We performed a cross-sectional analysis based on participants who were prospectively enrolled in the Copenhagen Chronic Kidney Disease Echocardiographic (CPH-CKD ECHO) Study. The study has previously been described in detail [12, 13]. Briefly, the CPH-CKD ECHO Study was a dual-centre study that comprised patients with non-dialysis-dependent CKD (n, 825) and age- and sex-matched control individuals (n, 175). Patients with CKD were enrolled from the outpatient clinic at the Departments of Nephrology at Rigshospitalet and Herlev-Gentofte Hospital in Denmark. Controls were recruited through posts in local newspapers and on a Danish website (forsoegsperson.dk) dedicated to highlighting ongoing research studies seeking voluntary individuals. The inclusion phase ran from October 2015 through October 2018. Inclusion criteria were known CKD and age between 30 and 75 years. CKD was defined as either kidney injury (albuminuria, renal cyst, or other structural abnormalities in the kidneys) or an eGFR < 60 mL/min/1.73 m2 lasting more than 3 months [14]. Exclusion criteria were as follows: kidney transplantation with a functioning graft, pregnancy, intellectual disability, dementia or psychosis, active malignancy, and retraction of informed consent.

The controls were not allowed to have known cardiovascular disease, kidney disease, malignant disease, or chronic disease, apart from mild hypertension, thyroid disease, mild depression, or hypercholesterolemia. Subjects were excluded if they had kidney damage or an eGFR < 60 mL/min/1.73 m2.

The total study population thus included 1000 individuals. Of these, 9 were excluded because they had aortic valve stenosis since the myocardial work analysis assumes that there is no obstruction between the left ventricle and the aorta [10]. In addition, 5 were excluded because they had severe aortic regurgitation since myocardial work analysis assumes negligible diastolic pressures, which is not the case with severe aortic regurgitation [10]. Finally, 55 were excluded because the myocardial work analysis was not technically feasible, leaving 931 for final analysis (757 patients with CKD and 174 controls).

The study was approved by the regional scientific ethics committee (ID: H-3–2011-069) and the Danish Data Protection Agency (ID: 30–0840). Informed consent was obtained from all participants, and the study adhered to the 2nd Helsinki Declaration.

Data on medical history and clinical characteristics were collected at an outpatient baseline visit upon inclusion. Medical history was obtained by interview and review of hospital medical records. A physical examination was performed to measure anthropometrics, heart rate, and brachial artery blood pressure. Hypertension was defined as a systolic blood pressure > 140 mmHg, diastolic blood pressure > 90 mmHg, or the use of antihypertensive medication. Urinary samples were collected to determine the degree of albuminuria (urine albumin-creatinine ratio, UACR). Details on UACR were available in 733/757 patients. Venous blood samples were drawn to acquire plasma creatinine level and for storage in a biobank. A 12-lead electrocardiogram was also performed at the visit.

eGFR was calculated from plasma creatinine level with the CKD-EPI formula [15]. The patients were grouped into eGFR categories in accordance with the 2012 Kidney Disease Improving Global Outcomes (KDIGO) guidelines [14]. In line with our previous publications, we also collated patients into three eGFR groups as follows: G1 + G2 (eGFR ≥ 60 mL/min/1.73 m2), G3 (eGFR 30–59 mL/min/1.73 m2), and G4 + G5 (eGFR ≤ 29 mL/min/1.73 m2) [13].

Albuminuria was characterized as follows: normoalbuminuria with UACR < 30 mg/g, microalbuminuria with UACR of 30–300 mg/g, and macroalbuminuria with UACR > 300 mg/g.

All echocardiographic examinations were performed using a GE Vivid E9 ultrasound machine according to a dedicated protocol. All images were acquired over three consecutive cycles. For individuals in sinus rhythm, measurements were performed in a single cardiac cycle — the one with the most optimal image quality — whereas all measurements were performed in three cardiac cycles for patients who had atrial fibrillation during the examination. All standard measures were performed with commercially available post-processing software (EchoPAC BT 203, GE Healthcare) by a single experienced investigator blinded to clinical information according to the 2015 American Society of Echocardiography/European Association of Cardiovascular Imaging (ASE/EACVI) guidelines [16]. The analysis process has been described meticulously elsewhere [13].

Valve disease was quantified according to the most recent guidelines [17, 18]. Significant valve disease was defined as either moderate aortic valve regurgitation, moderate or severe mitral regurgitation, or moderate or severe mitral stenosis.

Analyses of pressure-strain loops were performed according to published directions [19]. The analysis process and definition of work parameters are shown in Fig. 1. Pressure-strain loops were acquired by first analyzing myocardial speckle tracking in the three apical projections (minimum frame rate of 40 frames per second, mean ± SD, 58 ± 5 frames per second). The left ventricular myocardium was automatically traced using automated function imaging. This created a region of interest that covered the endocardial throughout the myo-epicardial layer. The tracing and region of interest could be adjusted at the discretion of the investigator. Segments could be excluded if the tracing did not follow the myocardial speckles adequately, however, only one segment in total could be excluded, otherwise, the analysis was considered infeasible. Speckle tracking analysis was feasible in 941 (94%). Pressure-strain loops were then created by inputting blood pressure and visually estimating valvular event timing. The following myocardial work parameters were derived from the pressure-strain loop analyses: global work index (GWI), global work efficiency (GWE), global constructive work (GCW), and global wasted work (GWW).

The pressure-strain loop analysis process used to obtain myocardial work measures. First, left ventricular speckle tracking is performed (top left panel), then the brachial artery blood pressure is added, and then the timing of valvular opening and closure is visually estimated (top right panel with orange arrow at the mitral valve and green arrow at the aortic valve). The results are depicted in the bottom left panel and include a pressure-strain loop with the area reflecting global myocardial work index, the bulls-eye plot shows segmental values of myocardial work, and bar charts show the relative distribution of constructive and wasted work. The bottom right panel presents definitions of the four work measures derived from the pressure-strain loop analysis 

Abnormal work indices were defined as follows according to published reference material [20]: GWI < 1576 mmHg%, GCW < 1708 mmHg%, GWW > 159 mmHg%, GWE < 93.0%.

In line with our previous publication, GLS below 18% (numerical value) was considered abnormal [13].

Clinical and echocardiographic characteristics were compared for the patients with CKD stratified by normal vs. abnormal GWI. Furthermore, myocardial work measures were compared between patients with CKD and controls. For these comparisons, Gaussian-distributed continuous variables were analyzed with Student’s T-test and reported as mean with standard deviation. Non-Gaussian distributed variables were compared with the Wilcoxon rank-sum test and reported as median with interquartile ranges. Gaussian distribution was assessed from histograms. Categorical variables were compared with either the chi2 test or Fisher’s exact test as appropriate and reported as total numbers with percentages.

Linear multivariable regressions were made to account for confounders between patients with CKD and controls and calculate predicted means. Adjustments were made for body mass index, diabetes, hypertension, smoking status, alcohol consumption, eGFR, UACR, heart rate, left bundle branch block, atrial fibrillation, heart failure, ischemic heart disease, significant valve disease, and left ventricular mass index. For all regression analyses involving work measures, the GWW and GWE variables were log- and logit-transformed, respectively.

Comparisons were further made across groups of left ventricular remodeling (controls vs. patients with CKD without and with left ventricular hypertrophy (LVH), respectively), across eGFR groups (G1, G2, G3, G4, G5), strata of albuminuria (normoalbuminuria, microalbuminuria, and macroalbuminuria). For these comparisons, the ANOVA test was applied for Gaussian-distributed variables, and non-Gaussian distributed variables were compared with a non-parametric trend test.

Linear regression analysis was also applied to examine the association between eGFR and myocardial work measures and between UACR and myocardial work measures. For all regression analyses, UACR underwent a log-transformation. Multivariable adjustments were made for relevant confounders: age, sex, hypertension, diabetes, heart rate, significant valve disease, ischemic heart disease, smoking status, alcohol consumption, body mass index, known heart failure, and left ventricular mass index. For the analyses concerning the association between UACR and work measures, the multivariable model also included adjustment for eGFR and vice versa. The same analyses were carried out in a subgroup of patients with CKD who exhibited current signs of functional kidney disease (defined as either reduced eGFR (< 60 mL/min/1.73 m2) or albuminuria (UACR > 30 mg/g)).

Logistic regression was applied to investigate which stages of CKD were associated with increased likelihood of abnormal work. Multivariable adjustments were similar to the linear regressions.

Tests for interactions from diabetes, hypertension, and CKD etiology, respectively, were applied in both linear and logistic regression analysis. Since diabetes significantly modified the association between eGFR and work measures, these regression analyses were stratified by diabetes status.

All statistical analyses were performed using STATA v. 15 SE (StataCorp LP, College Station, TX). p-values < 0.05 were considered significant in all analyses.