conception of the trial idea and as a trial management group member, AA worked closely with the trial team, including the PPI facilitator (EC), ensuring a patient focus throughout. PPI was integrated throughout the trial from design to dissemination. NIHR INVOLVE guidance for good practice for effective involvement was followed, including payment and reimbursement of expenses for patient advisors. An advisory group of two people with experience of caring for a family member after a hip fracture met on three occasions, with additional correspondence via email. This group informed the production of study materials, gave feedback on the qualitative interview schedule and was involved in the interpretation of findings.
Qualitative study
The embedded qualitative study examined participants’ experiences of their postoperative recovery period and compared findings between trial arms. 18 semistructured telephone interviews were conducted by one researcher (EC). A subsample of trial participants (8 in SPAIRE arm and 10 in lateral arm) were recruited from 3 sites. Participants were interviewed around 130 days after surgery. Interviewees were asked about impact of surgery and experience of recovery during the postoperative period.
Interviews were transcribed verbatim and uploaded to NVivo. Data were analyzed using adapted thematic analysis, employing a combined deductive-inductive approach to explore known factors that would be experienced during recovery but also allow for unexpected findings.24 Four interviews were double-coded (SM-T, EC) and SM-T and EC met periodically during the analysis to develop and refine codes and themes. Findings were then compared between trial arms by EC and SM-T. Researchers were blinded to the trial arm of interviewees until analyses were complete.