The Malmö Diet and Cancer Cardiovascular (MDC-CV) cohort study
The MDCS is a large prospective cohort study with participants recruited from Malmö, a city in southern Sweden [15]. During 1991-1994, a random sample of 6,103 participants was taken from MDCS to investigate the epidemiology of carotid artery atherosclerosis (MDC-CV cohort study) [14]. Among them, 5,540 donated fasting blood samples.
We excluded participants with missing baseline data on eGFR, other covariates or GDF-15, or participants with previously diagnosed CKD or lost to follow-up. Therefore, 4,318 participants (Figure 1, mean aged 57.5 ± 5.95 years, male 39.4 %) remained for analyses of incident CKD, as detected by national registers of hospital inpatients and outpatients [16]. During 2007-2012, MDC-CV participants who were still alive and living in the Malmö area were invited to a re-examination. A total of 3,734 attended, which corresponds to 75.8% of the eligible population [17]. Among the 4318 individuals in this study, 2,827 attended re-examination and 2,744 had follow-up data available for eGFR. This sub-cohort study was then analyzed for decline in eGFR (Figure 1). Incident CKD based on eGFR was further analyzed as the outcome in 2,713 participants with baseline eGFR ≥60 mL/min/1.73 m2 (Figure 1). Written informed consent was obtained from all included participants. The study conformed to the Declaration of Helsinki and was approved by the ethical committee at Lund University, Lund, Sweden (LU 51/90).
GDF-15 measurement
Fasting blood samples were collected from the cubital vein and stored at −80°C until assay. GDF-15 levels were measured by the SciLifeLab analysis service (Uppsala, Sweden) using Proseek® Multiplex CVD I96×96 reagent kit where a Proximity Extension Assay technology was applied [14, 18]. Briefly, the assay procedure consisted of three key steps: incubation, extension and detection. Raw Proseek data went through a pre-processing normalization procedure and were set relative to a fixed background level, after which Normalized Protein Expression (log2 scale) values were generated, measured in arbitrary units (AU). High AU values corresponded to a high protein concentration. GDF-15 levels in 987 subjects measured by Proseek assay closely correlated (r =0.89 [L. Lind, unpublished data]) with the values by an electrochemiluminescence immunoassay (Roche Diagnostics, Mannheim, Germany) [14].
CKD based on the ICD codes from the national register
Information on CKD diagnosis was obtained from the Swedish patient register with nation-wide coverage. Moreover, the Swedish renal registry was searched for any additional cases of CKD [19]. CKD was defined as codes 585-586 according to ICD-9, and N18 and N19 according to ICD-10. All participants without any previous diagnosis of CKD were followed from baseline until the occurrence of a diagnosis of CKD (registry-based CKD), emigration from Sweden, death or December 31st, 2013, whichever came first.
The CKD diagnosis in the Swedish patient register has been previously described and validated [20]. Briefly, for validation, CKD diagnoses were evaluated by two experienced specialists in nephrology. Patient records and laboratory data were reviewed and CKD cases were defined following the 2012 KDIGO criteria [21]. Validation showed that 94% of patients had a correct diagnosis of CKD [20].
CKD based on eGFR, and eGFR decline from baseline to follow-up
The eGFR at baseline and follow-up was determined from a combination of plasma creatinine and cystatin C using the CKD-Epidemiology Collaboration 2012 equation [22]. Single measurements of eGFR were assessed at each time point. A cut-off value of 60 mL/min/1.73 m2 was used to identify participants with eGFR-based CKD [20]. The difference between these two measurements was defined as eGFR change.
At baseline, creatinine and cystatin C were analyzed with the Jaffé method (Beckman Synchron LX20-4; Beckman-Coulter) and with a particle-enhanced immunonephelometry assay (N Latex Cystatin; Dade Behring, Deerfield, IL), respectively. Since the world calibrator was not introduced until 2010, cystatin C values were not standardized (reference value: 0.53~0.95 mg/L). During 2007-2012, creatinine was determined in follow-up samples using an enzymatic method (Cobas autoanalyzer; Roche Diagnostics) calibrated by isotope-dilution mass spectrometry-traceable (IDMS) creatinine [23], and cystatin C was analyzed using an automated particle-based immunoassay, adjusted to the international reference preparation ERM-DA 71/IFCC.38 T [24]. Therefore, values of creatinine and cystatin C could not be directly compared between baseline and follow-up.
Other variables and definitions
Baseline characteristics were obtained from self-administered questionnaire, physical examination, and blood measurements. Data on medication, smoking habits and alcohol consumption were collected by questionnaires. Participants were classified into current smokers, former smokers and never smokers. An average daily alcohol consumption >40 g for males or >30 g for females was considered as high alcohol consumption. Waist circumference was determined as being midpoint between the end of the 12th rib and the iliac crest. Blood pressure was measured with a mercury-column sphygmomanometer after 10 min of rest while the subject was in a supine position. Participant with a history of coronary event or stroke was considered to have CVD at baseline.
Glucose concentration was measured in fresh whole blood samples after an overnight fasting, following standard procedures at the Department of Clinical Chemistry, University Hospital Malmö. Diabetes was defined as self-reported physician diagnosis of diabetes, use of anti-diabetic drugs or fasting whole blood glucose ≥6.1 mmol/L (corresponding to plasma glucose ≥7.0 mmol/L). Low density lipoprotein (LDL) concentration was estimated using the Friedewald’s formula. Measurements of biomarkers were conducted later using frozen (−80°C) plasma samples. C-reactive protein (CRP) was measured with a Tina-quant® CRP latex assay (Roche Diagnostics, Basel, Switzerland). Methods to measure N-terminal pro-B-type natriuretic peptide (NT-proBNP) levels was the same way as that for GDF-15 [14, 18].
Statistical analyses
Baseline characteristics are presented for participants divided into quartiles (Q1-Q4) according to GDF-15 concentration, using sex-specific quartile limits. For skewed variables, log-transformation was performed to achieve a normal distribution. Differences across GDF-15 quartiles were examined using analysis of variance for continuous variables and logistic regression analysis for categorized variables.
Cox proportional hazard regression was used to analyze the association between baseline GDF-15 and incident CKD discovered by the national register. Hazard ratios (HRs) and 95% confidential intervals (CIs) were obtained. GDF-15 was treated both as a continuous variable (per standard deviation (SD) change) and as a categorized variable (in quartiles). In multivariate-adjusted models, potential covariates taken into consideration were age, sex, waist circumference, smoking, high alcohol consumption, systolic blood pressure, LDL, CRP, diabetes, CVD, anti-hypertensive drug medication, and baseline eGFR. Since GDF-15 has been frequently considered as a cardiovascular biomarker in recent years, NT-proBNP, a traditional cardiovascular marker was additionally adjusted for in a sensitivity analysis to explore whether the association of GDF-15 with CKD could be mediated by cardiac function. A restricted cubic spline function was incorporated into the Cox model to test for possible non-linearity, with knots placed at 20, 40, 60 and 80 percentages of GDF-15 concentration. Possible effect modifications were examined by introducing an interaction term between GDF-15 levels and risk factors into the multivariate model one by one. The competing risks of death was accounted for in a sensitivity analysis by the Fine and Gray proportional subdistribution hazards models method. In another sensitivity analysis, the association between GDF-15 and CKD was analyzed while participants with baseline eGFR <60 mL/min/1.73 m2 were excluded. In addition, for participants with follow-up data available for eGFR, multiple linear regression was used to analyze the association between GDF-15 and eGFR change from baseline to the end of the follow-up. A multiple logistic regression analysis was conducted for the association between GDF-15 and eGFR-based CKD.
All analyses were performed using the Statistical Analysis System version 9.3 for Windows (SAS Institute Inc., Cary, NC, USA). A 2-tailed p <0.05 was considered statistically significant.