The Osteoporosis Prospective Risk Assessment (OPRA) cohort, previously described (20), is a population-based study, originally designed to study fracture. During 1995–1999 women, being 75 years old at point of invitation, were randomly and without exclusion criteria, selected from the city archives of Malmö and invited by letter to participate. Of the 1604 women approached, 152 stated illness and 376 unwillingness as cause for non-participation, while 32 could not be reached despite several attempts, resulting in 1044 women attending baseline investigation (65% response rate). Participants were followed up twice over a period of ten years. The first follow-up investigation took place after five years (age 80, n=715) and the second after ten years (age 85, n=382). The present study uses data from women with available FGF23 values, corresponding to 995, 667 and 324 at respective visits. Missing values reflect a random loss across the cohort (for example, lack of serum, inability to provide a blood sample, haemolysis or failed analysis) and not a systematic error or selection. The study was approved by the Regional Ethics Committee in Lund (Dnr: 2014804) and in accordance with the Helsinki declaration. All participants provided written informed consent.

Non-fasting blood samples were collected between 8 a.m. and 1 p.m, centrifuged and stored at −80°C. Routine blood chemistry was analysed by standard methods at the time of the study (21). Kidney function (estimated glomerular filtration rate, eGFR) was based on cystatin C, using the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) study equation (22).

Serum levels of FGF23 were measured using the C-terminal assay, which identifies 2 epitopes on the C-terminus part of the hormone, capturing both intact FGF23 and the C-terminal fragments of cleaved and inactivated hormone (23). Samples from all visits (age 75, 80 and 85) were analysed in batch in 2016 using ELISA (rev.no. 150407, Biomedica Medizinprodukte GmbH & Co KG, A-1210 Wien, Divischgasse 4). Intra-assay CVs were: ≤12% at 6 pmol/L and ≤6% at 10 pmol/L. Inter-assay CVs were: ≤10% at 6 pmol/L and ≤5% at 10 pmol/L.

Muscle mass was measured by dual-energy x-ray absorptiometry (DXA) using the same machine (GE Lunar, Madison, WI) throughout the study. Appendicular skeletal lean mass (ASM) was adjusted for body size by dividing with height squared (ASM/height2, kg/m2). The threshold for low muscle mass was <5.5 kg/m2.

Muscle strength was measured using knee strength, as hand grip strength (HGS) was not collected at all visits and isometric torque methods are suggested as an alternative by the European Working Group on Sarcopenia in Older People (EWGSOP2) (24). At all visits knee extension was tested using a computerized isokinetic dynamometer (Biodex Medical Systems, v4.5.0, Biodex Corporation, New York). The best out of three maximum isometric contractions (knee extension at 90°) lasting 5 seconds each was recorded (Nms). We identified a threshold for low knee strength equating to a HGS threshold of <16kg, previously applied in defining sarcopenic participants (17), using two steps:

Step 1: Using low gait speed (≤0.8 m/s) as the dependent variable, as suggested by Alley et al (25), and maximum repetition knee extension (Nms) as the scale dependent variable, we performed a Classification and Regression Tree (CRT) analysis to establish a clinically relevant threshold for knee strength that would closely mirror a HGS <16kg. The CRT segmented the data and identified the knee strength threshold with the strongest relationship to low gait speed. Based on this a threshold for low knee strength at ≤174.24 Nms was identified. As an additional check, we performed CRT analysis again, this time using low HGS (<16kg) as the dependent variable; the identified threshold was <175.9 Nm which is close to the cutoff derived from using the recommended low gait speed.

Step 2: To assess how well the tree structure can be generalized and to ensure selection of the best performing cutoff, a split-sample validation was also performed, in which test- and training sample groups were randomly assigned with 50% of the cohort data in each group. Based on this, a threshold at ≤174.85 Nms was identified in both segments. Taking the average from steps 1 and 2, knee strength <175 Nms was equated to be equivalent to HGS <16kg.

Sarcopenia was defined in accordance with the updated 2018 guidelines from the EWGSOP (17), as low muscle mass (i.e. ASM<5.5 kg/m2) plus low muscle strength (i.e. knee strength <175 Nms). With reference to the same guidelines, a group of women with probable sarcopenia was defined based on muscle strength alone (i.e. knee strength <175 Nms) without reference to muscle mass.

Muscle function. Gait speed (m/s) was tested at all visits through a 10m walk with one turn (i.e. 2×5m); a higher value indicating a poorer function. The participants wore ordinary shoes with the instruction to walk as fast as they could.

Using the principles described by Searle et al (26) and described in full elsewhere (27), a frailty index (Fl) was created using data collected at each visit (age 75, 80 and 85). Concisely, this frailty index contains thirteen variables covering several physiological domains (time spent outdoors, daily physical activity, balance, walking speed, number of steps taken, muscle strength, diabetes, cancer/severe disease, diseases affecting balance, self-reported fall risk, polypharmacy, CRP and creatinine). The index represents ‘deficits in health’, scored from 0.0–1.0; the higher the score, the more frail. The FI correlated very strongly (r = 0.80) to a full 40-variable index (28), created for the two follow-up visits, and both the 13- and 40-variable frailty index show similar capacity to predict mortality (27). The frailty index was used as a continuous variable, but also using a threshold ≥ 0.25 to define frail individuals. This empirical cutoff is supported by other studies (29, 30).

At each visit, all women completed an extensive questionnaire, from which information about falls and mobility was collected. Falls were defined as having had a fall within the previous year (yes/no). Impaired mobility was defined as using a walking aid inside or outside or moving with other people or bedbound.

Incident fractures (hip or major osteoporotic, the latter defined as any of the following: hip, vertebra, distal radius, shoulder), were followed for ten years through the X-ray files at the Radiology Department, Malmö, Skåne University Hospital. Since the department of orthopedics was the sole unit treating fractures in the catchment area, information loss was exceptionally low (31). Osteoporotic fractures defined as any of the following: hip, vertebra, distal radius, shoulder. Pathological and high energy fractures were excluded.

Descriptive fdata is presented as mean with standard deviation or median with interquartile range, as appropriate. In case of non-normally distributed data, non-parametric tests were performed. At all three visits (age 75, 80 and 85), women were stratified into tertiles of FGF23 (T1 = lowest FGF23 level; T3 = highest FGF23 level). Overall comparisons between tertiles of FGF23 with outcomes frailty, sarcopenia, muscle strength, muscle function and muscle mass, were performed using ANOVA, Kruskal-Wallis or Chi-square test as appropriate.

To investigate if women in the highest FGF23 tertile had higher odds of a) sarcopenia, b) probable sarcopenia and c) frailty, logistic regression analyses were performed at age 75, and then repeated at age 80 (based on FGF23 measurements at age 80) and 85 (based on FGF23 measurements at age 85). These analyses are presented unadjusted (model 1) and adjusted for: PTH, vitamin D, estimated kidney function (eGFR), phosphate and calcium (model 2). Additionally, to investigate if FGF23 at baseline (age 75) was associated with outcomes at follow up, longitudinal analyses were performed. Using logistic regression analyses and tertiles of FGF23 at age 75, association with sarcopenia, probable sarcopenia and frailty at age 80 and 85 was investigated. Data is presented unadjusted and adjusted for PTH, vitamin D, estimated kidney function (eGFR), phosphate and calcium at age 75. Kaplan Meier curves describe fracture free survival during the ten-year study period (i.e. from age 75–85) for hip and major osteoporotic fractures. P-value for difference was calculated using the log rank test.

All statistical analyses were performed using SPSS (IBM Corp. Released 2021. IBM SPSS Statistics for Windows, Version 28.0. Armonk, NY). A p-value <0.05 was considered nominally significant.