The Chronic Arterial Disease, quality of life, and mortality in chronic KIDney injury (CADKID) study is a prospective follow-up study protocol assessing arterial disease, quality of life and mortality in patients with CKD stage G4–5. A total of 210 consecutive patients referred to the predialysis outpatient clinic of Turku University Hospital were recruited to the main study protocol between August 2013 and September 2017. The present study is a post hoc report from the CADKID study. A total of two hundred and ten patients were recruited from the predialysis outpatient clinic in the Kidney Center at Turku University Hospital. Inclusion criteria were estimated glomerular filtration rate (eGFR) < 30 ml/min per 1.73m2 using the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation and age > 18 years.

Medications, relevant medical history, and documented fractures were gathered by the researchers from the electronic patient records of the research hospital. The fractures were divided into low and high energy fractures. Falling from height (> 1 m), road traffic accidents, and injuries due to external violence were considered high energy fractures, while other injuries such as falling from ground level were considered low energy fractures. Information was gathered during a follow-up time of 5 years from the time of recruitment. The study participants were followed up every 3–4 months, including laboratory parameters (analyzed by the laboratory (TYKSLAB) of the research hospital) and clinical status as per clinical guidelines [12]. All study patients resided in the catchment area of the research hospital and all incident fractures are referred to the research hospital for assessment and care. Dual-energy X-ray absorptiometry (DXA) was not part of the study protocol and performed as per the discretion of the attending clinician.

Assessment of Cardiovascular HealthAbdominal Aortic Calcification (AAC) Score

Lateral lumbar radiography was performed using standard radiographic equipment. A validated score system by Kauppila et al. [13] was used to grade the calcification of the abdominal aorta. In this score system, the anterior and posterior aspects of the abdominal aorta in the region corresponding to the 1st to the 4th lumbar vertebra were analyzed in a total of eight segments. The degree of calcification per segment is scored between 0 and 3, so that the total score range is from 0 to 24. Individual aortic segments both from the anterior and the posterior walls were summed to get the anteroposterior severity score (0—24). All radiographs were analyzed independently by two researchers. The mean score was used in the analysis.

Maximal Stress Ergometry

All study participants underwent maximal bicycle stress ergometry, if they were in physical condition to perform the test. Altogether, 174 (82.9%) patients performed an incremental, symptom-limited cycling exercise test. The target speed was 60 rpm and it was reached after a warm-up phase of 30 s. The participants were encouraged to continue the test until exhaustion. An individual primary test workload was determined based on an estimated maximum workload (using as reference datasets of the Mini Suomi study [14]) and duration of the test workload increase/min (10, 15, or 20W) was accomplished automatically by the ergometer software. The percentage of the mean workload of the last 4 min of exercise (wlast4%) was obtained. Also heart rate and maximal oxygen uptake for all patients were obtained.

Echocardiography

The echocardiographic measures were obtained at baseline from a standardized transthoracic echocardiography performed by the Department of Clinical Physiology of Turku University Hospital before the stress ergometry. A commercially available ultrasound system (Vivid E9; GE Vingmed Ultrasound, Horten, Norway) with a 3.5-MHz phased-array transducer (M5S) was used. The systolic and diastolic function and dimensions of the left ventricle were measured, as well as left ventricular wall thicknesses, the left ventricular mass index, and left ventricular ejection fraction (EF).

Ultrasound Assessment of Brachial Artery Flow-Mediated Dilatation and IMT

Ultrasound methods employed and their reproducibility have been previously published in detail [15]. Briefly, for IMT measurements, the left common carotid artery ca. 1 cm proximal to the carotid bulb was scanned using B-mode and several images were saved digitally for later blinded offline analysis. To assess brachial flow-mediated dilatation (FMD), a measure of endothelial function, the right brachial artery diameter was measured both at rest and during the reactive hyperemia induced by inflation of a pneumatic tourniquet placed around the forearm, followed by release. The increase in vessel diameter after reactive hyperemia was expressed as the percentage relative to the resting scan.

As the focus of the present study was to assess the association between cardiovascular health and fracture risk, a composite risk score (3–12 points) was formed by summarizing TnT, ProBNP, and Wlast4% coded into four subgroups (1–4 points each) with a lower score indicating lower TnT or ProBNP or better ergometry test performance and a higher score indicating higher TnT or ProBNP or poorer ergometry test performance or the lack of ergometry test assessment due to poor physical condition.

Statistical Analysis

The results for categorical covariates were reported as absolute and relative (percentage) frequencies and continuous covariates as mean ± standard deviation (SD) or median [inter-quartile range (IQR)] for normally distributed or non-parametric, respectively. Kolmogorov–Smirnov and Shapiro–Wilk tests were used to test normality in continuous covariates.

Categorical covariates were compared using the Fisher’s exact or Pearson × 2 test. The comparisons of continuous covariates were performed using Student’s T test for normally distributed covariates and Mann–Whitney U test for skewed covariates.

The associations between covariates of interest and incident fractures within 5 years of follow-up were explored separately with univariate Cox proportional hazards models. All covariates with an association at p < 0.05 level were entered in respective multivariable Cox proportional hazards models. The multivariable Cox models were performed by including age, gender and eGFR as covariates together with a single variable of interest (e.g., TnT or proBNP or urea etc.) in each respective model (Supplementary Table S1). With a moderate number of events (patients with fractures) in the cohort, adjustment for multiple covariates in the same multivariable model could not be done as including these variables in the models would have led to increased risk of multicollinearity.

To further explore the association between cardiovascular biomarkers and incident fracture risk, TnT and ProBNP were divided into quartiles (coded as one for the lowest quartile and four as the highest quartile of measurements) and Wlast4% measurements were divided into thirds. The patients who did not undergo the ergometry test formed the fourth Wlast4% subgroup. Next, the three subgroups (thirds) of patients with available Wlast4% measurements and the subgroup without ergometry test measurements were coded into four groups as follows: one as the highest third, two as the middle third, three as the lowest third of the available Wlast4% measurements, and four as the patients who did not undergo the ergometry test. Finally, the coded quartiles of TnT and ProBNP as well as the coded four subgroups of Wlast4% were summed together to form a composite risk score (3–12 points) covariate with the lowest value of three indicating patients with the Wlast4% performance in the highest third and TnT and ProBNP, both, in the lowest quartiles and the highest value of 12 indicating patients with no available Wlast4% measurement and TnT and ProBNP, both, in the highest quartiles.

The association between 5-year fracture risk and the composite risk score covariate was explored in a univariate Cox proportional hazards analysis and a multivariable Cox model adjusted with age, gender, and prior diagnosis of coronary artery disease (CAD) as covariates. The association between fracture risk and the composite risk score was further tested with a Kaplan–Meier plot and log rank test by dividing the composite risk score into four groups (one coded as the lowest score of three; two coded as the score 4–6; three coded as the score 7–9; and four coded as the score 10–12).

Finally, the association between incident fractures and all-cause mortality within 5 years of follow-up was assessed in a univariate Cox proportional hazards analysis.

All analyses were two-sided and p < 0.05 was considered significant. IBM SPSS statistics software version 27.0 was used for all analyses.