C-terminal fragment of agrin (CAF) levels predict acute kidney injury after acute myocardial infarction

The study design of this prospective observational cohort study has been previously published [16]. In short consecutive patients with acute ST-elevation MI (STEMI) or non ST-elevation MI (NSTEMI) were recruited if they fulfilled the following inclusion criteria: 1) age ≥ 18 years; 2) ability to provide written, informed consent; and 3) acute, spontaneous (type 1) AMI. The main exclusion criteria were presence of pre-existing renal disease and AMI-related symptom onset ≥72 h from hospital admission. Patients referred for urgent coronary artery bypass grafting and those suffering a fatal event during the index hospitalization were also excluded from the study.

Consecutive patients admitted to the Coronary Care Unit from 3 different Cardiology Departments (Alexandroupolis; Kavala; Athens) in Greece were recruited from July 2010 to May 2014 [16]. In 436 patients, baseline plasma and urine samples were available for additional assessment of CAF concentrations and 403 patients had a valid result for all plasma and urine biomarkers. STEMI patients underwent either primary percutaneous coronary intervention (PCI), or fibrinolysis followed by rescue or elective PCI as indicated [17, 18]. NSTEMI patients underwent invasive or primarily conservative (followed by elective PCI) strategy according to current recommendations [18, 19]. Patients were assessed for presence of AKI at 48 h post admission using the Acute Kidney Injury Network (AKIN) [20] and the Acute Dialysis Quality Initiative [Risk, Injury and Failure (RIFLE)] criteria [21] and also at hospital discharge using the RIFLE criteria and the Kidney Disease: Improving Global Outcomes (KDIGO) criteria [22]. Changes in absolute or relative changes in creatinine, or eGFR were presumed to have occurred within hospitalization. In addition, in each patient a transthoracic echocardiography study was performed during hospitalization using standard techniques [23]. Death from any or cardiovascular causes, repeat hospitalization and need for dialysis or other renal replacement therapy was assessed with telephone contacts during the 2-years follow up.

Blood and urine sampling for CAF, FE-Na, NGAL, IL-18 and cyst-C assessment was performed at hospital admission and before any critical intervention according to previously established analytical protocols. Peripheral blood samples for measurement of blood chemistry (renal function) and full blood count were obtained from all patients on admission, 48 h after the index event, and also daily until discharge. Blood samples were drawn from a peripheral vein of the patients in vacutainer tubes containing EDTA as an anticoagulant. Samples were immediately centrifuged at 4000 rpm for 10 min at ambient temperature, and the extracted plasma was stored in aliquots and frozen at −70 °C until use. Urine samples were collected in aliquots from spot urine during admission and were immediately frozen at -70 °C until use. [24]. The coefficients of variation for intra- and inter-assay precision were <8% and <10% respectively, for all assays. CAF levels were measured using the NTtotalCAF ELISA kit from Neurotune [14]. The lower detection limit was 40 pM. Data for plasma were valid when the coefficient of variation (%CV) was below 20%. Independent analysis revealed a mixed intra- and inter-assay %CV below 12.6% for plasma samples.

The study was approved by the institutional Ethics Committee, and all subjects gave written informed consent at the time of enrollment. The CAF sample kits measurements were performed blindly by Neurotune AG Schweiz, which had no role in data analysis.

Data are presented as percentages for categorical data, as means ± standard deviation (SD) for continuous variables that were normally distributed and as medians with interquartile range (IQR) for non-normally distributed data. Comparisons between categorical variables were performed by chi- square test or Fisher’s exact test when required. Differences in continuous variables between two groups were assessed using the Student’s t-test or the Mann-Whitney’s U-test as appropriate.

The diagnostic accuracy of the under investigation biomarkers was determined by calculating the area under the curve (AUC), sensitivity, specificity, and positive and negative predictive values using receiver operating characteristic (ROC) curve analysis. The association between CAF levels and incidence of AKI during hospitalization was assessed by univariable logistic regression analyses. For both assessments serum creatinine on admission were considered as baseline. Independence of the association was assessed in multivariable logistic regression analysis models arbitrarily using variables that could act as possible cofounders for prognosis (age, diabetes mellitus, presence of anemia, glomerular filtration rate on admission, albuminuria, LV EF, invasive or not treatment strategy of the index event, presence of any adverse event during the index hospitalization and infarct size based on CPK maximum levels). For all (incidence or prognosis) logistic regression analysis models, odds ratios (OR) with 95% confidence intervals were calculated.

Baseline demographic, clinical, angiographic and laboratory characteristics of the cohort and in AKI versus non-AKI patents according to RIFLE-Criteria are listed in Table 1 . The majority of the patients were managed invasively during hospitalization and one fourth of the population experienced at least one in-hospital adverse event.

The incidence of AKI in our study population ranged from 7% to 15% (Additional file 1: Table S1) depending on timing (at 48 h vs. during hospitalization) and on definition used (AKIN vs. RIFLE vs. KDIGO). The majority of the patients had stage 1 kidney injury whereas none of the patients required dialysis during hospitalization. For further analysis, patients were considered to have AKI using the KDIGO or RIFLE criteria during hospitalization.

CAF concentrations in both mediums were significantly correlated with creatinine levels on admission (urine; Spearman’s rho 0.233, P < 0.001, plasma; Spearman’s rho 0.175, P < 0.001). This significant association was also observed with creatinine levels 48 h post admission (urine; Spearman’s rho 0.263, P < 0.001, plasma; Spearman’s rho 0.226, P < 0.001) and with peak creatinine levels (urine; Spearman’s rho 0.317, P < 0.001, plasma Spearman’s rho 0.225, P < 0.001) during hospitalization. The observed incidence of AKI increased across quartiles of urine CAF (Cochrane-Armitage test for trend with increasing quartile values; chi-square for trend, 10.99; P < 0.001). The association between increasing levels of urine CAF and the predicted incidence of AKI suggested a linear effect (r2 = 0.983). Finally, CAF concentrations were also significantly associated with Cyst-C levels both in plasma (Spearman’s rho 0.292, P < 0.001) and in urine (Spearman’s rho 0.267, P < 0.001).

From the under investigation variables only urine NGAL, urine (u) and plasma (p) CAF were capable of detecting AKI. The discriminating ability (regarding the incidence of AKI during hospitalization) of the urine concentration of the parameters under investigation ranged from good to moderate whereas from plasma levels only plasma CAF had a moderate discriminating ability for AKI occurrence (Table 2). In comparison, urinary CAF had a similar diagnostic accuracy for AKI as assessed with ROC analysis compared to urine NGAL (p = 0.73), and plasma CAF (p = 0.38) (Fig. 1). Indexing these markers to BMI, BSA, and urine creatinine or using their urine to plasma ratio did not improve their diagnostic accuracy (Additional file 1: Table S2).

Concerning diagnostic accuracy, ROC analysis identified a value of 1033 pM as optimal in predicting development of AKI. The sensitivity of urinary CAF was 37% (95%CI 25-51%) and the specificity 85% (95%CI 81-89) with a negative predictive value of 89% (95%CI 85-92%) and a positive predictive value of 30% (95%CI 20-42%). Moreover, the urinary CAF cut-off was associated with a positive likelihood ratio (+LR) of 2.52 (95% CI 1.7 -3.8) and a negative ratio (−LR) of 0.7 (95% CI,0.6 -0.9). Applying Bayes’ theorem, if we consider 15% as the pre-test probability for developing AKI, the post-test probability for developing AKI, when urinary CAF levels are ≥1033 pM, is doubled to 30% (95% CI, 22-40). Similarly, the post-test probability for developing AKI, when the urinary CAF concentrations are <1033 pM, is only 11% (95% CI, 9-13). Applying the Bayes theorem in terms of number needed to diagnose using the cut-off value of 1033 pM, 1 in 3 positive tests are truly predictive of the disease whilst 1 in 1 negative tests are truly non-predictive of the disease.

Levels of urinary CAF were associated with AKI incidence in a univariate model (OR per 1 SD, 1.45 95%CI 1.15-1.82, P = 0.002). The observed association was more robust compared to those observed with plasma CAF levels (OR per 1 SD, 1.41 95%CI 1.11-1.79, P = 0.005) and urine NGAL levels (OR per 1 SD, 1.15 95%CI 0.93-1.42, P = 0.2).

Levels of urinary CAF continued to be associated with AKI incidence in a multivariate model (Table 3) that included multiple variables (age, diabetes mellitus, presence of anemia, glomerular filtration rate on admission, albuminuria, LV EF, invasive or not treatment strategy of the index event, presence of any adverse event during the index hospitalization and infarct size based on CPK maximum levels) with an OR 1.35 95%CI 1.05 -1.74. Of interest, neither urine NGAL (P = 0.62), nor plasma CAF (P = 0.24) concentrations were independent predictors of AKI in multivariable models. The addition of urinary CAF levels in predictive models including each one of the 3 comparator biomarkers resulted in increased discriminating ability. Furthermore, the addition of urinary CAF levels to a predictive model including all 3 biomarkers (urine NGAL and plasma CAF) showed a better predictive performance. However, the additive value of urinary CAF levels was marginally non-significant. (Additional file 1: Table S3)

During the 2-year follow up, 33 deaths were observed among the whole study population. Twenty-six deaths were attributed to cardiovascular causes whereas the rest 7 were attributed to other causes mainly cancer. 57 patients required at least one repeat hospitalization during the follow up period, whilst 15 patients developed deterioration in their kidney function that required an inpatient intervention or outpatient follow up. Of interest only 2 out of the 15 patients required permanent renal replacement therapy.

Only plasma CAF levels were associated with death from any cause (OR 4.1 95%CI 1.7-9.7; P = 0.001) whereas urinary CAF concentration were not associated (P = 0.84). Multivariable analysis showed that plasma CAF levels remained a significant independent predictor of mortality (OR 2.5 95%CI 1.02-6.2; P = 0.04). Neither plasma CAF (P = 0.08) nor urinary CAF (P = 0.8) were predictive of deterioration of kidney function albeit plasma CAF levels were marginally non-significant.