Retrospectively, we identified all consecutive patients of ≥ 18 years of age with a confirmed diagnosis of AF on VKAs attending outpatient cardiology consultations of a tertiary hospital between January 2011 and February 2013. Only patients who fulfilled the following criteria were included in this study: Patients with permanent or paroxysmal AF recently started on VKAs (i.e., not more than 8 mo passed since the beginning of their VKAs therapy), and who have regular visits for INR measurements. Patients with prosthetic valve (n = 452), rheumatic heart disease (n = 43), active cancer (n = 41), dementia (n = 26), and/or interrupted vitamin K antagonist > 3 d (n = 73) were excluded. Thus, the final analyzed cohort consisted of 911 patients. A detailed medical history was recorded for each patient, and the basal clinical characteristics at study entry together with information on follow up were carefully gathered by cardiologists.

The vast majority of patients were on acenocoumarol (93%; and the remaining patients were on warfarin).

The study was approved by the Clinical Research Ethics Committee of our hospital.

For each patient, Serum creatinine was measured by the modified kinetic Jaffe method in a single clinical laboratory in our institution. All creatinine measurements were performed with an isotope dilution mass spectroscopy (IDMS)-traceable enzymatic assay that has previously been shown to provide very reliable eGFR results compared with the measured GFR[13]; these measurements were analyzed automatically using the ADVIA 2400 Chemistry System (Siemens Diagnostics, Tarrytown, NY, United States).

We calculated the eGFR using the IDMS-traceable version of the MDRD-4 equation[8]: 175 × [standardized serum creatinine (mg/dL)]^-1.154× age^-0.203× (0.742 if female) × (1.212 if black).

The new CKD-EPI equation was also used[9]: 141 × (minimum of standardized serum creatinine (mg/dL)/κ or 1)^α× [maximum of standardized serum creatinine (mg/dL)/κ or 1]^-1.209× 0.993^age× (1.018 if female) × (1.159 if black). Where κ is 0.7 for females and 0.9 for males and α is -0.329 for females and -0.411 for males.

We categorized the eGFR obtained from each formula into three categories: ≥ 60 mL/min per 1.73 m² (normal or mild renal dysfunction), 30-59 mL/min per 1.73 m² (moderate renal dysfunction) and < 30 mL/min per 1.73 m² (severe renal dysfunction). No patients were on renal replacement therapy.

Patients were followed up to 1-year after the enrolment. The primary endpoint of the present study was a composite endpoint of major bleeding, TE complications, or death; whichever comes first. The secondary endpoint was each individual component of the composite endpoint.

Data on major bleeding, and TE complications were gathered from the cardiology clinic visits and records, and through hospital files as well as through primary care centers reports.

We used the 2005 International Society on Thrombosis and Haemostasis (ISTH) criteria to define major bleeding[14]. Thus, a major bleeding event was adjudicated if one of the following criteria was met: fatal bleeding and/or symptomatic bleeding in a critical area or organ (e.g., such as intracranial, intraspinal, intraocular, retroperitoneal, atraumatic intraarticular, pericardial, or intramuscular with compartment syndrome); and/or bleeding causing drop of hemoglobin of ≥ 2 g/dL, or leading to transfusion of ≥ 2 units of whole blood or packed red blood cells.

A TE complication was defined as the occurrence of ischemic stroke, transient ischemic attack, or peripheral embolism (including fatal TE events). Diagnosis of stroke or transient ischemic attack required an acute neurological deficit lasting for more or less than 24 h, respectively, which could not be explained by other causes and with at least 1 image test (computed tomography or magnetic resonance) compatible with the diagnosis, as well as confirmation from a neurologist. A diagnosis of peripheral embolism was defined as non-central nervous system embolism leading to an abrupt vascular insufficiency associated with clinical or radiographic evidence of arterial occlusion in absence of another mechanism such as atherosclerosis, instrumentation, or trauma.

Qualitative data were expressed as frequencies and percentages while quantitative data were summarized as mean and standard deviation. Comparison between qualitative data was performed using the χ² test or the Fisher exact test, as appropriate. The t-Student test was used to compare quantitative data.

The relationship between the primary endpoint and eGFR according to both formulas was evaluated using separate Cox proportional hazard regression models. The candidate variables to construct the multivariate Cox models were those variables presented P < 0.10 in the univariate Cox analysis, or those co-variables of recognized prognostic value in the medical literature. Once the initial Cox models had been established, they were simplified by stepdown elimination. Thus, the final Cox models to determine the adjusted effect of eGFR on the composite endpoint, included: age, sex, previous stroke, basal hemoglobin, chronic obstructive pulmonary disease, diabetes mellitus, congestive heart failure or left ventricular ejection fraction ≤ 40%, history of malignant disease and coronary artery disease.

The association between eGFR formulas and the individual endpoints of either major bleeding or TE events was determined using competing-risks regression based on Fine and Gray’s proportional subhazards models. The Fine and Gray models were adjusted for HAS-BLED score[4] in the case of testing the relationship between eGFR formulas and major bleeding, and for CHA₂DS₂-VASc score[15] in the case of testing the relationship between eGFR formulas and TE events. For all-cause mortality, we used a Cox regression model. Once the initial Cox model for predicting all-cause mortality had been established, it was simplified by stepdown elimination; and finally included the following covariables: age, sex, diabetes mellitus, and history of malignant disease, previous stroke, basal hemoglobin, and congestive heart failure or ejection fraction ≤ 40%.

The discriminatory capacity of each formula at predicting either the primary or secondary endpoint was determined by calculating the c- statistic. We used the Delong test to compare the c-statistic values from each formula.

The calibration of the model was assessed with the Grønnesby and Borgan goodness-of-fit test. This test determines how closely the predicted event rate approximates the observed event rate over a range of scores. A significant value of P indicates a lack of fit.

The estimated coefficients were expressed as the hazard ratio (HR) with the respective 95%CI. A 2-sided P < 0.05 was considered statistically significant for all analyses.

Finally, we also assessed the incremental prognostic value of using one equation over another; using the concept of net reclassification improvement (NRI) as described by Pencina et al[16], to determine whether the reclassification of patients by one of the formulas regarding to each other, would result in a more accurate risk estimation.

All the analyses were performed with STATA 13, and by using the MedCalc statistical software version 12.2.1.

The study was reviewed by our expert Biostatistic Emad Abu-Assi, MD, PhD.

The mean eGFR was higher when computed by the new CKD-EPI than with the re-expressed MDRD-4 (69.8 ± 23, 67.2 ± 19 mL/min per 1.73 m²), respectively (P < 0.0001 for comparison).

There was lower prevalence of eGFR < 60 mL/min per 1.73 m² with the new CKD-EPI than with the re-expressed MDRD-4 (32.9% vs 34.1%).

During a follow up of 10 ± 3 mo, the composite endpoint occurred in 98 (10.8%) patients: 30 (3.3%) patients developed major bleeding, 18 (2%) had TE events, and 60 (6.6%) patients died.

The rate of the composite endpoint increased monotonically from the higher to the lower eGFR categories for both formulas (Figure 1).

Open in New Tab   Full Size Figure   Download Figure

Figure 1 Distribution of major cardiovascular events according to the categories of estimated glomerular filtration rate using the re-expressed Modification of Diet in Renal Disease-4 and the new Chronic Kidney Disease Epidemiology Collaboration equations. eGFR: Estimated glomerular filtration rate; MDRD-4 indicates: Four variables Modification of Diet in Renal Disease; CKD-EPI: Chronic Kidney Disease Epidemiology Collaboration.

Significant association was observed between the eGFR using both formulas as continuous variables and the composite endpoint. The adjusted hazard ratios of eGFR by each formula on the composite endpoint were: 0.98 (95%CI: 0.967-0.988) and 0.97 (95%CI: 0.963-0.987) for the re-expressed MDRD-4 and the new CKD-EPI, respectively (Table 2).

Similarly, the eGFR as a categorical variable was a strong independent predictor of the occurrence of the composite endpoint regardless of the formula used (Table 3).

The discriminative capacity of both formulas at predicting the composite endpoint, were quite similar, regardless of the eGFR was used as continuous (0.683 vs 0.695 for the re-expressed MDRD-4 and the new CKD-EPI, respectively; P = 0.748) or categorical variable (0.632 vs 0.639 for the re-expressed MDRD-4 and the new CKD-EPI, respectively; P = 0.45) (Table 4).

There was a step increase in the major bleeding rate, as the eGFR declines, independently of the formula used to calculate the eGFR (Figure 1).

After adjusting for HAS-BLED bleeding risk score, the re-expressed MDRD-4 eGFR as well as the new CKD-EPI eGFR, as continuous variables, showed a tendency to predict major bleeding: HR for both formulas = 0.98 (95%CI: 0.965-1.000; P = 0.07) (Table 2).

No significant association was observed between categorical eGFR from both formulas and major bleeding, either in the unadjusted or by using the adjusted competing-risk models (Table 3).

At predicting major bleeding, the discriminative ability of the continuous re-expressed MDRD-4 eGFR was modest: 0.666; quite similar to that obtained from using the continuous new CKD-EPI eGFR: c-statistic = 0.677 (P = 0.85).

When eGFR was considered as a categorical variable, the discriminative capacity of each formula at predicting major bleeding was of 0.550 and of 0.571 for the re-expressed MDRD-4 and the new CKD-EPI, respectively (P = 0.79) (Table 4).

As shown in Figure 1, the distribution of the TE event rate in the different eGFR categories, demonstrated a consistent gradient of risk, regardless of the formula used.

After adjusting for the CHA₂DS₂-VASc risk score, no significant association was observed between eGFR as a continuous variable and TE events: HR = 0.98 (95%CI: 0.965-1.000) and 0.98 (95%CI: 0.965-1.001), for the re-expressed MDRD-4 and the new CKD-EPI, respectively (Table 2).

When eGFR was considered as a categorical variable, only significant association existed between eGFR < 30 mL/min per 1.73 m² and the TE complications, after controlling for CHA₂DS₂-VASc score: HR = 5.1 (95%CI: 1.04-25.4) for the re-expressed MDRD-4, and HR = 5.0 (95%CI: 1.0-24.9) for the new CKD-EPI (Table 3).

The discriminative power of GFR estimates determined by both formulas was also modest. For continuous eGFR, the c-statistic values were of 0.616 and 0.644 for the re-expressed MDRD-4 and the new CKD-EPI, respectively, (P = 0.27), and for categorical eGFR, the c-statistic values were 0.617 and 0.622 when using the re-expressed MDRD-4 and the new CKD-EPI, respectively, (P = 0.76) (Table 4).

The rate of all-cause mortality increased progressively from the higher to the lower eGFR values for both formulas (Figure 1).

Continuous eGFR calculated by either the reexpressed MDRD-4 or the new CKD-EPI was an independent predictor of all-cause mortality; adjusted HR = 0.98; (P < 0.01) (Table 2).

A strong association was also found between categorical eGFR and all-cause mortality after adjusting for several confounders (Table 3).

Good discrimination was obtained from continuous eGFR: c-statistic = 0.715 for the re-expressed MDRD-4 and 0.722 for the new CKD-EPI (P = 0.52).

The discriminative power of eGFR as a categorical variable in terms of c-statistic was: 0.679 and 0.678 when using the re-expressed MDRD-4 and the new CKD-EPI, respectively, (P = 0.91) (Table 4).

Estimated GFR from both formulas demonstrated good calibration for the major cardiovascular events with P value > 0.1 (Table 4).

The NRI analysis did not significantly favor the new CKD-EPI over the re-expressed MDRD-4 whether for predicting the composite endpoint, major bleeding and all-cause mortality (NRI = 2.13%, 4.35%, and 0.9%, with P = 0.27, 0.19, and 0.7, respectively).

However, at predicting the TE event, the NRI favored the new CKD-EPI formula with NRI of 1% (95%CI: -0.08 to +2.0, P = 0.07) indicating a strong tendency to reclassify better the patients according to their risk of developing TE event, compared with the re-expressed MDRD-4.

The main limitation of our study is its retrospective design, but it has interesting strong points as it reflects real world practice by enrolment of consecutive patients with NVAF who have full range of eGFR and were attending our outpatient cardiology clinics with the advantage of careful follow up and data collection by cardiologists.

The sample size might be another limitation of our study that could limit the likelihood of detecting small effects or significant relationships from the data. Important to mention here that we did not have the direct measured GFR, so we cannot determine the extent to which the two formulas reflect the GFR as determined by the gold standard method. However, eGFR is the practical way to estimate renal function which has been used in several patient populations. The fact that we have only one serum creatinine measure for every patient could limit the verification of the acute vs chronic nature of the renal dysfunction in some patients, but this limitation was present in several related studies[23-25]. The lack of cystatin C data might be considered a limitation of our study. However, it should be taken into account that all the creatinine measurements in our study cohort were performed with the IDMS-traceable enzymatic assay method, which has been shown to provide very reliable eGFR results[13] and is considered the standard method to assess renal function[29].

Finally, all of the enrolled patients in our cohort have Caucasian race, so the applicability of our findings in other populations with different races should be addressed in other studies.

The new CKD-EPI reduced the prevalence of patients with renal dysfunction, in a real world cohort of patients with NVAF on VKAs. Renal dysfunction reflected by GFR estimates from the re-expressed MDRD-4 or the new CKD-EPI was an independent predictor of the composite endpoint and all-cause mortality. Both formulas had similar prognostic impacts regarding the prediction of composite endpoint, major bleeding, TE events and all-cause mortality. Our analysis indicates that the more widespread adoption of the new CKD-EPI instead of the re-expressed MDRD-4 may improve the management of patients with NVAF.

Renal dysfunction is a frequent comorbidity seen in patients with atrial fibrillation. Moreover, renal dysfunction is a strong predictor of thromboembolic event and also of bleeding event (when the patients are anticoagulated). This reflects the need for more accurate estimate of renal function to guarantee the optimal management of patients with atrial fibrillation. The standard way to assess renal function is the glomerular filtration rate. Among the available equations to estimate the glomerular filtration rate are: the re-expressed Modification of Diet in Renal Disease equation which is still the commonly used equation by many laboratories all over the world and the new Chronic Kidney Disease Epidemiology Collaboration equation which has been recently proposed to be used instead of previous equation in daily practice as the new equation has an assumed ability to reduce the prevalence of patients with renal dysfunction and better reclassification of patients. There is limited information about the performance of both equations in patients with atrial fibrillation.

The authors think that the new Chronic Kidney Disease Epidemiology Collaboration equation to estimate glomerular filtration rate must have a wide diffusion as an alternative to the re-expressed Modification of Diet in Renal Disease equation. In this paper the authors provide support to the hypothesis, reporting the superiority of the new Chronic Kidney Disease Epidemiology Collaboration equation over the re-expressed Modification of Diet in Renal Disease equation in the clinical context of patients with atrial fibrillation on anticoagulation.

The results derived from our analysis, state that the new Chronic Kidney Disease Epidemiology Collaboration equation reduced the prevalence of patients with renal dysfunction (i.e., estimated glomerular filtration rate < 60 ml/min per 1.73 m²), and at the same time continued to have the prognostic impact similar to the re-expressed Modification of Diet in Renal Disease equation at predicting the major adverse events. Although there are still some concerns about the performance of the new equation in subgroups of elderly and obese patients, the study from a real world cohort encourages the cardiologists to use of the new Chronic Kidney Disease Epidemiology Collaboration equation to assess renal function in patients with atrial fibrillation and increase the confidence to use it in all clinical situations.

The millions of patients with atrial fibrillation will get benefit and better management if there is wide spread adoption of the new Chronic Kidney Disease Epidemiology Collaboration equation instead of the re-expressed Modification of Diet in Renal Disease equation, giving the ability of the new equation to correctly reclassify patients in comparison with the re-expressed equation.

The Chronic Kidney Disease Epidemiology Collaboration equation (CKD-EPI) was published in May 2009 as a reliable tool to estimate glomerular filtration rate. It was developed in an effort to create an equation more accurate than the re-expressed Modification of Diet in Renal Disease equation. Researchers pooled data from multiple studies to develop and validate this new equation. They used 10 studies that included 8254 participants, randomly using 2/3 of the data sets for development and the other 1/3 for internal validation. Sixteen additional studies, which included 3896 participants, were used for external validation. The CKD-EPI equation performed better than the Modification of Diet in Renal Disease equation, as the prevalence of chronic kidney disease was 11.5% vs 13.1% according to the National Health and Nutrition Examination Survey data in the United States of America.

First of all I would like to congratulate the authors with their achievement. In this retrospective study including relatively limited sample size of Caucasian subjects, the findings encourage the use and application of the new CKD-EPI equation for assessment not only of renal function in patients with non-valvular atrial fibrillation but also in all clinical situations. For the first time, Abumuaileq RRY et al evaluated the re- expressed MDRD-4 and the new CKD-EPI formulas at predicting the occurrence of major adverse outcomes in a real world cohort of patients with non-valvular atrial fibrillation on anticoagulation. The study was well conducted and clinically relevant.