Patient selection

Over an 18-month period (from January 2006 to June 2007), 143 patients at various stages of CKD were recruited by the Nephrology Department at Amiens Hospital (Amiens, France). The present study was approved by the local investigational review board (Comité de Protection des Personnes Nord-Ouest II, approval number 06H3) and written, informed consent was provided from all participants, in line with the tenets of the Declaration of Helsinki.

The inclusion criteria included age over 40, close monitoring by a nephrologist, and CKD, as defined by (i) two estimated glomerular filtration rates < 90 ml/min per 1.73 m² (calculated according to Cockcroft and Gault's equation) at a 3– to 6–month interval [22] or (ii) the use of hemodialysis.

The main exclusion criteria were the presence of chronic inflammatory disease, atrial fibrillation, complete heart block, abdominal aorta aneurysm, an aortic and/or femoral artery prosthesis, primary hyperparathyroidism, kidney transplantation or any acute CV event in the three months prior to screening.

Study protocol

One hundred and forty-three patients were day-hospitalized for research purposes: laboratory blood tests, blood pressure measurements, IMT, PWV, and ABI determinations, a lateral lumbar X-ray and a multislice spiral computed tomography (CT) scan (to determine CV calcification).

Hemodialysis patients were examined on a dialysis-free day or, if this was not possible, the morning before a dialysis session. These patients underwent a four-hour, standard hemodialysis session three times a week. A patient interview focused on comorbidities, the disease history (especially previous CV events), concomitant medications, and TRFs (age, gender, diabetes, triglyceride levels, hypertension, and smoking status). The patient’s medical files were reviewed in order to gather more exhaustive information. Previous CVD was defined as a history of any of the following events in the patient's medical records: myocardial infarction, stroke, heart failure, angina pectoris or surgical procedures for angina or coronary/peripheral artery disease (including percutaneous transluminal angioplasty). Hypertension was defined as treatment with at least one antihypertensive agent. Diabetes mellitus was defined as treatment with insulin or oral antidiabetic medications.

Laboratory tests

Blood samples were collected on the morning of the study day (before any other investigations), prepared and then stored at -80°C. All assays were performed on subsequently thawed plasma or serum samples.

Serum total calcium, phosphate, albumin, LDL cholesterol, hemoglobin, creatinine and C-reactive protein levels were measured in an on-site biochemistry laboratory with standard auto-analyzer techniques (the Modular IIP system from Roche Diagnostics, Basel, Switzerland). The laboratory's normative range of phosphate concentrations was 0.8–1.45 mmol/l. Serum intact parathyroid hormone and 25 hydroxyvitamin D levels were assayed with a chemiluminometric immunoassay (the Liaison N-tact PTH CLIA and the Liaison 25 OH vitamin D TOTAL CLIA [which measures both D2 and D3], Diasorin, Stillwater, MN). Serum 1.25 (OH)₂ vitamin D levels were determined in an RIA (¹²⁵I RIA, Diasorin). Serum cystatin C (Cys C) levels were determined by immunonephelometry (N latex Cystatin C, Siemens, Erlangen, Germany) and serum creatinine (Scr) levels were measured using the Jaffe method based on standard isotope-dilution mass spectrometry (IDMS). The Chronic Kidney Disease Epidemiology Collaboration equation was then used to calculate a more accurate estimate of the glomerular filtration rate in all non-dialyzed patients [23]: 177.6 x Scr^-0.62 x Cys C^-0.57 x age^-0.20 x (0.82 if female). Next, patients were classified into CKD stages according to the National Kidney Foundation’s Kidney Disease Outcomes Quality Initiative [23].

Evaluation of the intima media thickness

The IMT was measured semi-automatically from carotid ultrasound data using dedicated software (M'Ath, Intelligence in Medical Technologies SARL, Paris, France). All measurements were performed by the same sonographer. Using an automated edge detection algorithm, the IMT was measured along a 10 mm-long segment of the far wall of the common carotid artery (CCA, at a site free of any discrete plaques) and corresponded to the distance between the lumen-intima interface and the media-adventitia interface. A mean of 50 measurements were automatically performed on each image (two images per body side) and on each body side (left and right). The greatest measured values for the right and left CCA-IMTs were used in the analysis. Data on reproducibility have been previously published by our group; the correlation coefficient for repeated readings of the CCA-IMT was 0.96 [24].

Evaluation of the PWV

The PWV between the carotid and femoral arteries was determined transcutaneously by trained physicians, using an automatic, dedicated, validated device (Complior Colson, Createch Industrie, Massy, France), as previously described [16, 24]. The velocity was calculated as the distance L between the two recording sites (measured along the body surface), divided by the time interval t between the feet of the flow waves (i.e. PWV = L/t) [25]. To avoid possible methodological bias, a single senior investigator (MT) performed the PWV measurements. All measurements were performed twice (to confirm reproducibility) and the mean value was subsequently analyzed. Data on the validation of this automatic method and its reproducibility have been published previously; the intra-observer repeatability coefficient was 0.93 and the interobserver reproducibility coefficient was 0.89 [25].

Calculation of the ankle-brachial index

The ABI was calculated after the participant had rested in the supine position for 5 minutes. A hand-held Doppler probe was used to measure the systolic pressures at the right brachial artery, right posterior tibial artery, left posterior tibial artery and left brachial artery (in that order). According to a previously validated method, the ABI was calculated as the ratio of brachial to ankle artery pressure for each leg. The lower of the two values was subsequently analyzed (unless the ABI was 1.3 or more, in which case the higher value was selected).[26]

Abdominal aorta imaging with plain radiography

A technique similar to that described by Kauppila et al. [27] was used to obtain images of the lower abdominal aorta and thus determine the aortic calcification score. On a plain radiograph of the abdomen (in the sagittal plane), the aorta was identified as a tubular structure passing in front of the anterior surface of the spine. We used a semi-quantitative scoring system; only the abdominal aorta segments in front of the first to fourth lumbar vertebrae were considered. Points were assigned on a 0–3 scale (0: absence; 1: small; 2: moderate; 3: large), according to the length of each calcified plaque identified along the length of the lumbar vertebrae under consideration. Hence, the score could vary from 0 (no calcification) to 24 (maximal calcification). The score was noted independently by two investigators (SL and LD). In the event of disagreement between the scorers, a consensus was reached for all radiographs. To validate the reproducibility of our vascular calcification scoring, 73 randomly selected radiographies were scored blindly by the two independent readers. The calcification scores were compared by calculating Pearson's correlation coefficient (r = 0.925, p = 0.01); this value demonstrates that the calcification scores were correctly assessed on the basis of our chosen criteria.

Multislice spiral CT

In order to quantify the presence and extent of aortic and coronary artery calcifications, each patient underwent a multislice spiral CT scan. All examinations were performed with a 64-detector CT scanner (Lightspeed VCT, GE Healthcare, Milwaukee, WI). Detailed technical information on the procedure has been provided elsewhere[16]. Briefly, the total aortic calcification volume was obtained by calculating the sum of all voxels in the remaining volume after thresholding. The abdominal aorta calcification score was calculated as follows: [(total calcification volume) / (aorta wall surface area) x 100]. The Agatston score [28] was used to quantify coronary artery calcification.

Follow-up

Patients were followed up prospectively for the occurrence of non-fatal CV events or all-cause death (whichever occurred first). A total of 64 events (composite endpoints) were analyzed. Cardiovascular-related hospitalizations and deaths were collected prospectively every six months and adjudicated by two nephrologists. In each case, the patient's medical records were reviewed and the cause of death or the CV event was assigned on the basis of all the available clinical information. For out-of-hospital deaths, the patient’s family doctor was interviewed to obtain pertinent information on the cause. The following criteria were used for the adjudication: angina pectoris was defined as symptoms with objective evidence of ischemia (electrocardiogram or coronary artery angiography); adjudication as congestive heart failure had to be supported by a physician's diagnosis or treatment for heart failure, pulmonary congestion or cardiac dysfunction.

Cardiovascular mortality was defined as any death directly to CV system dysfunction (stroke, myocardial infarction, congestive heart failure or sudden death). Non-fatal CV events were defined as systemic CV dysfunction (stroke, angina pectoris, myocardial infarction, congestive cardiac failure, peripheral ischemia or new-onset arrhythmia) or surgical procedures for angina or coronary/peripheral artery disease. When several events occurred in the same patient, only the time to the first event was considered in our analysis.

Statistical analysis

Data were expressed as the mean ± SD, median and range or frequency, as appropriate. In the study population, patients were stratified according to the presence or absence of events. Intergroup comparisons were performed with a chi-squared test for categorical variables and Student’s t test or the Mann-Whitney test for continuous variables. The Kaplan-Meier method was used to estimate percentage survival as a function of the presence or absence of the composite factor (a CV event or death). Survival curves were compared in a log-rank test. Putative associations between clinical factors, biological factors and mortality were assessed in univariate and multivariate Cox models. The proportional hazard assumption was checked with the Kolmogorov-type supremum test. In the multivariate analysis, the initial model included TRFs only: age, gender, diabetes, triglyceride level, hypertension, and smoking status. In the additional models (Models 2 to 7), the TRFs were supplemented with the various non-invasive measurements. The discriminant ability of the seven Cox models was assessed with Harrell’s C-index [29], which estimates the proportion of correct predictions (i.e. the proportions of patients with better staging and better survival). The C-index varies from 0.5 (no discrimination) to 1 (perfect discrimination). The net reclassification improvement (NRI) was estimated for predefined risk categories of composite criteria (<10%, 10%-19% and ≥20%), based on existing clinical thresholds for cardiovascular disease [30]. Lastly, the integrated discrimination improvement (IDI) was calculated [31]. The NRI assess the putative clinical utility of different non-invasive measures (in addition to TRFs) as a marker of CV event occurrence or mortality. The IDI can be viewed as a continuous version of the NRI and does not require predefined risk categories; only the predicted probability of an event is used in the reclassification. According to a receiver operating characteristic curve built for the composite endpoint, the best cutoff for the coronary artery calcification score was 259.5 AU (with a sensibility of 0.71 and a specificity of 0.73) and the best cutoff for the aortic calcification score was 1.32% (with a sensibility of 0.735 and a specificity of 0.61).

The threshold for statistical significance was set to p≤0.05. All statistical analyses were performed using SPSS software (version 18.0, SPSS Inc., Chicago, IL) for Windows (Microsoft Corp, Redmond, WA) and SAS software (version 9.2, SAS Institute Inc., Cary, NC).