Study design:
The DIACART (for “DIAbète et Calcification ARTerielle”) study is a prospective monocentric cohort study(4). The recruitment period extended from February to October 2014. Inclusion criteria were: Type 2 Diabetes with at least a history of coronary artery disease and/or peripheral arterial occlusive disease and/or age > 50 years for men and > 60 years for women. Exclusion criteria were: (1) an estimated glomerular filtration rate (GFR) calculated by the modification of diet in renal disease equation < 30 ml/min, (2) a history of lower limbs angioplasty and/or bypass, or (3) type 1 diabetes. The study was approved by the local ethics committee and registered in ClinicalTrials.gov (Identifier: NCT02431234). All patients were informed of the study objectives and procedures. Participants gave their written informed consent of participation prior to inclusion.
Study follow-up
At the inclusion visit and after a mean follow up of 31.2+/-3.9 months (median 30.5 months and range 26-44.8 months), all patients had a clinical evaluation, laboratory blood tests, and helicoidal CT (computerized tomography) scans. Patient interviews focused on comorbidities and their personal disease history. Their medical records were reviewed to check the clinical information and to record concomitant treatments. Ischemic cardiomyopathy was defined as a history of myocardial infarction, acute coronary syndrome or any surgical procedures undergone due to coronary artery disease.
Assessment of peripheral neuropathy
The physician who performed the clinical tests was not informed of the calcification score nor the laboratory results. Peripheral neuropathy was assessed by the Neuropathy Disability Score (NDS)(16). It assesses sensory vibration of the big toe using a 128-Hz tuning fork, temperature sense on the dorsum of the foot using a tuning fork with a beaker of ice or warm water, pinprick sense by applying a pin near the nail of the big toe and the Achilles reflex test. Each sensory test scores 0 for normal and 1 for abnormal sensation, on each foot. Achilles reflexes score 0 if they are present, 1 if they are present with reinforcement and 2 if they are absent, for each foot. An NDS ≥ 6 leads to the diagnosis of diabetic peripheral neuropathy.
Laboratory evaluations
Blood and urine samples were collected after an overnight fast for the measurement of routine biochemistry diagnostic tests, glucose, glycated hemoglobin (HbA1c), cholesterol (total and HDL), triglycerides, high-sensitivity C-reactive protein (hsCRP), calcium, phosphate, urine albumin and creatinin.
Total OPG was measured by ELISA (ELISA MicroVue; Quidel Corporation), and soluble RANKL (sRANKL) concentrations were evaluated using the human RANKL Single Plex kit from Millipore (Ref: HBN51K1RANKL; eBioscienceo). With this kit, the minimum detectable concentration of sRANKL is 4.8 pg/mL and inter- and intra-assay precision is below 6% coefficient of variation. The OPG to sRANKL concentration ratio was calculated for each patient without any postanalytical modifications.
Circulating total adiponectin concentrations were measured in serum using an enzyme-linked immunosorbent assay kit (ALPCO, Eurobio, Paris, France) as recommended by the manufacturer, with the lowest detection limit of 0.4 µg/mL for total adiponectin. Inter-assay coefficients of variation of low and high human pool controls for total adiponectin were 7.93% and 8.46%, respectively.
Selected assays (including desphospho--uncarboxylated MGP (dp-ucMGP) and total-uncarboxylated MGP (t-ucMGP) assays) were performed on thawed samples, which had been frozen and stored at -80°C. A dual-antibody ELISA was used to measure dp-ucMGP levels; the capture antibody was directed against the non-phosphorylated MGP sequence 3–15 (mAb-dpMGP; VitaK BV, Maastricht, Netherlands) and the detecting antibody was directed against the uncarboxylated MGP sequence 35–49 (mAb-ucMGP; VitaK BV). The same antibodies have already been used elsewhere for immunohistochemical staining(15, 17). Intra-assay variability was 5,6% for dp-ucMGP and 8,9% for t-ucMGP, while inter-assay variability was 9,9% for dp-ucMGP and 11,4% for t-ucMGP. A competitive (single-antibody) ELISA was used to measure t-ucMGP levels, as described previously(18, 19).
Soluble RAGE (sRAGE) was measured on plasma samples with a commercially available ELISA kit (Quantikine Human RAGE Immunoassay, R&D Systems, Minneapolis, USA; see http://www.rndsystems.com/Products/DRG00 for detailed description of the measurement method).
Advanced Glycation End products (Carboxymethyllysine (CML), Methylglyoxal-derived hydroimidazolone1 (MG‐H1) and pentosidine) were determined on serum samples by liquid chromatography coupled with tandem mass spectrometry (API4000 system ABSciex, Les Ulis, France)(20). AGEs concentrations were expressed as a ratio of total protein concentrations.
Total cholesterol and triglyceride concentrations were determined by an automated enzymatic method (Konelab, Thermoclinical Labsystems, Cergy Pontoise, France and Biomerieux, Marcy L’Etoile, France, respectively). HDL-cholesterol was measured by a direct method (Konelab). LDL‑cholesterol was calculated using Friedewald’s equation on triglycerides ≤ 3.9 mmol/L or directly measured when triglycerides were > 3.9 mmol/L using a Konelab kit. Inter- and intra-assay coefficients of variation were 2.2% and 0.9%, 1.3% and 0.9%, 3.5% and 0.97%, 1.3% and 0.9%, respectively for total cholesterol, HDL-cholesterol, triglycerides and LDL-cholesterol, respectively.
Serum human Fetuin A was evaluated using ELISA kit (TECOmedical Group, France) and was performed according to instructions provided by Epitope Diagnostics (intra-assay coefficient of variation: <5.5%, inter-assay coefficient of variation: <6.8%; detection limit of the assay: 5.0 ng/mL).
Serum human C-terminal FGF23 (Fibroblast Growth Factor 23) was determined using an ELISA kit (TECOmedical Group, France) and was performed according to the instructions provided by Immutopics (intra-assay coefficient of variation: <2.4%, inter-assay coefficient of variation: <4.7%; detection limit of the assay: 1.5 RU/mL).
IL-6 (Interleukine-6) was quantified using automated assays with Access 2 (Beckman Coulter Inc, Villepinte, France) according to the manufacturer’s instructions. IGF-1 (Insulin like Growth Factor 1), 25-hydroxyvitamin D, intact parathyroid hormone (iPTH) were measured by a single step chemiluminescence sandwich method on the Liaison XL (DiaSorin) analyzer. Because IGF-1 decreases with age, a standardized IGF-1 score was calculated as previously described [IGF-1 score = (log [IGF-1 (micrograms per liter)] + 0.00625 × age - 2.555)/0.104](21).
Assessment of insulin resistance
Insulin resistance was evaluated by the TyG index (triglycerides glucose index), which was calculated as ln(fasting triglycerides [mg/dL]*fasting glucose [mg/dL]/2)(22).
Imaging for below-knee arterial calcification score
Below-knee artery calcification score was obtained after scanning with a 128-slice multidetector dual source CT-scanner (Somatom Definition Flash, Siemens Healthineers Healthcare, Erlangen, Germany) without contrast, from the base of the patella down to the ankle. Three millimeter cross-sectional slices were analyzed. The analysis was performed by radiologists who were blinded to the results of color duplex ultrasonography, laboratory tests or clinical examination, using a commercially available software package (Heartbeat CaScore, Philips Healthcare, Eindhoven, Netherlands). On cross-sectional images, areas of calcification along below-knee arteries with a density of ≥130 Hounsfield units attenuation and a surface area of >1mm2 were identified semi automatically. Calcification score, determined according to the method described by Agatston et al., was obtained separately for each of the main below-knee arteries (distal popliteal, anterior tibial, posterior tibial and peroneal arteries) and then added up to obtain the total calcification score(23). Below-knee artery calcification scoring was performed at the inclusion visit and at the end of the study(4).
Statistical analysis
Quantitative variables were described by their mean, standard deviation, median and quartiles Q1-Q3. Qualitative variables were described by their frequency and percentage. The effect of baseline serum RANKL on the progression of calcification of leg arteries between baseline and the end of the follow-up was analyzed using univariable and multivariable linear regression model adjusted on arterial calcification score at baseline. This method is fully equivalent to an analysis of covariance (ANCOVA). Distributions of biological parameters were checked graphically and those with a log-normal distribution were subsequently transformed before any analysis to improve normality. This model was adjusted on baseline cardiovascular risk factors and other factors known to be associated with vascular calcification (age, gender, tobacco use, hypertension, waist circumference, BMI, triglycerides, total cholesterol, HDL-cholesterol, LDL-cholesterol, ApoA1, urinary albumin-creatinine ratio, HbA1c, lower limb log calcification score at baseline, hsCRP, parathormone, glomerular filtration rate (MDRD), and duration of follow up). Cook distance was used for the detection of highly influential observations on the coefficient estimates. A backward stepwise variable selection procedure based on the Akaike Information Criterion was used to select the final multivariate model. Coefficient estimates were provided with their corresponding 95% confidence intervals.
The effects of other bone remodeling factors, markers of inflammation and glycation (Calcium, iPTH, OPG, MGP, IGF-1, Fetuin A, FGF-23, hsCRP, IL-6, carboxyméthyllysine, MG-H1, pentosidine, RAGE) were evaluated with multivariate linear regression model using the same procedure.
Significance was defined as p-values of less than 0.05. Statistical analyses were performed using R 3.5.1 (http://www.R-project.org).
Sample size
It was assumed that the correlation between the logarithm of the serum RANKL level and the logarithm of the artery calcification score at 2 years, adjusted on the covariates, is about 0.20 (or an R² of 0.04). In addition, it was assumed that the multivariate linear model (based on RANKL and associated risk factors) will explain about 15% of the variability. Under these assumptions, at least 169 patients were needed to demonstrate a significant effect of RANKL on the calcification score with an alpha risk of 5% (bilateral formulation) and a power of 80%. In order to take into account 15% of patients lost to follow-up, 198 subjects were planned for the study.