Physical function and lean body mass as predictors of bone loss after hip fracture: a prospective follow-up study

This study utilizes data from a randomized controlled trial (ISRCTN53680197) investigating the effects of a yearlong home-based rehabilitation program compared to standard care on mobility recovery among over 60-year-old, ambulatory and community-dwelling older people with a recent hip fracture [18]. The design and recruitment procedure have been reported in detail earlier [19]. Briefly, all men and women who had been operated for a femoral neck or pretrochanteric fracture (ICD code S72.0 or S72.1) in the Central Finland Central Hospital (Jyväskylä, Finland) between 1.3.2008 and 31.12.2010, and fulfilling the inclusion criteria were informed about the study (n = 269). Of these, 161 were interested and further informed on the study. After preliminary assessment of eligibility, 136 persons were invited to the baseline measurements. Patients suffering from severe memory problems (Mini Mental State Examination, MMSE < 18), alcoholism, a severe cardiovascular or pulmonary condition or some other progressive disease, and severe depression (Beck Depression Inventory BDI-II > 29) were excluded from the study. Thereafter, 81 eligible patients participated in the study.

After the baseline measurements, conducted on average 10 weeks post fracture, the participants were randomized into an intervention (n = 40) and a standard care control (n = 41) group using a computer-generated group allocation list generated by a blinded statistician, who was not involved in either the recruitment or data collection process. Blocks of 10, stratified by gender and surgical procedure (internal fixation vs arthroplasty), were used. Follow-up measurements were arranged at 3, 6, and 12 months after the baseline measurements. All assessments were conducted at the research laboratory, and all outcome assessors were blinded to the treatment-group assignment. For the present analyses, data from the rehabilitation and standard care control groups were pooled, since the intervention had no effect on bone properties [10], and only baseline and 12-month follow-up bone data were utilized.

The sample size calculations have been reported in detail before [18, 19]. Briefly, an a priori sample size calculation was performed for the primary outcome, mobility limitation, based on previously published longitudinal data on mobility recovery after a hip fracture [20]. Based on calculations, a minimum of 44 participants were needed in each group (in total 88 participants) to detect the expected difference between the study groups at a level of significance of α = 0.05 and β = 0.20. Sample size was calculated using an online sample size calculator (DSS researcher’s toolkit).

All participants received standard care from the hospital. In addition, the intervention group received a year-long, physical rehabilitation program aimed at restoring mobility and physical functional capacity to the pre hip fracture level [18, 19]. The individually tailored program comprised an evaluation and modification of environmental hazards, guidance for safe walking, non-pharmacological pain management, motivational physical activity counselling and a progressive home exercise program. The intervention took place in the participants’ homes and included five to six home visits by a physiotherapist. The progressive home exercise program comprised strengthening exercises for the lower limb muscles using resistance bands, balance training in the standing position, stretching, and functional exercises including walking, reaching, turning in different directions and stair climbing [10, 18].

Bone scans from the distal tibia (5% of measured tibial length proximal to the distal end plate) and tibial shaft (55%) of both legs were obtained by pQCT (XCT-2000, Stratec Medizintechnik, Pforzheim, Germany) according to methods described earlier [10]. The pQCT scans were analyzed with an automated threshold-free cortical bone detection method [the outer boundary detection and subsequent shrinking (OBS) procedure, OBS cortical bone detection 2.1] [21, 22]. The outcome variables for the distal tibia were total volumetric bone mineral density (vBMD_TOT, mg/cm³) and compressive bone strength index (BSI, g²/cm⁴ = vBMD_TOT² × CSA_TOT) [8, 9]. For the midshaft site, the variables were cortical vBMD (vBMD_CO) and a strength-strain index (SSI, mm³; density-weighted polar section modulus) reflecting the bone’s resistance to bending and torsional loads. The root mean square coefficient of variation (CV_RMS) for the BMD and strength index measurements in our laboratory ranges from 0.4 to 1.6% [23].

Physical function at baseline was measured using the Short Physical Performance Battery (SPPB) test, which includes habitual walking speed, chair rise and standing balance tests [24]. A higher score (range, 0–12) indicates better performance. Perceived difficulty in walking outdoors was assessed by a questionnaire with the following response categories: 1) no difficulties, 2) some difficulties, 3) a great deal of difficulties, 4) manage only with help, and 5) unable to manage even with help.

Lean body mass (kg) was assessed with a bioimpedance device with eight polar electrodes (BC-418; TANITA, Tokyo, Japan). Participants were instructed to avoid caffeine for 2 h, alcohol for 36 h and physical exercise for 24 h before testing.

The presence of chronic conditions, use of prescription medications, fracture date and status, and type and date of surgery were assessed by means of a pre-structured questionnaire, current prescriptions and medical records and confirmed in a medical examination performed by a research nurse and a physician. Contraindications for the physical performance assessments were evaluated according to ACSM guidelines [25]. Body height and weight were measured using standard procedures, and body mass index was calculated as body weight divided by height squared (kg/m²). Fat percentage was assessed with bioimpedance device. Serum concentrations of 25-hydroxyvitamin D (25OHD, nmol/L) and parathyroid hormone (PTH, ng/L) were determined according to methods described earlier [10]. Smoking status was assessed by questionnaire and categorized as current, former and never smokers.

Mean values and standard deviations (SD) were calculated using standard procedures. To alleviate problems resulting from extreme outliers, we used the robust linear regression approach [26] to estimate the associations of the predictor variables with each dependent variable. The mean percentage changes in vBMD and the bone strength indices, used as outcome variables, were calculated as [(follow-up – baseline)/baseline × 100]. Baseline LBM, SPPB score and ability to walk outdoors were entered in the models at the same time as predictors. The SPPB scores were recoded into a single binary variable: 0) high performance (score ≥ 7) or 1) low performance (score < 7). A score below 7 indicates a high risk for disability [27]. The categories of perceived difficulty in walking outdoors were recoded as 0) major difficulties or unable (categories 3–5), or 1) no difficulties or minor difficulties (categories 1–2). Predictive mean matching of the ‘mice’ package [28] in the R programming environment was used to impute missing values in LBM for three subjects. The models were adjusted for potential confounders: age, gender, surgical procedure (internal fixation vs hemiarthroplasty vs total arthroplasty), number of chronic diseases and use of bisphosphonate medication (yes/no) at baseline. The study group was not included in the models, as no differences were observed in baseline characteristics between the groups and the intervention had no effect on bone properties [10]. The main reasons for missing bone data were inability to perform the measurements, inaccurate positioning of the leg, a technically invalid pQCT scan, substantial movement artifacts, and metal in tissues in the scanned region. For the distal tibia, a total of 154 valid scans were obtained at baseline, and 130 at 12 months. For the midshaft site, the corresponding numbers were 156 and 130. Descriptive analyses were performed using SPSS 24.0 software (IBM, NY, USA) and the robust linear regression models were analyzed using R version 3.5.1 (R core team, Vienna, Austria) with the significance level set at 5%. The study power, calculated for the main outcome, mobility limitation, was 78%.