Serum uromodulin and progression of kidney disease in patients with chronic kidney disease

The C-STRIDE is a multicenter prospective cohort of CKD patients, containing 39 clinical centers in different geographic regions of China. The criteria for the enrollment of participants are listed in Additional file 1. CKD participants have been enrolled from November 2011, a total of 3499 patients have completed screening until 30 June 2016, of which 686 patients were excluded due to missing values of serum creatinine and/or loss of follow-up data, altogether 2813 patients have the completed baseline and follow-up data. Due to the availability of the biosamples for measuring serum uromodulin, therefore, 2652 patients were included in the present study. For the etiologic diagnosis, there were 1707 patients with glomerular diseases, 547 patients with tubulointerstitial diseases and 398 patients with other or unknown causes. The design of C-STRIDE has been described elsewhere in detail [18].

We measured serum uromodulin levels of 2652 CKD patients at baseline and described the distribution of baseline data of these patients according to serum uromodulin levels. Then, we further investigated the associations of serum uromodulin with pre-specified end-points of CKD patients, including ESKD, cardiovascular events and all-cause mortality. The baseline data included detailed demographics, underlying disease, behavioral habits, medical and medication history, anthropometric measures (height, weight, resting blood pressure), chemistry indexes (triglyceride, total cholesterol, low-density lipoprotein cholesterol, high-density lipoprotein cholesterol, fasting blood glucose, prealbumin, serum creatinine, high-sensitivity C-reactive protein), and urinary albumin/creatinine ratio. The end-point events of CKD patients were collected before 30 June 2017. All these patients gave written informed consent before data collection. The study was approved by the Ethics Committee of Peking University First Hospital and was in adherence with the Declaration of Helsinki.

Fasting venous blood samples were obtained at the study visit. All participants’ blood samples were transported by cold chain to the Central Laboratory of Peking University First Hospital and stored at − 80 °C until use. Serum uromodulin was measured in batches from stored material in the Central Laboratory of Peking University First Hospital from May 2016 to August 2016. We measured serum uromodulin by a commercially available ELISA kit (Euroimmun AG, Lübeck, Germany) according to the manufacturer’s instructions. The process of the assay was described in Additional file 1.

The covariates in the multivariable models were either potential confounders, or the independent predictors of the adverse outcomes. All blood and urine samples were analyzed in the Central Laboratory of Peking University First Hospital to avoid the variation of testing values between laboratories. Serum total, low-density lipoprotein, and high-density lipoprotein cholesterol, triglycerides were measured with commercially available reagents. Urinary albumin and creatinine were measured from a fresh morning spot urine sample or morning urine sample stored at 4 °C for less than 1 week. Albuminuria was measured with immunoturbidimetric tests. Urinary creatinine was measured with the ammonia iminohydrolase method. The urinary albumin to creatinine ratio (mg/g creatinine) was calculated. Serum creatinine was measured by the same methods as urinary creatinine. The eGFR was evaluated by the equation developed by adaptation of the Modification of Diet in Renal Disease (MDRD) equation on the basis of data from the Chinese CKD participants: eGFR = 175 × (serum creatinine [in μmol/L]/88.4)^−1.234 × age^−0.179 × (if female × 0.79) [19]. All eGFR values of more than 120 mL/min/1.73 m² were set at 120 mL/min/1.73 m².

Body mass index was calculated by using the following formula: weight (kg)/height² (m²). Blood pressure was measured three times at 5-min intervals by a sphygmomanometer. The mean value of the three readings was calculated. The use of anti-hypertensive medications in the past 2 weeks before baseline examination was recorded. Diabetes was defined as the fasting plasma glucose of 7.0 mmol/L or more, or the use of hypoglycemic agents or a self-reported history of diabetes.

The pre-specified end-point events of the CKD patients included ESKD, cardiovascular event, and all-cause mortality. ESKD is defined as the initiation of chronic dialysis or renal kidney transplantation or irreversible development of eGFR < 15 mL/min/1.73 m². All the ESKD events included in the current study have initiated hemodialysis, or peritoneal dialysis, or kidney transplantation. Cardiovascular events included acute myocardial infarction, unstable angina, hospitalization for congestive heart failure, cerebrovascular events (intraparenchymal hemorrhage, subarachnoid hemorrhage, cerebral infarction, etc.), and peripheral vascular diseases. Echocardiogram and electrocardiogram examination were requested in the validation of cardiovascular events; however, cardio-angiography was not routinely performed. Only one event per patient was included in the current analysis. Finally, we separately evaluated the associations of baseline serum uromodulin with different clinical outcomes of CKD patients, including ESKD, cardiovascular events and all-cause mortality.

The end-point events were collected at 3- to 6-month intervals until 30 June 2017 in the current analysis. The director at each clinical center asked for end-point events according to the new-onset events registration form either by phone calls or routine clinical visits. Once the end-point events occurred, the director at the clinical center filled out the new-onset events registration form and submitted the related clinical data to the Renal Institute of Peking University via email within 1 month. The suspected clinical outcomes were then adjudicated by an independent committee consisting of specialist physicians.

All CKD patients were stratified according to tertiles of baseline serum uromodulin levels. Continuous variables are presented as the means and standard deviations, except for highly skewed variables that are shown as median and interquartile ranges (IQR), and categorical variables are presented as proportions. One-way ANOVA was used to compare continuous variables, and Chi squared tests were used to compare categorical variables.

The incidence rates of end-point events (including ESKD, cardiovascular event and all-cause mortality) were calculated as number of events per 100 person-years. We depicted cumulative hazard function for the three events separately according to uromodulin levels by using a Kaplan–Meier curve and compared the event rates by using log-rank test.

To investigate the association between serum uromodulin and outcomes, Cox proportional hazards regression models were used to estimate hazards ratios (HRs) and 95% confidence intervals (CIs). We found that the risk of ESKD was linearly increased through the decline of uromodulin as shown in the linear spline analysis (Additional file 1: Figure S1), so we treated uromodulin either as a categorical variable (using the highest tertile as the reference) or a continuous variable (per standard deviation change) to represent the exposure variable. Multivariable models were constructed to adjust for potential confounding variables of the adverse outcomes, including age (continuous), gender (male vs. female), body-mass index (continuous), current smoker (yes vs. no), previous history of cardiovascular disease (yes vs. no), systolic blood pressure (continuous), using anti-hypertensive medications in the past 2 weeks (yes vs. no), diabetes (yes vs. no), prealbumin (continuous), logarithm transformed low-density lipoprotein cholesterol, high-density lipoprotein cholesterol [20], triglyceride, high-sensitivity C-reactive protein, urinary albumin/creatinine ratio (all in continuous), and eGFR (continuous). The missing values were filled before they were entered in the regression model. The proportional hazards assumption was assessed via Kaplan–Meier curves using log–log plots. We fitted logistic regression model by using the same covariates in the Cox regression model and calculated area under receiver operating characteristic curve (AUC). We compared AUCs inclusion or exclusion of serum uromodulin in the model with traditional cardiovascular risk factors, eGFR and ACR, in order to evaluate the change in discriminating ability for ESKD after inclusion of serum uromodulin. Statistical analyses were performed using the SAS software (version 9.4, SAS institute, CA, USA). P < 0.05 (two-sided) was considered statistically significant.

Among the 2652 CKD patients included in our study, 1548 (58.4%) were male, and 1104 (41.6%) were female, with an age of 48.7 ± 13.8 years. The baseline eGFR was 49.6 ± 29.4 mL/min/1.73 m². A total of 788 (29.7%) participants had an eGFR greater than 60 mL/min/1.73 m² at baseline. Altogether, 1053 (39.7%) and 811 (30.6%) patients were in CKD stages 3 and 4, respectively.

The median level of serum uromodulin was 77.2 ng/mL (IQR 48.3–125.9 ng/mL) in 2652 CKD patients. The levels in the three etiologic types of CKD were 100.4 ± 65.6 ng/mL in glomerular diseases, 69.0 ± 40.6 ng/mL in tubulointerstitial diseases and 73.4 ± 51.4 ng/mL in other or unknown causes, respectively (P < 0.001). The baseline characteristics of CKD patients according to tertiles of the serum uromodulin levels are presented in Table 1. Compared with patients with higher serum uromodulin levels, those with lower uromodulin levels were older, had a higher proportion of previous history of smoking, diabetes, and cardiovascular disease, higher proportion of current use of antihypertensive medications, and had higher levels of blood pressure, triglyceride, high-sensitivity C-reactive protein and urinary albumin/creatinine ratio but lower levels of total cholesterol and eGFR. The patients with missing value in the clinical characteristics tended to have a lower uromodulin level in our study. For example, the uromodulin levels were 79.20 ng/mL and 92.19 ng/mL, respectively, among those with and without missing value of systolic blood pressure (P < 0.001). In addition, we found a positive correlation between serum uromodulin and eGFR in multivariable linear correlation analysis (r = 0.68, P < 0.001).

The incidence rates of end-point events according to levels of serum uromodulin are shown in Table 2. During the median follow-up of 53.6 (IQR 44.0–64.0) months, there were 404 ESKD, 189 cardiovascular events and 69 deaths occurred. ESKD, cardiovascular events and death rates were 3.60, 1.60 and 0.57 per 100 person-years, respectively. Higher incidence rates of all the three end-point events were seen with the decreased levels of uromodulin (Figs. 1, 2, 3, all P-values for log-rank test < 0.05).

The association of serum uromodulin with outcomes is shown in Table 3. After adjusting for demographic and traditional cardiovascular risk factors, as well as the baseline eGFR levels, baseline serum uromodulin levels were independently associated with the risk of incident ESKD, with an HR of 1.92 (95% CI 1.26–2.90) in the bottom tertile compared with the top tertile. Every standard deviation increase of uromodulin was associated with a decreased risk of ESKD, with an HR of 0.69 (95% CI 0.55–0.86). However, we did not detect significant associations between serum uromodulin and the risk of cardiovascular events as well as all-cause mortality in multivariable adjusted model. Similar results were found among subgroup of patients with glomerular diseases and tubulointerstitial diseases (Additional file 1: Tables S1, S2).

In the fully adjusted logistic regression model for ESKD, the AUCs with inclusion or exclusion of serum uromodulin level were 0.8623 (95% CI 0.8438–0.8808) and 0.8601 (95% CI 0.8413–0.8789), respectively. The difference of the AUCs was 0.0023 (95% CI − 0.0007 to 0.0052) (P-value for the difference = 0.1).