A cross-sectional study was conducted in patients with a clinical vertebral or non-vertebral fracture attending the Fracture liaison Service (FLS) in VieCuri Medical Center located in the south of the Netherlands from November 2013 to June 2016. Subsequent fracture risk evaluation includes clinical evaluation, laboratory tests, medication review, vertebral fracture assessment (VFA), bone mineral density (BMD) measurement using DXA, and ideally also a fall risk assessment [6, 18]. All adults aged 50–90 that visited the emergency department with a radiologically confirmed fracture were screened for fracture risk evaluation by trained nurses. Those eligible for fracture risk evaluation were invited to attend the FLS as part of usual care. Ineligible patients were persons living outside of the region, with a low life expectancy (less than a year), or already under care for osteoporotic or oncological care. Further, patients with facial and skull fractures, prosthetic failure, pathological fracture, or osteomyelitis were not invited. For this study, patients with missing values on weight and height, or missing values on all functional performance tests (CST, HGS, TUG, and 6MWT) due to logistical issues (e.g., issues around availability of equipment, appointment scheduling, or lack of time or personnel) or unknown causes were excluded. Patients who were unable to perform these tests due to physical inability were included. Patients were evaluated and treated according to the Dutch guidelines for osteoporosis and fracture prevention [19].

Age at the time of fracture, weight (kg), height (cm), and BMI (kg/m2) were measured during the FLS visit. During the FLS visit, the patients were asked to complete a questionnaire containing information about medical history and medication use, residential status, cause of the fracture, and information on previous fractures. Furthermore, medical history was extracted from medical records of the emergency department visit. Comorbidities were classified according to the tenth revision of International Classification of Disease (ICD-10). Dual-energy X-ray absorptiometry (DXA, Hologic Inc, Bedford, MA, USA), was used to measure bone mineral density (BMD) and measurements of body composition. Osteoporosis was defined as bone mineral density T-score ≤  − 2.5, osteopenia as T-score between − 1.0 and − 2.5, and a normal BMD as a T-score ≥ 1.0 at the hip, femoral neck, or lumbar spine [20]. Prevalent vertebral fractures (VFs) were assessed on lateral spine images acquired with DXA. The grading of VFs was done morphometrically using the classification of Genant [21], based on percentage height loss and were categorized according to the most severe VF, as follows; grade 1 (20 to 25%), grade 2 (> 25 to 40%), or grade 3 (height loss > 40%). Appendicular lean mass (ALM) was calculated as the sum of lean tissue in the arms and legs (kg) [22] and was corrected for squared height [14].

Physical functioning tests at baseline included CST, HGS, TUG, and 6MWT. All tests have a good to excellent inter-rater reliability in older populations with and without comorbidities (supplemental Table 1). All tests were conducted by trained nurses. Lower body strength was assessed by 30-s CST [23]. Participants were instructed to start in seated position, to not use the armrests, cross the arms over the chest, stand up, and fully sit back down in-between stands. The measured outcome was the number of times a person can fully stand up from a chair in 30 s time. HGS was measured by handheld dynamometer (JAMAR, Sammons Preston, Bolingbrook, Illinois). HGS testing was performed in seated position, with the elbows flexed at 90°. The maximum handgrip strength from three attempts for the left hand and three attempts for the right hand was used for analysis and was defined as the best score out the six attempts. TUG was used to measure balance, walking ability and overall mobility [24]. Participants were asked to rise from a chair with armrests, walk for three meters, turn, walk back to the chair, and sit down. The use of walking aids was permitted. The measured outcome was time to perform TUG in seconds. The TUG was performed three times and the mean score was calculated. If patients could only perform the test one or two times, the mean of those scores was used. Lastly, 6MWT was used to measure walking ability and functional exercise capacity [25]. Patients were asked to walk up and down between the safety cones on either end of a 10 m level linoleum hallway, at a comfortable speed, while covering as much distance as possible. Use of walking aid was permitted and patients were allowed to rest or stop when needed. No encouragements were given during the test. Patients performed the test one time, and walking distance in meters (6MWD) was scored. Confirmed sarcopenia was defined according to the EWGSOP2 criteria, as having either low HGS or low CST scores, and low ALM, e.g., “confirmed sarcopenia” [14]. “Probable sarcopenia” was defined as having either low CST or low HGS scores. The guideline’s cut-off scores for HGS are < 16 kg for females and < 27 kg for males and of CST > 15 s for five rises. As the 30-s CST was performed in this study, the cut-off score for low performance on CST was adapted to < 10 stands in 30 s. The guideline defines low ALM as < 7.0 kg/m2 for males and < 5.5 kg/m2 for females. Osteosarcopenia was defined as having “confirmed sarcopenia,” combined with having osteoporosis or osteopenia [26].

Fractures were grouped in three ways; first, according to the center classification into four different fracture categories: 1. hip fractures, 2. major fractures (vertebra, multiple rib, humerus, pelvis, distal femur, and proximal tibia), 3. minor fractures; all remaining except fingers and toes which were grouped separately in group 4 [2]. Multiple fractures were allocated according to the most severe fracture. The center classification groups’ fractures based on mortality risk after the fracture. Second, fractures were grouped according to the most important osteoporotic fractures of the IOF classification: major osteoporotic fracture (MOF) including lower arm, hip, humerus, clinical vertebrae fractures, and non-MOF fractures (all other fractures). The third grouping covered all lower extremity fractures (any fracture from the pelvis down) versus all upper body fractures (all other fractures) and was chosen to differentiate if proximity of the fracture is of influence on the physical performance.

To adequately assess physical performance of the FLS patients, a comparison to a healthy age- and sex-related reference population using Z-scores was chosen, comparable with the use of T-scores in osteoporosis measurements. While many test cut-off points for poor physical performance are proposed in the literature, some methodological issues could lead, when used, to a biased overview of poor physical performance in a population. First, a wide variety of different cut-off scores exist for all tests without established consensus on the optimal cut-off point. Cut-off scores are developed based on different outcomes for poor performance (e.g., the TUG cut-off score for is falls 15 s, for fractures 13.5 s, and sarcopenia 20 s [14, 15, 27]. Last, most are a “one-size fits all” cut-off score and not stratified to age and sex [14, 15, 27], which could lead to over- or underestimation of test results in specific groups. To secure optimal reference data of the general population, we carried out a semi-structured literature search in March 2022. The search strategy incorporated a combination of “Medical Subject Headings” or “Title/Abstract” terms describing “reference values” and physical performance test, e.g., “TUG, 6MWT, CST or HGS.” The search strategy, outcome and rationale of the literature search is presented in Supplementary Table 1 and 2. For the TUG we applied the reference data from Svinøy et al. [28], a large (N = 5400) Norwegian cross-sectional study including community-dwelling older persons aged 60–84 and published mean and SD’s that were comparable to other study in a Western population of Kenny et al. [29]. Results of Kenny et al. for the age categories 50–60 were studied and it was decided that the reference values of Svinøy et al. for 60 year olds could be extrapolated to 50–60 year olds. For 30CST and 6MWT, reference data was provided by Rikli and Jones et al. including 7183 community dwelling adults from the United States aged 60–94 [30]. The reference data from Rikli and Jones dates from 1999. However, reference data for the 30 s CST were particularly scarce and the testing method of Rikli and Jones was completely identical to ours as they were the first to develop this test. CST outcome of Rikli and Jones were compared to a more recent German cohort study of Albrecht et al. that did not meet inclusion criteria of the age range and results between the two studies were largely comparable [31]. Based on a comparison with other literature [32], it was decided that the reference values of Rikli and Jones et al. for 60 year olds could be extrapolated to 50–60 year olds. For the 6WMT, not all inclusion criteria were met; no study had identical testing methods except for Beekman et al. [33]; however, they did not present results stratified to age and sex groups. Means and SD of Rikli and Jones were used and their sex-stratified outcomes were compared to Beekman et al. and Casanova et al. [34], both of the latter reported higher means in both sexes. Thus, possible low performance of the FLS population might be even higher if other reference populations were used [30]. Finally, for the HGS, the reference data were derived from the meta-analysis of Dodds et al. which included data on HGS of 49,964 persons from 12 general population studies from across the United Kingdom [35], with similar testing methods and high comparability with reference data from other developed countries.

In the descriptive analysis mean and SD were calculated for normal distributed data. For non-normally distributed data medians and interquartile ranges were presented. Normality was visually assessed, and due to the large sample size approximately normal distributions were accepted. Patient specific Z-scores were calculated for each test, using the age- and sex-specific means and SD derived from the literature. The mean Z-scores were tested against the theoretical expected value of 0 of the reference population (the healthy age- and sex-related peers) using one-sample T-tests. Fracture group-specific boxplots of the Z-scores were created, according to center classification, stratified to sex and age. All categories shown have > 5 persons included in the analysis. Impaired and poor performance were defined as scoring 1SD and 2SD below age and gender norms, respectively, and proportions of impaired and poor performance were calculated. Patients who were physically unable to perform these tests were included in the poor performance group (> 2SD deviance from normative expectations). Uni- and multivariate linear regression models were used to investigate the association between the different fracture groupings and the Z-scores of the different performance tests. Following groups were tested: center major and hip fractures (with Center Minor as reference group), IOF MOF (non-MOF as reference group), lower extremity (upper extremity fractures as reference group), prevalent VF, excluding all patients with a clinical vertebral fracture (no prevalent VF as reference group). Multivariable analyses were adjusted for age, sex, and time since fracture. As a sensitivity analysis the descriptive analysis of the performance tests, comparison of Z-scores using T-tests and the proportions of impaired and poor performance were also calculated for the population excluding finger and toe fractures, as these fractures are often not included in FLS populations. A P value of < 0.05 was considered statistically significant. Statistical analyses were run in SPSS (IBM SPSS Statistics 28).