p-Cresol and Cardiovascular Risk in Mild-to-Moderate Kidney Disease

Patients

Prevalent chronic kidney disease patients, followed at the CKD outpatient clinic of the University Hospital Gasthuisberg, 18 years or older and able to provide consent, were eligible for inclusion. Patients were screened between November 2005 and September 2006. The study was performed according to the Declaration of Helsinki and approved by the ethics committee of the University Hospital Leuven. Informed consent was obtained from all patients. The trial was prospectively registered at clinicaltrials.gov (NCT00441623).

Baseline Evaluation and Biochemical Measurements

Data on baseline demographics, smoking habit, presence of diabetes, prevalent cardiovascular disease, and cause of kidney disease were collected at time of informed consent. At inclusion, blood was taken by venous puncture for measurement of creatinine (mg/dl), hemoglobin (g/dl), biointact parathormone (PTH) (ng/L), calcium (mg/dl), phosphate (mg/dl), albumin (g/dl), C-reactive protein (CRP) (mg/L), cholesterol (mg/dl), and total and free p-cresol (mg/L).

Creatinine, hemoglobin, PTH, calcium, phosphate, CRP, and cholesterol were all measured using standard laboratory techniques. The eGFR was calculated using the four-variable MDRD (Modification of Diet in Renal Disease study) equation (15). Albumin was measured using the bromcresol green method. Total and free (not bound to proteins) p-cresol serum concentrations were measured using a gas chromatography-mass spectrometry (GC-MS) method on serum stored at −80°C until batch analysis. After heat-acid denaturation of binding proteins, p-cresol was extracted in ethyl acetate and injected on the GC-MS (Trace GC-MS; Thermo Finnigan, San José, CA). p-Cresol-d8 was used as internal standard. Free p-cresol concentrations were measured in serum ultrafiltered at 37°C using 30,000 D molecular cutoff filters (Centifree UF devices; Amicon, Beverly, MA). We and others have recently demonstrated that in vivo p-cresol is almost entirely sulfated (16,17). Both p-cresol and its conjugated metabolites are measured by this analytical technique (17). We demonstrated good agreement between the GC-MS–based method for indirect quantification of p-cresyl sulfate after hydrolysis to p-cresol and a HPLC method for direct quantification of total serum concentrations (18). We performed additional method comparison for unbound p-cresol/p-cresyl sulfate (Figure 1). We chose to use the GC-MS–based method because of its lower limit of quantification. Values below the limit of quantification (0.02 mg/L; 0.18 μM) were treated as 0.09 μM for statistical analysis. Intra- and interassay coefficients of variation were 3.3% and 5.3%, respectively.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Method comparison between indirect quantification of unbound p-cresyl sulfate concentrations after hydrolysis to p-cresol by gas chromatography-mass spectrometry and direct quantification of unbound p-cresyl sulfate concentrations by HPLC. Deming fit demonstrated excellent agreement. As expected, we observed a nonsignificant (P = 0.1) small proportional bias (y = 0.00 + 0.91 · x). Indeed, besides p-cresyl sulfate, small amounts of p-cresyl glucuronide and unconjugated p-cresol are quantified by the indirect GC-MS method.

End Point Evaluation

Patients were followed at the nephrology CKD outpatient clinic at 3- to 6-month intervals. Patients were prospectively followed until December 1, 2008. If the patient's last study visit was before October 2008 or a scheduled visit was missed, patients and/or their general practitioner were contacted by phone to ensure completeness of the study data. If information could not be obtained, the patient was assumed to be lost to follow-up starting from the date of the last actual visit. Cardiovascular end points were prospectively recorded and coded, blinded from clinical and biochemical data. The primary end point (first cardiovascular event) was a composite of death from cardiac causes, nonlethal myocardial infarction, myocardial ischemia, coronary intervention, ischemic stroke, or new peripheral vascular disease, whichever occurred first. Only one event per subject was included in the analysis.

After review of available information, cause of death was classified as either cardiovascular, infectious, malignancy, or other. Cardiovascular deaths included fatal myocardial infarction, sudden death, and death due to congestive heart failure. Cases of unobserved sudden death were considered cardiovascular death only when other potential causes could be excluded. Otherwise, they were classified as “other cause of death.” Out-of-hospital deaths were coded after consultation of the general practitioner. Nonlethal cardiovascular events included myocardial infarction, diagnosed based on elevated levels of cardiac enzymes and/or typical electrocardiography changes, myocardial ischemia with typical electrocardiography changes without elevated cardiac enzymes, and coronary intervention (thrombolysis, percutaneous coronary intervention, or coronary artery bypass grafting). Ischemic stroke was defined as a neurologic deficit lasting more than 24 hours. Hemorrhagic stroke was excluded from the primary end point. Peripheral vascular disease included new-onset ischemic pain in the lower limbs, with abnormal ankle brachial pressure index or radiologic evidence of peripheral vascular disease, new-onset ischemic necrotic lesions, or surgical arterial intervention.

Statistical Analysis

Continuous variables are expressed as mean (SD) for normally distributed variables or median (interquartile range) otherwise. Differences between baseline variables according to free p-cresol tertiles were analyzed by the Kruskal-Wallis test. Associations between p-cresol free serum concentrations and other variables were analyzed using a Spearman rank correlation matrix. The Kaplan-Meier method was used to estimate cumulative incidence of the primary end point. Time to first cardiovascular event analysis was performed using Cox proportional hazards analysis. The relative risk of a new cardiovascular event was expressed as a hazard ratio. A two-sided P < 0.05, unadjusted for multiple comparisons, was considered statistically significant. Several models were built, including different sets of covariates, to adjust for demographics (age and gender), medical history (coronary artery disease and peripheral artery disease), presence of diabetes, CKD markers [hemoglobin, calcium, phosphate, and PTH (Ln) concentrations], markers of inflammation [C-reactive protein (Ln) concentrations], markers of nutrition (body mass index and albumin concentrations), Framingham risk factors (age, gender, systolic BP, smoking, diabetes, and serum cholesterol concentrations), and medical therapy (use of converting enzyme inhibitors/Angiotensin receptor blockers, phosphate binders, statin intake, and 25-hydroxy vitamin D supplements). For each model, the proportionality assumption was tested against log(time). In case the proportionality assumption was violated, follow-up time-adjusted variables were entered as covariates. Owing to relatively low event number, candidate covariates for the final model were selected from these models (hierarchical elimination approach). Variables with P < 0.2 in individual models were retained. For sensitivity testing, additional best-fit models were constructed using backward elimination (P_exclude > 0.05) and stepwise selection (P_include < 0.20, P_exclude > 0.05). Two different approaches were used to select covariates. In a first approach we included all covariates associated at the P = 0.2 level with new cardiovascular events on univariate analysis. Using this approach, we included age, gender, history of coronary artery disease, history of peripheral vascular disease, diabetes, eGFR, hemoglobin, calcium, phosphate, CRP, albumin, systolic BP, and free p-cresol concentrations. As the number of covariates is sizeable, we performed backward elimination at P ≥ 0.2 followed by backward elimination at P ≥ 0.05 to reduce the risk of overfitting. In a second approach we included the most widely accepted cluster of covariates used to estimate cardiovascular disease risk. This approach yielded age, gender, diabetes, smoking habit, systolic BP, and serum cholesterol as covariates. To study the potential confounding effect of kidney function, eGFR was entered both as covariate and as stratifying variable. As eGFR is modeled from age and gender, creatinine was entered as a traditional marker of kidney function in multivariate models where age and gender were included. For statistics, SAS (version 9.1; the SAS Institute, Cary, NC) and SPSS (version 15.0; Chicago, IL) software packages were used. The authors had full access to the data and take responsibility for its integrity. All authors have read and agree with the manuscript as written.